git/convert.c

2044 lines
50 KiB
C
Raw Normal View History

Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
#include "cache.h"
#include "advice.h"
#include "config.h"
#include "convert.h"
#include "copy.h"
#include "gettext.h"
#include "hex.h"
#include "object-store.h"
#include "attr.h"
#include "run-command.h"
#include "quote.h"
#include "sigchain.h"
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
#include "pkt-line.h"
#include "sub-process.h"
#include "trace.h"
#include "utf8.h"
#include "ll-merge.h"
#include "wrapper.h"
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
/*
* convert.c - convert a file when checking it out and checking it in.
*
* This should use the pathname to decide on whether it wants to do some
* more interesting conversions (automatic gzip/unzip, general format
* conversions etc etc), but by default it just does automatic CRLF<->LF
* translation when the "text" attribute or "auto_crlf" option is set.
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
*/
/* Stat bits: When BIN is set, the txt bits are unset */
#define CONVERT_STAT_BITS_TXT_LF 0x1
#define CONVERT_STAT_BITS_TXT_CRLF 0x2
#define CONVERT_STAT_BITS_BIN 0x4
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
struct text_stat {
/* NUL, CR, LF and CRLF counts */
unsigned nul, lonecr, lonelf, crlf;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
/* These are just approximations! */
unsigned printable, nonprintable;
};
static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
{
unsigned long i;
memset(stats, 0, sizeof(*stats));
for (i = 0; i < size; i++) {
unsigned char c = buf[i];
if (c == '\r') {
if (i+1 < size && buf[i+1] == '\n') {
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
stats->crlf++;
i++;
} else
stats->lonecr++;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
continue;
}
if (c == '\n') {
stats->lonelf++;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
continue;
}
if (c == 127)
/* DEL */
stats->nonprintable++;
else if (c < 32) {
switch (c) {
/* BS, HT, ESC and FF */
case '\b': case '\t': case '\033': case '\014':
stats->printable++;
break;
case 0:
stats->nul++;
/* fall through */
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
default:
stats->nonprintable++;
}
}
else
stats->printable++;
}
/* If file ends with EOF then don't count this EOF as non-printable. */
if (size >= 1 && buf[size-1] == '\032')
stats->nonprintable--;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
}
/*
* The same heuristics as diff.c::mmfile_is_binary()
* We treat files with bare CR as binary
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
*/
static int convert_is_binary(const struct text_stat *stats)
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
{
if (stats->lonecr)
return 1;
if (stats->nul)
return 1;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
if ((stats->printable >> 7) < stats->nonprintable)
return 1;
return 0;
}
static unsigned int gather_convert_stats(const char *data, unsigned long size)
{
struct text_stat stats;
int ret = 0;
if (!data || !size)
return 0;
gather_stats(data, size, &stats);
if (convert_is_binary(&stats))
ret |= CONVERT_STAT_BITS_BIN;
if (stats.crlf)
ret |= CONVERT_STAT_BITS_TXT_CRLF;
if (stats.lonelf)
ret |= CONVERT_STAT_BITS_TXT_LF;
return ret;
}
static const char *gather_convert_stats_ascii(const char *data, unsigned long size)
{
unsigned int convert_stats = gather_convert_stats(data, size);
if (convert_stats & CONVERT_STAT_BITS_BIN)
return "-text";
switch (convert_stats) {
case CONVERT_STAT_BITS_TXT_LF:
return "lf";
case CONVERT_STAT_BITS_TXT_CRLF:
return "crlf";
case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF:
return "mixed";
default:
return "none";
}
}
const char *get_cached_convert_stats_ascii(struct index_state *istate,
const char *path)
{
const char *ret;
unsigned long sz;
void *data = read_blob_data_from_index(istate, path, &sz);
ret = gather_convert_stats_ascii(data, sz);
free(data);
return ret;
}
const char *get_wt_convert_stats_ascii(const char *path)
{
const char *ret = "";
struct strbuf sb = STRBUF_INIT;
if (strbuf_read_file(&sb, path, 0) >= 0)
ret = gather_convert_stats_ascii(sb.buf, sb.len);
strbuf_release(&sb);
return ret;
}
static int text_eol_is_crlf(void)
{
if (auto_crlf == AUTO_CRLF_TRUE)
return 1;
else if (auto_crlf == AUTO_CRLF_INPUT)
return 0;
if (core_eol == EOL_CRLF)
return 1;
if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF)
return 1;
return 0;
}
static enum eol output_eol(enum convert_crlf_action crlf_action)
{
switch (crlf_action) {
case CRLF_BINARY:
return EOL_UNSET;
case CRLF_TEXT_CRLF:
return EOL_CRLF;
case CRLF_TEXT_INPUT:
return EOL_LF;
case CRLF_UNDEFINED:
case CRLF_AUTO_CRLF:
return EOL_CRLF;
case CRLF_AUTO_INPUT:
return EOL_LF;
case CRLF_TEXT:
case CRLF_AUTO:
/* fall through */
return text_eol_is_crlf() ? EOL_CRLF : EOL_LF;
}
warning(_("illegal crlf_action %d"), (int)crlf_action);
return core_eol;
}
static void check_global_conv_flags_eol(const char *path,
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
struct text_stat *old_stats, struct text_stat *new_stats,
int conv_flags)
safecrlf: Add mechanism to warn about irreversible crlf conversions CRLF conversion bears a slight chance of corrupting data. autocrlf=true will convert CRLF to LF during commit and LF to CRLF during checkout. A file that contains a mixture of LF and CRLF before the commit cannot be recreated by git. For text files this is the right thing to do: it corrects line endings such that we have only LF line endings in the repository. But for binary files that are accidentally classified as text the conversion can corrupt data. If you recognize such corruption early you can easily fix it by setting the conversion type explicitly in .gitattributes. Right after committing you still have the original file in your work tree and this file is not yet corrupted. You can explicitly tell git that this file is binary and git will handle the file appropriately. Unfortunately, the desired effect of cleaning up text files with mixed line endings and the undesired effect of corrupting binary files cannot be distinguished. In both cases CRLFs are removed in an irreversible way. For text files this is the right thing to do because CRLFs are line endings, while for binary files converting CRLFs corrupts data. This patch adds a mechanism that can either warn the user about an irreversible conversion or can even refuse to convert. The mechanism is controlled by the variable core.safecrlf, with the following values: - false: disable safecrlf mechanism - warn: warn about irreversible conversions - true: refuse irreversible conversions The default is to warn. Users are only affected by this default if core.autocrlf is set. But the current default of git is to leave core.autocrlf unset, so users will not see warnings unless they deliberately chose to activate the autocrlf mechanism. The safecrlf mechanism's details depend on the git command. The general principles when safecrlf is active (not false) are: - we warn/error out if files in the work tree can modified in an irreversible way without giving the user a chance to backup the original file. - for read-only operations that do not modify files in the work tree we do not not print annoying warnings. There are exceptions. Even though... - "git add" itself does not touch the files in the work tree, the next checkout would, so the safety triggers; - "git apply" to update a text file with a patch does touch the files in the work tree, but the operation is about text files and CRLF conversion is about fixing the line ending inconsistencies, so the safety does not trigger; - "git diff" itself does not touch the files in the work tree, it is often run to inspect the changes you intend to next "git add". To catch potential problems early, safety triggers. The concept of a safety check was originally proposed in a similar way by Linus Torvalds. Thanks to Dimitry Potapov for insisting on getting the naked LF/autocrlf=true case right. Signed-off-by: Steffen Prohaska <prohaska@zib.de>
2008-02-06 11:25:58 +00:00
{
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
if (old_stats->crlf && !new_stats->crlf ) {
safecrlf: Add mechanism to warn about irreversible crlf conversions CRLF conversion bears a slight chance of corrupting data. autocrlf=true will convert CRLF to LF during commit and LF to CRLF during checkout. A file that contains a mixture of LF and CRLF before the commit cannot be recreated by git. For text files this is the right thing to do: it corrects line endings such that we have only LF line endings in the repository. But for binary files that are accidentally classified as text the conversion can corrupt data. If you recognize such corruption early you can easily fix it by setting the conversion type explicitly in .gitattributes. Right after committing you still have the original file in your work tree and this file is not yet corrupted. You can explicitly tell git that this file is binary and git will handle the file appropriately. Unfortunately, the desired effect of cleaning up text files with mixed line endings and the undesired effect of corrupting binary files cannot be distinguished. In both cases CRLFs are removed in an irreversible way. For text files this is the right thing to do because CRLFs are line endings, while for binary files converting CRLFs corrupts data. This patch adds a mechanism that can either warn the user about an irreversible conversion or can even refuse to convert. The mechanism is controlled by the variable core.safecrlf, with the following values: - false: disable safecrlf mechanism - warn: warn about irreversible conversions - true: refuse irreversible conversions The default is to warn. Users are only affected by this default if core.autocrlf is set. But the current default of git is to leave core.autocrlf unset, so users will not see warnings unless they deliberately chose to activate the autocrlf mechanism. The safecrlf mechanism's details depend on the git command. The general principles when safecrlf is active (not false) are: - we warn/error out if files in the work tree can modified in an irreversible way without giving the user a chance to backup the original file. - for read-only operations that do not modify files in the work tree we do not not print annoying warnings. There are exceptions. Even though... - "git add" itself does not touch the files in the work tree, the next checkout would, so the safety triggers; - "git apply" to update a text file with a patch does touch the files in the work tree, but the operation is about text files and CRLF conversion is about fixing the line ending inconsistencies, so the safety does not trigger; - "git diff" itself does not touch the files in the work tree, it is often run to inspect the changes you intend to next "git add". To catch potential problems early, safety triggers. The concept of a safety check was originally proposed in a similar way by Linus Torvalds. Thanks to Dimitry Potapov for insisting on getting the naked LF/autocrlf=true case right. Signed-off-by: Steffen Prohaska <prohaska@zib.de>
2008-02-06 11:25:58 +00:00
/*
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
* CRLFs would not be restored by checkout
safecrlf: Add mechanism to warn about irreversible crlf conversions CRLF conversion bears a slight chance of corrupting data. autocrlf=true will convert CRLF to LF during commit and LF to CRLF during checkout. A file that contains a mixture of LF and CRLF before the commit cannot be recreated by git. For text files this is the right thing to do: it corrects line endings such that we have only LF line endings in the repository. But for binary files that are accidentally classified as text the conversion can corrupt data. If you recognize such corruption early you can easily fix it by setting the conversion type explicitly in .gitattributes. Right after committing you still have the original file in your work tree and this file is not yet corrupted. You can explicitly tell git that this file is binary and git will handle the file appropriately. Unfortunately, the desired effect of cleaning up text files with mixed line endings and the undesired effect of corrupting binary files cannot be distinguished. In both cases CRLFs are removed in an irreversible way. For text files this is the right thing to do because CRLFs are line endings, while for binary files converting CRLFs corrupts data. This patch adds a mechanism that can either warn the user about an irreversible conversion or can even refuse to convert. The mechanism is controlled by the variable core.safecrlf, with the following values: - false: disable safecrlf mechanism - warn: warn about irreversible conversions - true: refuse irreversible conversions The default is to warn. Users are only affected by this default if core.autocrlf is set. But the current default of git is to leave core.autocrlf unset, so users will not see warnings unless they deliberately chose to activate the autocrlf mechanism. The safecrlf mechanism's details depend on the git command. The general principles when safecrlf is active (not false) are: - we warn/error out if files in the work tree can modified in an irreversible way without giving the user a chance to backup the original file. - for read-only operations that do not modify files in the work tree we do not not print annoying warnings. There are exceptions. Even though... - "git add" itself does not touch the files in the work tree, the next checkout would, so the safety triggers; - "git apply" to update a text file with a patch does touch the files in the work tree, but the operation is about text files and CRLF conversion is about fixing the line ending inconsistencies, so the safety does not trigger; - "git diff" itself does not touch the files in the work tree, it is often run to inspect the changes you intend to next "git add". To catch potential problems early, safety triggers. The concept of a safety check was originally proposed in a similar way by Linus Torvalds. Thanks to Dimitry Potapov for insisting on getting the naked LF/autocrlf=true case right. Signed-off-by: Steffen Prohaska <prohaska@zib.de>
2008-02-06 11:25:58 +00:00
*/
if (conv_flags & CONV_EOL_RNDTRP_DIE)
die(_("CRLF would be replaced by LF in %s"), path);
else if (conv_flags & CONV_EOL_RNDTRP_WARN)
warning(_("in the working copy of '%s', CRLF will be"
" replaced by LF the next time Git touches"
" it"), path);
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
} else if (old_stats->lonelf && !new_stats->lonelf ) {
safecrlf: Add mechanism to warn about irreversible crlf conversions CRLF conversion bears a slight chance of corrupting data. autocrlf=true will convert CRLF to LF during commit and LF to CRLF during checkout. A file that contains a mixture of LF and CRLF before the commit cannot be recreated by git. For text files this is the right thing to do: it corrects line endings such that we have only LF line endings in the repository. But for binary files that are accidentally classified as text the conversion can corrupt data. If you recognize such corruption early you can easily fix it by setting the conversion type explicitly in .gitattributes. Right after committing you still have the original file in your work tree and this file is not yet corrupted. You can explicitly tell git that this file is binary and git will handle the file appropriately. Unfortunately, the desired effect of cleaning up text files with mixed line endings and the undesired effect of corrupting binary files cannot be distinguished. In both cases CRLFs are removed in an irreversible way. For text files this is the right thing to do because CRLFs are line endings, while for binary files converting CRLFs corrupts data. This patch adds a mechanism that can either warn the user about an irreversible conversion or can even refuse to convert. The mechanism is controlled by the variable core.safecrlf, with the following values: - false: disable safecrlf mechanism - warn: warn about irreversible conversions - true: refuse irreversible conversions The default is to warn. Users are only affected by this default if core.autocrlf is set. But the current default of git is to leave core.autocrlf unset, so users will not see warnings unless they deliberately chose to activate the autocrlf mechanism. The safecrlf mechanism's details depend on the git command. The general principles when safecrlf is active (not false) are: - we warn/error out if files in the work tree can modified in an irreversible way without giving the user a chance to backup the original file. - for read-only operations that do not modify files in the work tree we do not not print annoying warnings. There are exceptions. Even though... - "git add" itself does not touch the files in the work tree, the next checkout would, so the safety triggers; - "git apply" to update a text file with a patch does touch the files in the work tree, but the operation is about text files and CRLF conversion is about fixing the line ending inconsistencies, so the safety does not trigger; - "git diff" itself does not touch the files in the work tree, it is often run to inspect the changes you intend to next "git add". To catch potential problems early, safety triggers. The concept of a safety check was originally proposed in a similar way by Linus Torvalds. Thanks to Dimitry Potapov for insisting on getting the naked LF/autocrlf=true case right. Signed-off-by: Steffen Prohaska <prohaska@zib.de>
2008-02-06 11:25:58 +00:00
/*
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
* CRLFs would be added by checkout
safecrlf: Add mechanism to warn about irreversible crlf conversions CRLF conversion bears a slight chance of corrupting data. autocrlf=true will convert CRLF to LF during commit and LF to CRLF during checkout. A file that contains a mixture of LF and CRLF before the commit cannot be recreated by git. For text files this is the right thing to do: it corrects line endings such that we have only LF line endings in the repository. But for binary files that are accidentally classified as text the conversion can corrupt data. If you recognize such corruption early you can easily fix it by setting the conversion type explicitly in .gitattributes. Right after committing you still have the original file in your work tree and this file is not yet corrupted. You can explicitly tell git that this file is binary and git will handle the file appropriately. Unfortunately, the desired effect of cleaning up text files with mixed line endings and the undesired effect of corrupting binary files cannot be distinguished. In both cases CRLFs are removed in an irreversible way. For text files this is the right thing to do because CRLFs are line endings, while for binary files converting CRLFs corrupts data. This patch adds a mechanism that can either warn the user about an irreversible conversion or can even refuse to convert. The mechanism is controlled by the variable core.safecrlf, with the following values: - false: disable safecrlf mechanism - warn: warn about irreversible conversions - true: refuse irreversible conversions The default is to warn. Users are only affected by this default if core.autocrlf is set. But the current default of git is to leave core.autocrlf unset, so users will not see warnings unless they deliberately chose to activate the autocrlf mechanism. The safecrlf mechanism's details depend on the git command. The general principles when safecrlf is active (not false) are: - we warn/error out if files in the work tree can modified in an irreversible way without giving the user a chance to backup the original file. - for read-only operations that do not modify files in the work tree we do not not print annoying warnings. There are exceptions. Even though... - "git add" itself does not touch the files in the work tree, the next checkout would, so the safety triggers; - "git apply" to update a text file with a patch does touch the files in the work tree, but the operation is about text files and CRLF conversion is about fixing the line ending inconsistencies, so the safety does not trigger; - "git diff" itself does not touch the files in the work tree, it is often run to inspect the changes you intend to next "git add". To catch potential problems early, safety triggers. The concept of a safety check was originally proposed in a similar way by Linus Torvalds. Thanks to Dimitry Potapov for insisting on getting the naked LF/autocrlf=true case right. Signed-off-by: Steffen Prohaska <prohaska@zib.de>
2008-02-06 11:25:58 +00:00
*/
if (conv_flags & CONV_EOL_RNDTRP_DIE)
die(_("LF would be replaced by CRLF in %s"), path);
else if (conv_flags & CONV_EOL_RNDTRP_WARN)
warning(_("in the working copy of '%s', LF will be"
" replaced by CRLF the next time Git touches"
" it"), path);
safecrlf: Add mechanism to warn about irreversible crlf conversions CRLF conversion bears a slight chance of corrupting data. autocrlf=true will convert CRLF to LF during commit and LF to CRLF during checkout. A file that contains a mixture of LF and CRLF before the commit cannot be recreated by git. For text files this is the right thing to do: it corrects line endings such that we have only LF line endings in the repository. But for binary files that are accidentally classified as text the conversion can corrupt data. If you recognize such corruption early you can easily fix it by setting the conversion type explicitly in .gitattributes. Right after committing you still have the original file in your work tree and this file is not yet corrupted. You can explicitly tell git that this file is binary and git will handle the file appropriately. Unfortunately, the desired effect of cleaning up text files with mixed line endings and the undesired effect of corrupting binary files cannot be distinguished. In both cases CRLFs are removed in an irreversible way. For text files this is the right thing to do because CRLFs are line endings, while for binary files converting CRLFs corrupts data. This patch adds a mechanism that can either warn the user about an irreversible conversion or can even refuse to convert. The mechanism is controlled by the variable core.safecrlf, with the following values: - false: disable safecrlf mechanism - warn: warn about irreversible conversions - true: refuse irreversible conversions The default is to warn. Users are only affected by this default if core.autocrlf is set. But the current default of git is to leave core.autocrlf unset, so users will not see warnings unless they deliberately chose to activate the autocrlf mechanism. The safecrlf mechanism's details depend on the git command. The general principles when safecrlf is active (not false) are: - we warn/error out if files in the work tree can modified in an irreversible way without giving the user a chance to backup the original file. - for read-only operations that do not modify files in the work tree we do not not print annoying warnings. There are exceptions. Even though... - "git add" itself does not touch the files in the work tree, the next checkout would, so the safety triggers; - "git apply" to update a text file with a patch does touch the files in the work tree, but the operation is about text files and CRLF conversion is about fixing the line ending inconsistencies, so the safety does not trigger; - "git diff" itself does not touch the files in the work tree, it is often run to inspect the changes you intend to next "git add". To catch potential problems early, safety triggers. The concept of a safety check was originally proposed in a similar way by Linus Torvalds. Thanks to Dimitry Potapov for insisting on getting the naked LF/autocrlf=true case right. Signed-off-by: Steffen Prohaska <prohaska@zib.de>
2008-02-06 11:25:58 +00:00
}
}
static int has_crlf_in_index(struct index_state *istate, const char *path)
autocrlf: Make it work also for un-normalized repositories Previously, autocrlf would only work well for normalized repositories. Any text files that contained CRLF in the repository would cause problems, and would be modified when handled with core.autocrlf set. Change autocrlf to not do any conversions to files that in the repository already contain a CR. git with autocrlf set will never create such a file, or change a LF only file to contain CRs, so the (new) assumption is that if a file contains a CR, it is intentional, and autocrlf should not change that. The following sequence should now always be a NOP even with autocrlf set (assuming a clean working directory): git checkout <something> touch * git add -A . (will add nothing) git commit (nothing to commit) Previously this would break for any text file containing a CR. Some of you may have been folowing Eyvind's excellent thread about trying to make end-of-line translation in git a bit smoother. I decided to attack the problem from a different angle: Is it possible to make autocrlf behave non-destructively for all the previous problem cases? Stealing the problem from Eyvind's initial mail (paraphrased and summarized a bit): 1. Setting autocrlf globally is a pain since autocrlf does not work well with CRLF in the repo 2. Setting it in individual repos is hard since you do it "too late" (the clone will get it wrong) 3. If someone checks in a file with CRLF later, you get into problems again 4. If a repository once has contained CRLF, you can't tell autocrlf at which commit everything is sane again 5. autocrlf does needless work if you know that all your users want the same EOL style. I belive that this patch makes autocrlf a safe (and good) default setting for Windows, and this solves problems 1-4 (it solves 2 by being set by default, which is early enough for clone). I implemented it by looking for CR charactes in the index, and aborting any conversion attempt if this is found. Signed-off-by: Finn Arne Gangstad <finag@pvv.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2010-05-11 22:37:57 +00:00
{
unsigned long sz;
void *data;
const char *crp;
int has_crlf = 0;
autocrlf: Make it work also for un-normalized repositories Previously, autocrlf would only work well for normalized repositories. Any text files that contained CRLF in the repository would cause problems, and would be modified when handled with core.autocrlf set. Change autocrlf to not do any conversions to files that in the repository already contain a CR. git with autocrlf set will never create such a file, or change a LF only file to contain CRs, so the (new) assumption is that if a file contains a CR, it is intentional, and autocrlf should not change that. The following sequence should now always be a NOP even with autocrlf set (assuming a clean working directory): git checkout <something> touch * git add -A . (will add nothing) git commit (nothing to commit) Previously this would break for any text file containing a CR. Some of you may have been folowing Eyvind's excellent thread about trying to make end-of-line translation in git a bit smoother. I decided to attack the problem from a different angle: Is it possible to make autocrlf behave non-destructively for all the previous problem cases? Stealing the problem from Eyvind's initial mail (paraphrased and summarized a bit): 1. Setting autocrlf globally is a pain since autocrlf does not work well with CRLF in the repo 2. Setting it in individual repos is hard since you do it "too late" (the clone will get it wrong) 3. If someone checks in a file with CRLF later, you get into problems again 4. If a repository once has contained CRLF, you can't tell autocrlf at which commit everything is sane again 5. autocrlf does needless work if you know that all your users want the same EOL style. I belive that this patch makes autocrlf a safe (and good) default setting for Windows, and this solves problems 1-4 (it solves 2 by being set by default, which is early enough for clone). I implemented it by looking for CR charactes in the index, and aborting any conversion attempt if this is found. Signed-off-by: Finn Arne Gangstad <finag@pvv.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2010-05-11 22:37:57 +00:00
data = read_blob_data_from_index(istate, path, &sz);
if (!data)
autocrlf: Make it work also for un-normalized repositories Previously, autocrlf would only work well for normalized repositories. Any text files that contained CRLF in the repository would cause problems, and would be modified when handled with core.autocrlf set. Change autocrlf to not do any conversions to files that in the repository already contain a CR. git with autocrlf set will never create such a file, or change a LF only file to contain CRs, so the (new) assumption is that if a file contains a CR, it is intentional, and autocrlf should not change that. The following sequence should now always be a NOP even with autocrlf set (assuming a clean working directory): git checkout <something> touch * git add -A . (will add nothing) git commit (nothing to commit) Previously this would break for any text file containing a CR. Some of you may have been folowing Eyvind's excellent thread about trying to make end-of-line translation in git a bit smoother. I decided to attack the problem from a different angle: Is it possible to make autocrlf behave non-destructively for all the previous problem cases? Stealing the problem from Eyvind's initial mail (paraphrased and summarized a bit): 1. Setting autocrlf globally is a pain since autocrlf does not work well with CRLF in the repo 2. Setting it in individual repos is hard since you do it "too late" (the clone will get it wrong) 3. If someone checks in a file with CRLF later, you get into problems again 4. If a repository once has contained CRLF, you can't tell autocrlf at which commit everything is sane again 5. autocrlf does needless work if you know that all your users want the same EOL style. I belive that this patch makes autocrlf a safe (and good) default setting for Windows, and this solves problems 1-4 (it solves 2 by being set by default, which is early enough for clone). I implemented it by looking for CR charactes in the index, and aborting any conversion attempt if this is found. Signed-off-by: Finn Arne Gangstad <finag@pvv.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2010-05-11 22:37:57 +00:00
return 0;
crp = memchr(data, '\r', sz);
if (crp) {
unsigned int ret_stats;
ret_stats = gather_convert_stats(data, sz);
if (!(ret_stats & CONVERT_STAT_BITS_BIN) &&
(ret_stats & CONVERT_STAT_BITS_TXT_CRLF))
has_crlf = 1;
}
autocrlf: Make it work also for un-normalized repositories Previously, autocrlf would only work well for normalized repositories. Any text files that contained CRLF in the repository would cause problems, and would be modified when handled with core.autocrlf set. Change autocrlf to not do any conversions to files that in the repository already contain a CR. git with autocrlf set will never create such a file, or change a LF only file to contain CRs, so the (new) assumption is that if a file contains a CR, it is intentional, and autocrlf should not change that. The following sequence should now always be a NOP even with autocrlf set (assuming a clean working directory): git checkout <something> touch * git add -A . (will add nothing) git commit (nothing to commit) Previously this would break for any text file containing a CR. Some of you may have been folowing Eyvind's excellent thread about trying to make end-of-line translation in git a bit smoother. I decided to attack the problem from a different angle: Is it possible to make autocrlf behave non-destructively for all the previous problem cases? Stealing the problem from Eyvind's initial mail (paraphrased and summarized a bit): 1. Setting autocrlf globally is a pain since autocrlf does not work well with CRLF in the repo 2. Setting it in individual repos is hard since you do it "too late" (the clone will get it wrong) 3. If someone checks in a file with CRLF later, you get into problems again 4. If a repository once has contained CRLF, you can't tell autocrlf at which commit everything is sane again 5. autocrlf does needless work if you know that all your users want the same EOL style. I belive that this patch makes autocrlf a safe (and good) default setting for Windows, and this solves problems 1-4 (it solves 2 by being set by default, which is early enough for clone). I implemented it by looking for CR charactes in the index, and aborting any conversion attempt if this is found. Signed-off-by: Finn Arne Gangstad <finag@pvv.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2010-05-11 22:37:57 +00:00
free(data);
return has_crlf;
autocrlf: Make it work also for un-normalized repositories Previously, autocrlf would only work well for normalized repositories. Any text files that contained CRLF in the repository would cause problems, and would be modified when handled with core.autocrlf set. Change autocrlf to not do any conversions to files that in the repository already contain a CR. git with autocrlf set will never create such a file, or change a LF only file to contain CRs, so the (new) assumption is that if a file contains a CR, it is intentional, and autocrlf should not change that. The following sequence should now always be a NOP even with autocrlf set (assuming a clean working directory): git checkout <something> touch * git add -A . (will add nothing) git commit (nothing to commit) Previously this would break for any text file containing a CR. Some of you may have been folowing Eyvind's excellent thread about trying to make end-of-line translation in git a bit smoother. I decided to attack the problem from a different angle: Is it possible to make autocrlf behave non-destructively for all the previous problem cases? Stealing the problem from Eyvind's initial mail (paraphrased and summarized a bit): 1. Setting autocrlf globally is a pain since autocrlf does not work well with CRLF in the repo 2. Setting it in individual repos is hard since you do it "too late" (the clone will get it wrong) 3. If someone checks in a file with CRLF later, you get into problems again 4. If a repository once has contained CRLF, you can't tell autocrlf at which commit everything is sane again 5. autocrlf does needless work if you know that all your users want the same EOL style. I belive that this patch makes autocrlf a safe (and good) default setting for Windows, and this solves problems 1-4 (it solves 2 by being set by default, which is early enough for clone). I implemented it by looking for CR charactes in the index, and aborting any conversion attempt if this is found. Signed-off-by: Finn Arne Gangstad <finag@pvv.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2010-05-11 22:37:57 +00:00
}
static int will_convert_lf_to_crlf(struct text_stat *stats,
enum convert_crlf_action crlf_action)
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
{
if (output_eol(crlf_action) != EOL_CRLF)
return 0;
/* No "naked" LF? Nothing to convert, regardless. */
if (!stats->lonelf)
return 0;
if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
/* If we have any CR or CRLF line endings, we do not touch it */
/* This is the new safer autocrlf-handling */
if (stats->lonecr || stats->crlf)
return 0;
if (convert_is_binary(stats))
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
return 0;
}
return 1;
}
static int validate_encoding(const char *path, const char *enc,
const char *data, size_t len, int die_on_error)
{
const char *stripped;
/* We only check for UTF here as UTF?? can be an alias for UTF-?? */
if (skip_iprefix(enc, "UTF", &stripped)) {
skip_prefix(stripped, "-", &stripped);
/*
* Check for detectable errors in UTF encodings
*/
if (has_prohibited_utf_bom(enc, data, len)) {
const char *error_msg = _(
"BOM is prohibited in '%s' if encoded as %s");
/*
* This advice is shown for UTF-??BE and UTF-??LE encodings.
* We cut off the last two characters of the encoding name
* to generate the encoding name suitable for BOMs.
*/
const char *advise_msg = _(
"The file '%s' contains a byte order "
"mark (BOM). Please use UTF-%.*s as "
"working-tree-encoding.");
int stripped_len = strlen(stripped) - strlen("BE");
advise(advise_msg, path, stripped_len, stripped);
if (die_on_error)
die(error_msg, path, enc);
else {
return error(error_msg, path, enc);
}
} else if (is_missing_required_utf_bom(enc, data, len)) {
const char *error_msg = _(
"BOM is required in '%s' if encoded as %s");
const char *advise_msg = _(
"The file '%s' is missing a byte order "
"mark (BOM). Please use UTF-%sBE or UTF-%sLE "
"(depending on the byte order) as "
"working-tree-encoding.");
advise(advise_msg, path, stripped, stripped);
if (die_on_error)
die(error_msg, path, enc);
else {
return error(error_msg, path, enc);
}
}
}
return 0;
}
static void trace_encoding(const char *context, const char *path,
const char *encoding, const char *buf, size_t len)
{
static struct trace_key coe = TRACE_KEY_INIT(WORKING_TREE_ENCODING);
struct strbuf trace = STRBUF_INIT;
int i;
strbuf_addf(&trace, "%s (%s, considered %s):\n", context, path, encoding);
for (i = 0; i < len && buf; ++i) {
strbuf_addf(
&trace, "| \033[2m%2i:\033[0m %2x \033[2m%c\033[0m%c",
i,
(unsigned char) buf[i],
(buf[i] > 32 && buf[i] < 127 ? buf[i] : ' '),
((i+1) % 8 && (i+1) < len ? ' ' : '\n')
);
}
strbuf_addchars(&trace, '\n', 1);
trace_strbuf(&coe, &trace);
strbuf_release(&trace);
}
static int check_roundtrip(const char *enc_name)
{
/*
* check_roundtrip_encoding contains a string of comma and/or
* space separated encodings (eg. "UTF-16, ASCII, CP1125").
* Search for the given encoding in that string.
*/
const char *found = strcasestr(check_roundtrip_encoding, enc_name);
const char *next;
int len;
if (!found)
return 0;
next = found + strlen(enc_name);
len = strlen(check_roundtrip_encoding);
return (found && (
/*
* check that the found encoding is at the
* beginning of check_roundtrip_encoding or
* that it is prefixed with a space or comma
*/
found == check_roundtrip_encoding || (
(isspace(found[-1]) || found[-1] == ',')
)
) && (
/*
* check that the found encoding is at the
* end of check_roundtrip_encoding or
* that it is suffixed with a space or comma
*/
next == check_roundtrip_encoding + len || (
next < check_roundtrip_encoding + len &&
(isspace(next[0]) || next[0] == ',')
)
));
}
static const char *default_encoding = "UTF-8";
static int encode_to_git(const char *path, const char *src, size_t src_len,
struct strbuf *buf, const char *enc, int conv_flags)
{
char *dst;
size_t dst_len;
int die_on_error = conv_flags & CONV_WRITE_OBJECT;
/*
* No encoding is specified or there is nothing to encode.
* Tell the caller that the content was not modified.
*/
if (!enc || (src && !src_len))
return 0;
/*
* Looks like we got called from "would_convert_to_git()".
* This means Git wants to know if it would encode (= modify!)
* the content. Let's answer with "yes", since an encoding was
* specified.
*/
if (!buf && !src)
return 1;
if (validate_encoding(path, enc, src, src_len, die_on_error))
return 0;
trace_encoding("source", path, enc, src, src_len);
dst = reencode_string_len(src, src_len, default_encoding, enc,
&dst_len);
if (!dst) {
/*
* We could add the blob "as-is" to Git. However, on checkout
* we would try to re-encode to the original encoding. This
* would fail and we would leave the user with a messed-up
* working tree. Let's try to avoid this by screaming loud.
*/
const char* msg = _("failed to encode '%s' from %s to %s");
if (die_on_error)
die(msg, path, enc, default_encoding);
else {
error(msg, path, enc, default_encoding);
return 0;
}
}
trace_encoding("destination", path, default_encoding, dst, dst_len);
/*
* UTF supports lossless conversion round tripping [1] and conversions
* between UTF and other encodings are mostly round trip safe as
* Unicode aims to be a superset of all other character encodings.
* However, certain encodings (e.g. SHIFT-JIS) are known to have round
* trip issues [2]. Check the round trip conversion for all encodings
* listed in core.checkRoundtripEncoding.
*
* The round trip check is only performed if content is written to Git.
* This ensures that no information is lost during conversion to/from
* the internal UTF-8 representation.
*
* Please note, the code below is not tested because I was not able to
* generate a faulty round trip without an iconv error. Iconv errors
* are already caught above.
*
* [1] http://unicode.org/faq/utf_bom.html#gen2
* [2] https://support.microsoft.com/en-us/help/170559/prb-conversion-problem-between-shift-jis-and-unicode
*/
if (die_on_error && check_roundtrip(enc)) {
char *re_src;
size_t re_src_len;
re_src = reencode_string_len(dst, dst_len,
enc, default_encoding,
&re_src_len);
trace_printf("Checking roundtrip encoding for %s...\n", enc);
trace_encoding("reencoded source", path, enc,
re_src, re_src_len);
if (!re_src || src_len != re_src_len ||
memcmp(src, re_src, src_len)) {
const char* msg = _("encoding '%s' from %s to %s and "
"back is not the same");
die(msg, path, enc, default_encoding);
}
free(re_src);
}
strbuf_attach(buf, dst, dst_len, dst_len + 1);
return 1;
}
static int encode_to_worktree(const char *path, const char *src, size_t src_len,
struct strbuf *buf, const char *enc)
{
char *dst;
size_t dst_len;
/*
* No encoding is specified or there is nothing to encode.
* Tell the caller that the content was not modified.
*/
if (!enc || (src && !src_len))
return 0;
dst = reencode_string_len(src, src_len, enc, default_encoding,
&dst_len);
if (!dst) {
error(_("failed to encode '%s' from %s to %s"),
path, default_encoding, enc);
return 0;
}
strbuf_attach(buf, dst, dst_len, dst_len + 1);
return 1;
}
static int crlf_to_git(struct index_state *istate,
const char *path, const char *src, size_t len,
struct strbuf *buf,
enum convert_crlf_action crlf_action, int conv_flags)
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
{
struct text_stat stats;
char *dst;
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
int convert_crlf_into_lf;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
if (crlf_action == CRLF_BINARY ||
(src && !len))
return 0;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
/*
* If we are doing a dry-run and have no source buffer, there is
* nothing to analyze; we must assume we would convert.
*/
if (!buf && !src)
return 1;
gather_stats(src, len, &stats);
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
/* Optimization: No CRLF? Nothing to convert, regardless. */
convert_crlf_into_lf = !!stats.crlf;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
if (convert_is_binary(&stats))
return 0;
/*
* If the file in the index has any CR in it, do not
* convert. This is the new safer autocrlf handling,
* unless we want to renormalize in a merge or
* cherry-pick.
*/
if ((!(conv_flags & CONV_EOL_RENORMALIZE)) &&
has_crlf_in_index(istate, path))
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
convert_crlf_into_lf = 0;
}
if (((conv_flags & CONV_EOL_RNDTRP_WARN) ||
((conv_flags & CONV_EOL_RNDTRP_DIE) && len))) {
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
struct text_stat new_stats;
memcpy(&new_stats, &stats, sizeof(new_stats));
/* simulate "git add" */
if (convert_crlf_into_lf) {
new_stats.lonelf += new_stats.crlf;
new_stats.crlf = 0;
}
/* simulate "git checkout" */
if (will_convert_lf_to_crlf(&new_stats, crlf_action)) {
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
new_stats.crlf += new_stats.lonelf;
new_stats.lonelf = 0;
}
check_global_conv_flags_eol(path, &stats, &new_stats, conv_flags);
convert: Correct NNO tests and missing `LF will be replaced by CRLF` When a non-reversible CRLF conversion is done in "git add", a warning is printed on stderr (or Git dies, depending on checksafe) The function commit_chk_wrnNNO() in t0027 was written to test this, but did the wrong thing: Instead of looking at the warning from "git add", it looked at the warning from "git commit". This is racy because "git commit" may not have to do CRLF conversion at all if it can use the sha1 value from the index (which depends on whether "add" and "commit" run in a single second). Correct t0027 and replace the commit for each and every file with a commit of all files in one go. The function commit_chk_wrnNNO() should be renamed in a separate commit. Now that t0027 does the right thing, it detects a bug in covert.c: This sequence should generate the warning `LF will be replaced by CRLF`, but does not: $ git init $ git config core.autocrlf false $ printf "Line\r\n" >file $ git add file $ git commit -m "commit with CRLF" $ git config core.autocrlf true $ printf "Line\n" >file $ git add file "git add" calls crlf_to_git() in convert.c, which calls check_safe_crlf(). When has_cr_in_index(path) is true, crlf_to_git() returns too early and check_safe_crlf() is not called at all. Factor out the code which determines if "git checkout" converts LF->CRLF into will_convert_lf_to_crlf(). Update the logic around check_safe_crlf() and "simulate" the possible LF->CRLF conversion at "git checkout" with help of will_convert_lf_to_crlf(). Thanks to Jeff King <peff@peff.net> for analyzing t0027. Reported-By: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-08-13 21:29:27 +00:00
}
if (!convert_crlf_into_lf)
safecrlf: Add mechanism to warn about irreversible crlf conversions CRLF conversion bears a slight chance of corrupting data. autocrlf=true will convert CRLF to LF during commit and LF to CRLF during checkout. A file that contains a mixture of LF and CRLF before the commit cannot be recreated by git. For text files this is the right thing to do: it corrects line endings such that we have only LF line endings in the repository. But for binary files that are accidentally classified as text the conversion can corrupt data. If you recognize such corruption early you can easily fix it by setting the conversion type explicitly in .gitattributes. Right after committing you still have the original file in your work tree and this file is not yet corrupted. You can explicitly tell git that this file is binary and git will handle the file appropriately. Unfortunately, the desired effect of cleaning up text files with mixed line endings and the undesired effect of corrupting binary files cannot be distinguished. In both cases CRLFs are removed in an irreversible way. For text files this is the right thing to do because CRLFs are line endings, while for binary files converting CRLFs corrupts data. This patch adds a mechanism that can either warn the user about an irreversible conversion or can even refuse to convert. The mechanism is controlled by the variable core.safecrlf, with the following values: - false: disable safecrlf mechanism - warn: warn about irreversible conversions - true: refuse irreversible conversions The default is to warn. Users are only affected by this default if core.autocrlf is set. But the current default of git is to leave core.autocrlf unset, so users will not see warnings unless they deliberately chose to activate the autocrlf mechanism. The safecrlf mechanism's details depend on the git command. The general principles when safecrlf is active (not false) are: - we warn/error out if files in the work tree can modified in an irreversible way without giving the user a chance to backup the original file. - for read-only operations that do not modify files in the work tree we do not not print annoying warnings. There are exceptions. Even though... - "git add" itself does not touch the files in the work tree, the next checkout would, so the safety triggers; - "git apply" to update a text file with a patch does touch the files in the work tree, but the operation is about text files and CRLF conversion is about fixing the line ending inconsistencies, so the safety does not trigger; - "git diff" itself does not touch the files in the work tree, it is often run to inspect the changes you intend to next "git add". To catch potential problems early, safety triggers. The concept of a safety check was originally proposed in a similar way by Linus Torvalds. Thanks to Dimitry Potapov for insisting on getting the naked LF/autocrlf=true case right. Signed-off-by: Steffen Prohaska <prohaska@zib.de>
2008-02-06 11:25:58 +00:00
return 0;
/*
* At this point all of our source analysis is done, and we are sure we
* would convert. If we are in dry-run mode, we can give an answer.
*/
if (!buf)
return 1;
/* only grow if not in place */
if (strbuf_avail(buf) + buf->len < len)
strbuf_grow(buf, len - buf->len);
dst = buf->buf;
if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
/*
* If we guessed, we already know we rejected a file with
* lone CR, and we can strip a CR without looking at what
* follow it.
*/
do {
unsigned char c = *src++;
if (c != '\r')
*dst++ = c;
} while (--len);
} else {
do {
unsigned char c = *src++;
if (! (c == '\r' && (1 < len && *src == '\n')))
*dst++ = c;
} while (--len);
}
strbuf_setlen(buf, dst - buf->buf);
return 1;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
}
static int crlf_to_worktree(const char *src, size_t len, struct strbuf *buf,
enum convert_crlf_action crlf_action)
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
{
char *to_free = NULL;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
struct text_stat stats;
if (!len || output_eol(crlf_action) != EOL_CRLF)
return 0;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
gather_stats(src, len, &stats);
if (!will_convert_lf_to_crlf(&stats, crlf_action))
return 0;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
/* are we "faking" in place editing ? */
if (src == buf->buf)
to_free = strbuf_detach(buf, NULL);
strbuf_grow(buf, len + stats.lonelf);
for (;;) {
const char *nl = memchr(src, '\n', len);
if (!nl)
break;
if (nl > src && nl[-1] == '\r') {
strbuf_add(buf, src, nl + 1 - src);
} else {
strbuf_add(buf, src, nl - src);
strbuf_addstr(buf, "\r\n");
}
len -= nl + 1 - src;
src = nl + 1;
}
strbuf_add(buf, src, len);
free(to_free);
return 1;
Lazy man's auto-CRLF It currently does NOT know about file attributes, so it does its conversion purely based on content. Maybe that is more in the "git philosophy" anyway, since content is king, but I think we should try to do the file attributes to turn it off on demand. Anyway, BY DEFAULT it is off regardless, because it requires a [core] AutoCRLF = true in your config file to be enabled. We could make that the default for Windows, of course, the same way we do some other things (filemode etc). But you can actually enable it on UNIX, and it will cause: - "git update-index" will write blobs without CRLF - "git diff" will diff working tree files without CRLF - "git checkout" will write files to the working tree _with_ CRLF and things work fine. Funnily, it actually shows an odd file in git itself: git clone -n git test-crlf cd test-crlf git config core.autocrlf true git checkout git diff shows a diff for "Documentation/docbook-xsl.css". Why? Because we have actually checked in that file *with* CRLF! So when "core.autocrlf" is true, we'll always generate a *different* hash for it in the index, because the index hash will be for the content _without_ CRLF. Is this complete? I dunno. It seems to work for me. It doesn't use the filename at all right now, and that's probably a deficiency (we could certainly make the "is_binary()" heuristics also take standard filename heuristics into account). I don't pass in the filename at all for the "index_fd()" case (git-update-index), so that would need to be passed around, but this actually works fine. NOTE NOTE NOTE! The "is_binary()" heuristics are totally made-up by yours truly. I will not guarantee that they work at all reasonable. Caveat emptor. But it _is_ simple, and it _is_ safe, since it's all off by default. The patch is pretty simple - the biggest part is the new "convert.c" file, but even that is really just basic stuff that anybody can write in "Teaching C 101" as a final project for their first class in programming. Not to say that it's bug-free, of course - but at least we're not talking about rocket surgery here. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-02-13 19:07:23 +00:00
}
struct filter_params {
const char *src;
size_t size;
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
int fd;
const char *cmd;
const char *path;
};
static int filter_buffer_or_fd(int in UNUSED, int out, void *data)
{
/*
* Spawn cmd and feed the buffer contents through its stdin.
*/
struct child_process child_process = CHILD_PROCESS_INIT;
struct filter_params *params = (struct filter_params *)data;
int write_err, status;
/* apply % substitution to cmd */
struct strbuf cmd = STRBUF_INIT;
struct strbuf path = STRBUF_INIT;
struct strbuf_expand_dict_entry dict[] = {
{ "f", NULL, },
{ NULL, NULL, },
};
/* quote the path to preserve spaces, etc. */
sq_quote_buf(&path, params->path);
dict[0].value = path.buf;
/* expand all %f with the quoted path */
strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
strbuf_release(&path);
strvec_push(&child_process.args, cmd.buf);
child_process.use_shell = 1;
child_process.in = -1;
child_process.out = out;
if (start_command(&child_process)) {
strbuf_release(&cmd);
return error(_("cannot fork to run external filter '%s'"),
params->cmd);
}
sigchain_push(SIGPIPE, SIG_IGN);
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
if (params->src) {
write_err = (write_in_full(child_process.in,
params->src, params->size) < 0);
if (errno == EPIPE)
write_err = 0;
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
} else {
write_err = copy_fd(params->fd, child_process.in);
if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
write_err = 0;
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
}
if (close(child_process.in))
write_err = 1;
if (write_err)
error(_("cannot feed the input to external filter '%s'"),
params->cmd);
sigchain_pop(SIGPIPE);
status = finish_command(&child_process);
if (status)
error(_("external filter '%s' failed %d"), params->cmd, status);
strbuf_release(&cmd);
return (write_err || status);
}
static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd,
struct strbuf *dst, const char *cmd)
{
/*
* Create a pipeline to have the command filter the buffer's
* contents.
*
* (child --> cmd) --> us
*/
int err = 0;
struct strbuf nbuf = STRBUF_INIT;
struct async async;
struct filter_params params;
memset(&async, 0, sizeof(async));
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
async.proc = filter_buffer_or_fd;
async.data = &params;
async.out = -1;
params.src = src;
params.size = len;
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
params.fd = fd;
params.cmd = cmd;
params.path = path;
fflush(NULL);
if (start_async(&async))
return 0; /* error was already reported */
if (strbuf_read(&nbuf, async.out, 0) < 0) {
err = error(_("read from external filter '%s' failed"), cmd);
}
if (close(async.out)) {
err = error(_("read from external filter '%s' failed"), cmd);
}
if (finish_async(&async)) {
err = error(_("external filter '%s' failed"), cmd);
}
if (!err) {
strbuf_swap(dst, &nbuf);
}
strbuf_release(&nbuf);
return !err;
}
#define CAP_CLEAN (1u<<0)
#define CAP_SMUDGE (1u<<1)
#define CAP_DELAY (1u<<2)
struct cmd2process {
struct subprocess_entry subprocess; /* must be the first member! */
unsigned int supported_capabilities;
};
static int subprocess_map_initialized;
static struct hashmap subprocess_map;
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
static int start_multi_file_filter_fn(struct subprocess_entry *subprocess)
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
{
static int versions[] = {2, 0};
static struct subprocess_capability capabilities[] = {
{ "clean", CAP_CLEAN },
{ "smudge", CAP_SMUDGE },
{ "delay", CAP_DELAY },
{ NULL, 0 }
};
struct cmd2process *entry = (struct cmd2process *)subprocess;
return subprocess_handshake(subprocess, "git-filter", versions, NULL,
capabilities,
&entry->supported_capabilities);
}
static void handle_filter_error(const struct strbuf *filter_status,
struct cmd2process *entry,
const unsigned int wanted_capability)
{
if (!strcmp(filter_status->buf, "error"))
; /* The filter signaled a problem with the file. */
else if (!strcmp(filter_status->buf, "abort") && wanted_capability) {
/*
* The filter signaled a permanent problem. Don't try to filter
* files with the same command for the lifetime of the current
* Git process.
*/
entry->supported_capabilities &= ~wanted_capability;
} else {
/*
* Something went wrong with the protocol filter.
* Force shutdown and restart if another blob requires filtering.
*/
error(_("external filter '%s' failed"), entry->subprocess.cmd);
subprocess_stop(&subprocess_map, &entry->subprocess);
free(entry);
}
}
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
static int apply_multi_file_filter(const char *path, const char *src, size_t len,
int fd, struct strbuf *dst, const char *cmd,
const unsigned int wanted_capability,
convert: permit passing additional metadata to filter processes There are a variety of situations where a filter process can make use of some additional metadata. For example, some people find the ident filter too limiting and would like to include the commit or the branch in their smudged files. This information isn't available during checkout as HEAD hasn't been updated at that point, and it wouldn't be available in archives either. Let's add a way to pass this metadata down to the filter. We pass the blob we're operating on, the treeish (preferring the commit over the tree if one exists), and the ref we're operating on. Note that we won't pass this information in all cases, such as when renormalizing or when we're performing diffs, since it doesn't make sense in those cases. The data we currently get from the filter process looks like the following: command=smudge pathname=git.c 0000 With this change, we'll get data more like this: command=smudge pathname=git.c refname=refs/tags/v2.25.1 treeish=c522f061d551c9bb8684a7c3859b2ece4499b56b blob=7be7ad34bd053884ec48923706e70c81719a8660 0000 There are a couple things to note about this approach. For operations like checkout, treeish will always be a commit, since we cannot check out individual trees, but for other operations, like archive, we can end up operating on only a particular tree, so we'll provide only a tree as the treeish. Similar comments apply for refname, since there are a variety of cases in which we won't have a ref. This commit wires up the code to print this information, but doesn't pass any of it at this point. In a future commit, we'll have various code paths pass the actual useful data down. Signed-off-by: brian m. carlson <bk2204@github.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-03-16 18:05:02 +00:00
const struct checkout_metadata *meta,
struct delayed_checkout *dco)
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
{
int err;
int can_delay = 0;
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
struct cmd2process *entry;
struct child_process *process;
struct strbuf nbuf = STRBUF_INIT;
struct strbuf filter_status = STRBUF_INIT;
const char *filter_type;
if (!subprocess_map_initialized) {
subprocess_map_initialized = 1;
hashmap_init(&subprocess_map, cmd2process_cmp, NULL, 0);
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
entry = NULL;
} else {
entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
}
fflush(NULL);
if (!entry) {
entry = xmalloc(sizeof(*entry));
entry->supported_capabilities = 0;
if (subprocess_start(&subprocess_map, &entry->subprocess, cmd, start_multi_file_filter_fn)) {
free(entry);
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
return 0;
}
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
}
process = &entry->subprocess.process;
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
if (!(entry->supported_capabilities & wanted_capability))
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
return 0;
if (wanted_capability & CAP_CLEAN)
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
filter_type = "clean";
else if (wanted_capability & CAP_SMUDGE)
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
filter_type = "smudge";
else
die(_("unexpected filter type"));
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
sigchain_push(SIGPIPE, SIG_IGN);
assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n"));
err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type);
if (err)
goto done;
err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n");
if (err) {
error(_("path name too long for external filter"));
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
goto done;
}
err = packet_write_fmt_gently(process->in, "pathname=%s\n", path);
if (err)
goto done;
convert: permit passing additional metadata to filter processes There are a variety of situations where a filter process can make use of some additional metadata. For example, some people find the ident filter too limiting and would like to include the commit or the branch in their smudged files. This information isn't available during checkout as HEAD hasn't been updated at that point, and it wouldn't be available in archives either. Let's add a way to pass this metadata down to the filter. We pass the blob we're operating on, the treeish (preferring the commit over the tree if one exists), and the ref we're operating on. Note that we won't pass this information in all cases, such as when renormalizing or when we're performing diffs, since it doesn't make sense in those cases. The data we currently get from the filter process looks like the following: command=smudge pathname=git.c 0000 With this change, we'll get data more like this: command=smudge pathname=git.c refname=refs/tags/v2.25.1 treeish=c522f061d551c9bb8684a7c3859b2ece4499b56b blob=7be7ad34bd053884ec48923706e70c81719a8660 0000 There are a couple things to note about this approach. For operations like checkout, treeish will always be a commit, since we cannot check out individual trees, but for other operations, like archive, we can end up operating on only a particular tree, so we'll provide only a tree as the treeish. Similar comments apply for refname, since there are a variety of cases in which we won't have a ref. This commit wires up the code to print this information, but doesn't pass any of it at this point. In a future commit, we'll have various code paths pass the actual useful data down. Signed-off-by: brian m. carlson <bk2204@github.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-03-16 18:05:02 +00:00
if (meta && meta->refname) {
err = packet_write_fmt_gently(process->in, "ref=%s\n", meta->refname);
if (err)
goto done;
}
if (meta && !is_null_oid(&meta->treeish)) {
err = packet_write_fmt_gently(process->in, "treeish=%s\n", oid_to_hex(&meta->treeish));
if (err)
goto done;
}
if (meta && !is_null_oid(&meta->blob)) {
err = packet_write_fmt_gently(process->in, "blob=%s\n", oid_to_hex(&meta->blob));
if (err)
goto done;
}
if ((entry->supported_capabilities & CAP_DELAY) &&
dco && dco->state == CE_CAN_DELAY) {
can_delay = 1;
err = packet_write_fmt_gently(process->in, "can-delay=1\n");
if (err)
goto done;
}
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
err = packet_flush_gently(process->in);
if (err)
goto done;
if (fd >= 0)
err = write_packetized_from_fd_no_flush(fd, process->in);
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
else
err = write_packetized_from_buf_no_flush(src, len, process->in);
if (err)
goto done;
err = packet_flush_gently(process->in);
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
if (err)
goto done;
err = subprocess_read_status(process->out, &filter_status);
if (err)
goto done;
if (can_delay && !strcmp(filter_status.buf, "delayed")) {
string_list_insert(&dco->filters, cmd);
string_list_insert(&dco->paths, path);
} else {
/* The filter got the blob and wants to send us a response. */
err = strcmp(filter_status.buf, "success");
if (err)
goto done;
err = read_packetized_to_strbuf(process->out, &nbuf,
PACKET_READ_GENTLE_ON_EOF) < 0;
if (err)
goto done;
err = subprocess_read_status(process->out, &filter_status);
if (err)
goto done;
err = strcmp(filter_status.buf, "success");
}
done:
sigchain_pop(SIGPIPE);
if (err)
handle_filter_error(&filter_status, entry, wanted_capability);
else
strbuf_swap(dst, &nbuf);
strbuf_release(&nbuf);
convert: release strbuf to avoid leak apply_multi_file_filter and async_query_available_blobs both query subprocess output using subprocess_read_status, which writes data into the identically named filter_status strbuf. We add a strbuf_release to avoid leaking their contents. Leak output seen when running t0021 with LSAN: Direct leak of 24 byte(s) in 1 object(s) allocated from: #0 0x49ab49 in realloc ../projects/compiler-rt/lib/asan/asan_malloc_linux.cpp:164:3 #1 0xa8c2b5 in xrealloc wrapper.c:126:8 #2 0x9ff99d in strbuf_grow strbuf.c:98:2 #3 0x9ff99d in strbuf_addbuf strbuf.c:304:2 #4 0xa101d6 in subprocess_read_status sub-process.c:45:5 #5 0x77793c in apply_multi_file_filter convert.c:886:8 #6 0x77793c in apply_filter convert.c:1042:10 #7 0x77a0b5 in convert_to_git_filter_fd convert.c:1492:7 #8 0x8b48cd in index_stream_convert_blob object-file.c:2156:2 #9 0x8b48cd in index_fd object-file.c:2248:9 #10 0x597411 in hash_fd builtin/hash-object.c:43:9 #11 0x596be1 in hash_object builtin/hash-object.c:59:2 #12 0x596be1 in cmd_hash_object builtin/hash-object.c:153:3 #13 0x4ce83e in run_builtin git.c:475:11 #14 0x4ccafe in handle_builtin git.c:729:3 #15 0x4cb01c in run_argv git.c:818:4 #16 0x4cb01c in cmd_main git.c:949:19 #17 0x6bdc2d in main common-main.c:52:11 #18 0x7f42acf79349 in __libc_start_main (/lib64/libc.so.6+0x24349) SUMMARY: AddressSanitizer: 24 byte(s) leaked in 1 allocation(s). Direct leak of 120 byte(s) in 5 object(s) allocated from: #0 0x49ab49 in realloc ../projects/compiler-rt/lib/asan/asan_malloc_linux.cpp:164:3 #1 0xa8c295 in xrealloc wrapper.c:126:8 #2 0x9ff97d in strbuf_grow strbuf.c:98:2 #3 0x9ff97d in strbuf_addbuf strbuf.c:304:2 #4 0xa101b6 in subprocess_read_status sub-process.c:45:5 #5 0x775c73 in async_query_available_blobs convert.c:960:8 #6 0x80029d in finish_delayed_checkout entry.c:183:9 #7 0xa65d1e in check_updates unpack-trees.c:493:10 #8 0xa5f469 in unpack_trees unpack-trees.c:1747:8 #9 0x525971 in checkout builtin/clone.c:815:6 #10 0x525971 in cmd_clone builtin/clone.c:1409:8 #11 0x4ce83e in run_builtin git.c:475:11 #12 0x4ccafe in handle_builtin git.c:729:3 #13 0x4cb01c in run_argv git.c:818:4 #14 0x4cb01c in cmd_main git.c:949:19 #15 0x6bdc2d in main common-main.c:52:11 #16 0x7fa253fce349 in __libc_start_main (/lib64/libc.so.6+0x24349) SUMMARY: AddressSanitizer: 120 byte(s) leaked in 5 allocation(s). Signed-off-by: Andrzej Hunt <andrzej@ahunt.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-07-25 13:08:26 +00:00
strbuf_release(&filter_status);
return !err;
}
int async_query_available_blobs(const char *cmd, struct string_list *available_paths)
{
int err;
char *line;
struct cmd2process *entry;
struct child_process *process;
struct strbuf filter_status = STRBUF_INIT;
assert(subprocess_map_initialized);
entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
if (!entry) {
error(_("external filter '%s' is not available anymore although "
"not all paths have been filtered"), cmd);
return 0;
}
process = &entry->subprocess.process;
sigchain_push(SIGPIPE, SIG_IGN);
err = packet_write_fmt_gently(
process->in, "command=list_available_blobs\n");
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
if (err)
goto done;
err = packet_flush_gently(process->in);
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
if (err)
goto done;
while ((line = packet_read_line(process->out, NULL))) {
const char *path;
if (skip_prefix(line, "pathname=", &path))
string_list_insert(available_paths, xstrdup(path));
else
; /* ignore unknown keys */
}
err = subprocess_read_status(process->out, &filter_status);
if (err)
goto done;
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
err = strcmp(filter_status.buf, "success");
done:
sigchain_pop(SIGPIPE);
if (err)
handle_filter_error(&filter_status, entry, 0);
convert: release strbuf to avoid leak apply_multi_file_filter and async_query_available_blobs both query subprocess output using subprocess_read_status, which writes data into the identically named filter_status strbuf. We add a strbuf_release to avoid leaking their contents. Leak output seen when running t0021 with LSAN: Direct leak of 24 byte(s) in 1 object(s) allocated from: #0 0x49ab49 in realloc ../projects/compiler-rt/lib/asan/asan_malloc_linux.cpp:164:3 #1 0xa8c2b5 in xrealloc wrapper.c:126:8 #2 0x9ff99d in strbuf_grow strbuf.c:98:2 #3 0x9ff99d in strbuf_addbuf strbuf.c:304:2 #4 0xa101d6 in subprocess_read_status sub-process.c:45:5 #5 0x77793c in apply_multi_file_filter convert.c:886:8 #6 0x77793c in apply_filter convert.c:1042:10 #7 0x77a0b5 in convert_to_git_filter_fd convert.c:1492:7 #8 0x8b48cd in index_stream_convert_blob object-file.c:2156:2 #9 0x8b48cd in index_fd object-file.c:2248:9 #10 0x597411 in hash_fd builtin/hash-object.c:43:9 #11 0x596be1 in hash_object builtin/hash-object.c:59:2 #12 0x596be1 in cmd_hash_object builtin/hash-object.c:153:3 #13 0x4ce83e in run_builtin git.c:475:11 #14 0x4ccafe in handle_builtin git.c:729:3 #15 0x4cb01c in run_argv git.c:818:4 #16 0x4cb01c in cmd_main git.c:949:19 #17 0x6bdc2d in main common-main.c:52:11 #18 0x7f42acf79349 in __libc_start_main (/lib64/libc.so.6+0x24349) SUMMARY: AddressSanitizer: 24 byte(s) leaked in 1 allocation(s). Direct leak of 120 byte(s) in 5 object(s) allocated from: #0 0x49ab49 in realloc ../projects/compiler-rt/lib/asan/asan_malloc_linux.cpp:164:3 #1 0xa8c295 in xrealloc wrapper.c:126:8 #2 0x9ff97d in strbuf_grow strbuf.c:98:2 #3 0x9ff97d in strbuf_addbuf strbuf.c:304:2 #4 0xa101b6 in subprocess_read_status sub-process.c:45:5 #5 0x775c73 in async_query_available_blobs convert.c:960:8 #6 0x80029d in finish_delayed_checkout entry.c:183:9 #7 0xa65d1e in check_updates unpack-trees.c:493:10 #8 0xa5f469 in unpack_trees unpack-trees.c:1747:8 #9 0x525971 in checkout builtin/clone.c:815:6 #10 0x525971 in cmd_clone builtin/clone.c:1409:8 #11 0x4ce83e in run_builtin git.c:475:11 #12 0x4ccafe in handle_builtin git.c:729:3 #13 0x4cb01c in run_argv git.c:818:4 #14 0x4cb01c in cmd_main git.c:949:19 #15 0x6bdc2d in main common-main.c:52:11 #16 0x7fa253fce349 in __libc_start_main (/lib64/libc.so.6+0x24349) SUMMARY: AddressSanitizer: 120 byte(s) leaked in 5 allocation(s). Signed-off-by: Andrzej Hunt <andrzej@ahunt.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-07-25 13:08:26 +00:00
strbuf_release(&filter_status);
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
return !err;
}
static struct convert_driver {
const char *name;
struct convert_driver *next;
const char *smudge;
const char *clean;
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
const char *process;
int required;
} *user_convert, **user_convert_tail;
static int apply_filter(const char *path, const char *src, size_t len,
int fd, struct strbuf *dst, struct convert_driver *drv,
const unsigned int wanted_capability,
convert: permit passing additional metadata to filter processes There are a variety of situations where a filter process can make use of some additional metadata. For example, some people find the ident filter too limiting and would like to include the commit or the branch in their smudged files. This information isn't available during checkout as HEAD hasn't been updated at that point, and it wouldn't be available in archives either. Let's add a way to pass this metadata down to the filter. We pass the blob we're operating on, the treeish (preferring the commit over the tree if one exists), and the ref we're operating on. Note that we won't pass this information in all cases, such as when renormalizing or when we're performing diffs, since it doesn't make sense in those cases. The data we currently get from the filter process looks like the following: command=smudge pathname=git.c 0000 With this change, we'll get data more like this: command=smudge pathname=git.c refname=refs/tags/v2.25.1 treeish=c522f061d551c9bb8684a7c3859b2ece4499b56b blob=7be7ad34bd053884ec48923706e70c81719a8660 0000 There are a couple things to note about this approach. For operations like checkout, treeish will always be a commit, since we cannot check out individual trees, but for other operations, like archive, we can end up operating on only a particular tree, so we'll provide only a tree as the treeish. Similar comments apply for refname, since there are a variety of cases in which we won't have a ref. This commit wires up the code to print this information, but doesn't pass any of it at this point. In a future commit, we'll have various code paths pass the actual useful data down. Signed-off-by: brian m. carlson <bk2204@github.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-03-16 18:05:02 +00:00
const struct checkout_metadata *meta,
struct delayed_checkout *dco)
{
const char *cmd = NULL;
if (!drv)
return 0;
if (!dst)
return 1;
if ((wanted_capability & CAP_CLEAN) && !drv->process && drv->clean)
cmd = drv->clean;
else if ((wanted_capability & CAP_SMUDGE) && !drv->process && drv->smudge)
cmd = drv->smudge;
if (cmd && *cmd)
return apply_single_file_filter(path, src, len, fd, dst, cmd);
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
else if (drv->process && *drv->process)
return apply_multi_file_filter(path, src, len, fd, dst,
convert: permit passing additional metadata to filter processes There are a variety of situations where a filter process can make use of some additional metadata. For example, some people find the ident filter too limiting and would like to include the commit or the branch in their smudged files. This information isn't available during checkout as HEAD hasn't been updated at that point, and it wouldn't be available in archives either. Let's add a way to pass this metadata down to the filter. We pass the blob we're operating on, the treeish (preferring the commit over the tree if one exists), and the ref we're operating on. Note that we won't pass this information in all cases, such as when renormalizing or when we're performing diffs, since it doesn't make sense in those cases. The data we currently get from the filter process looks like the following: command=smudge pathname=git.c 0000 With this change, we'll get data more like this: command=smudge pathname=git.c refname=refs/tags/v2.25.1 treeish=c522f061d551c9bb8684a7c3859b2ece4499b56b blob=7be7ad34bd053884ec48923706e70c81719a8660 0000 There are a couple things to note about this approach. For operations like checkout, treeish will always be a commit, since we cannot check out individual trees, but for other operations, like archive, we can end up operating on only a particular tree, so we'll provide only a tree as the treeish. Similar comments apply for refname, since there are a variety of cases in which we won't have a ref. This commit wires up the code to print this information, but doesn't pass any of it at this point. In a future commit, we'll have various code paths pass the actual useful data down. Signed-off-by: brian m. carlson <bk2204@github.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-03-16 18:05:02 +00:00
drv->process, wanted_capability, meta, dco);
return 0;
}
static int read_convert_config(const char *var, const char *value, void *cb UNUSED)
{
const char *key, *name;
size_t namelen;
struct convert_driver *drv;
/*
* External conversion drivers are configured using
* "filter.<name>.variable".
*/
if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
return 0;
for (drv = user_convert; drv; drv = drv->next)
if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
break;
if (!drv) {
CALLOC_ARRAY(drv, 1);
drv->name = xmemdupz(name, namelen);
*user_convert_tail = drv;
user_convert_tail = &(drv->next);
}
/*
* filter.<name>.smudge and filter.<name>.clean specifies
* the command line:
*
* command-line
*
* The command-line will not be interpolated in any way.
*/
if (!strcmp("smudge", key))
return git_config_string(&drv->smudge, var, value);
if (!strcmp("clean", key))
return git_config_string(&drv->clean, var, value);
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
if (!strcmp("process", key))
return git_config_string(&drv->process, var, value);
if (!strcmp("required", key)) {
drv->required = git_config_bool(var, value);
return 0;
}
return 0;
}
static int count_ident(const char *cp, unsigned long size)
{
/*
* "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
*/
int cnt = 0;
char ch;
while (size) {
ch = *cp++;
size--;
if (ch != '$')
continue;
if (size < 3)
break;
if (memcmp("Id", cp, 2))
continue;
ch = cp[2];
cp += 3;
size -= 3;
if (ch == '$')
cnt++; /* $Id$ */
if (ch != ':')
continue;
/*
* "$Id: ... "; scan up to the closing dollar sign and discard.
*/
while (size) {
ch = *cp++;
size--;
if (ch == '$') {
cnt++;
break;
}
if (ch == '\n')
break;
}
}
return cnt;
}
static int ident_to_git(const char *src, size_t len,
struct strbuf *buf, int ident)
{
char *dst, *dollar;
if (!ident || (src && !count_ident(src, len)))
return 0;
if (!buf)
return 1;
/* only grow if not in place */
if (strbuf_avail(buf) + buf->len < len)
strbuf_grow(buf, len - buf->len);
dst = buf->buf;
for (;;) {
dollar = memchr(src, '$', len);
if (!dollar)
break;
memmove(dst, src, dollar + 1 - src);
dst += dollar + 1 - src;
len -= dollar + 1 - src;
src = dollar + 1;
if (len > 3 && !memcmp(src, "Id:", 3)) {
dollar = memchr(src + 3, '$', len - 3);
if (!dollar)
break;
if (memchr(src + 3, '\n', dollar - src - 3)) {
/* Line break before the next dollar. */
continue;
}
memcpy(dst, "Id$", 3);
dst += 3;
len -= dollar + 1 - src;
src = dollar + 1;
}
}
memmove(dst, src, len);
strbuf_setlen(buf, dst + len - buf->buf);
return 1;
}
static int ident_to_worktree(const char *src, size_t len,
struct strbuf *buf, int ident)
{
struct object_id oid;
char *to_free = NULL, *dollar, *spc;
int cnt;
if (!ident)
return 0;
cnt = count_ident(src, len);
if (!cnt)
return 0;
/* are we "faking" in place editing ? */
if (src == buf->buf)
to_free = strbuf_detach(buf, NULL);
hash_object_file(the_hash_algo, src, len, OBJ_BLOB, &oid);
strbuf_grow(buf, len + cnt * (the_hash_algo->hexsz + 3));
for (;;) {
/* step 1: run to the next '$' */
dollar = memchr(src, '$', len);
if (!dollar)
break;
strbuf_add(buf, src, dollar + 1 - src);
len -= dollar + 1 - src;
src = dollar + 1;
Fix mishandling of $Id$ expanded in the repository copy in convert.c If the repository contained an expanded ident keyword (i.e. $Id:XXXX$), then the wrong bytes were discarded, and the Id keyword was not expanded. The fault was in convert.c:ident_to_worktree(). Previously, when a "$Id:" was found in the repository version, ident_to_worktree() would search for the next "$" after this, and discarded everything it found until then. That was done with the loop: do { ch = *cp++; if (ch == '$') break; rem--; } while (rem); The above loop left cp pointing one character _after_ the final "$" (because of ch = *cp++). This was different from the non-expanded case, were cp is left pointing at the "$", and was different from the comment which stated "discard up to but not including the closing $". This patch fixes that by making the loop: do { ch = *cp; if (ch == '$') break; cp++; rem--; } while (rem); That is, cp is tested _then_ incremented. This loop exits if it finds a "$" or if it runs out of bytes in the source. After this loop, if there was no closing "$" the expansion is skipped, and the outer loop is allowed to continue leaving this non-keyword as it was. However, when the "$" is found, size is corrected, before running the expansion: size -= (cp - src); This is wrong; size is going to be corrected anyway after the expansion, so there is no need to do it here. This patch removes that redundant correction. To help find this bug, I heavily commented the routine; those comments are included here as a bonus. Signed-off-by: Andy Parkins <andyparkins@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-05-25 10:50:08 +00:00
/* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
if (len < 3 || memcmp("Id", src, 2))
continue;
/* step 3: skip over Id$ or Id:xxxxx$ */
if (src[2] == '$') {
src += 3;
len -= 3;
} else if (src[2] == ':') {
/*
* It's possible that an expanded Id has crept its way into the
* repository, we cope with that by stripping the expansion out.
* This is probably not a good idea, since it will cause changes
* on checkout, which won't go away by stash, but let's keep it
* for git-style ids.
*/
dollar = memchr(src + 3, '$', len - 3);
if (!dollar) {
/* incomplete keyword, no more '$', so just quit the loop */
break;
}
Fix mishandling of $Id$ expanded in the repository copy in convert.c If the repository contained an expanded ident keyword (i.e. $Id:XXXX$), then the wrong bytes were discarded, and the Id keyword was not expanded. The fault was in convert.c:ident_to_worktree(). Previously, when a "$Id:" was found in the repository version, ident_to_worktree() would search for the next "$" after this, and discarded everything it found until then. That was done with the loop: do { ch = *cp++; if (ch == '$') break; rem--; } while (rem); The above loop left cp pointing one character _after_ the final "$" (because of ch = *cp++). This was different from the non-expanded case, were cp is left pointing at the "$", and was different from the comment which stated "discard up to but not including the closing $". This patch fixes that by making the loop: do { ch = *cp; if (ch == '$') break; cp++; rem--; } while (rem); That is, cp is tested _then_ incremented. This loop exits if it finds a "$" or if it runs out of bytes in the source. After this loop, if there was no closing "$" the expansion is skipped, and the outer loop is allowed to continue leaving this non-keyword as it was. However, when the "$" is found, size is corrected, before running the expansion: size -= (cp - src); This is wrong; size is going to be corrected anyway after the expansion, so there is no need to do it here. This patch removes that redundant correction. To help find this bug, I heavily commented the routine; those comments are included here as a bonus. Signed-off-by: Andy Parkins <andyparkins@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-05-25 10:50:08 +00:00
if (memchr(src + 3, '\n', dollar - src - 3)) {
/* Line break before the next dollar. */
continue;
}
spc = memchr(src + 4, ' ', dollar - src - 4);
if (spc && spc < dollar-1) {
/* There are spaces in unexpected places.
* This is probably an id from some other
* versioning system. Keep it for now.
*/
continue;
}
len -= dollar + 1 - src;
src = dollar + 1;
} else {
/* it wasn't a "Id$" or "Id:xxxx$" */
continue;
}
Fix mishandling of $Id$ expanded in the repository copy in convert.c If the repository contained an expanded ident keyword (i.e. $Id:XXXX$), then the wrong bytes were discarded, and the Id keyword was not expanded. The fault was in convert.c:ident_to_worktree(). Previously, when a "$Id:" was found in the repository version, ident_to_worktree() would search for the next "$" after this, and discarded everything it found until then. That was done with the loop: do { ch = *cp++; if (ch == '$') break; rem--; } while (rem); The above loop left cp pointing one character _after_ the final "$" (because of ch = *cp++). This was different from the non-expanded case, were cp is left pointing at the "$", and was different from the comment which stated "discard up to but not including the closing $". This patch fixes that by making the loop: do { ch = *cp; if (ch == '$') break; cp++; rem--; } while (rem); That is, cp is tested _then_ incremented. This loop exits if it finds a "$" or if it runs out of bytes in the source. After this loop, if there was no closing "$" the expansion is skipped, and the outer loop is allowed to continue leaving this non-keyword as it was. However, when the "$" is found, size is corrected, before running the expansion: size -= (cp - src); This is wrong; size is going to be corrected anyway after the expansion, so there is no need to do it here. This patch removes that redundant correction. To help find this bug, I heavily commented the routine; those comments are included here as a bonus. Signed-off-by: Andy Parkins <andyparkins@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
2007-05-25 10:50:08 +00:00
/* step 4: substitute */
strbuf_addstr(buf, "Id: ");
strbuf_addstr(buf, oid_to_hex(&oid));
strbuf_addstr(buf, " $");
}
strbuf_add(buf, src, len);
free(to_free);
return 1;
}
static const char *git_path_check_encoding(struct attr_check_item *check)
{
const char *value = check->value;
if (ATTR_UNSET(value) || !strlen(value))
return NULL;
if (ATTR_TRUE(value) || ATTR_FALSE(value)) {
die(_("true/false are no valid working-tree-encodings"));
}
/* Don't encode to the default encoding */
if (same_encoding(value, default_encoding))
return NULL;
return value;
}
static enum convert_crlf_action git_path_check_crlf(struct attr_check_item *check)
{
const char *value = check->value;
if (ATTR_TRUE(value))
return CRLF_TEXT;
else if (ATTR_FALSE(value))
return CRLF_BINARY;
else if (ATTR_UNSET(value))
;
else if (!strcmp(value, "input"))
return CRLF_TEXT_INPUT;
else if (!strcmp(value, "auto"))
return CRLF_AUTO;
return CRLF_UNDEFINED;
}
static enum eol git_path_check_eol(struct attr_check_item *check)
{
const char *value = check->value;
if (ATTR_UNSET(value))
;
else if (!strcmp(value, "lf"))
return EOL_LF;
else if (!strcmp(value, "crlf"))
return EOL_CRLF;
return EOL_UNSET;
}
static struct convert_driver *git_path_check_convert(struct attr_check_item *check)
{
const char *value = check->value;
struct convert_driver *drv;
if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
return NULL;
for (drv = user_convert; drv; drv = drv->next)
if (!strcmp(value, drv->name))
return drv;
return NULL;
}
static int git_path_check_ident(struct attr_check_item *check)
{
const char *value = check->value;
return !!ATTR_TRUE(value);
}
static struct attr_check *check;
void convert_attrs(struct index_state *istate,
struct conv_attrs *ca, const char *path)
{
struct attr_check_item *ccheck = NULL;
if (!check) {
check = attr_check_initl("crlf", "ident", "filter",
"eol", "text", "working-tree-encoding",
NULL);
user_convert_tail = &user_convert;
git_config(read_convert_config, NULL);
}
attr: add flag `--source` to work with tree-ish The contents of the .gitattributes files may evolve over time, but "git check-attr" always checks attributes against them in the working tree and/or in the index. It may be beneficial to optionally allow the users to check attributes taken from a commit other than HEAD against paths. Add a new flag `--source` which will allow users to check the attributes against a commit (actually any tree-ish would do). When the user uses this flag, we go through the stack of .gitattributes files but instead of checking the current working tree and/or in the index, we check the blobs from the provided tree-ish object. This allows the command to also be used in bare repositories. Since we use a tree-ish object, the user can pass "--source HEAD:subdirectory" and all the attributes will be looked up as if subdirectory was the root directory of the repository. We cannot simply use the `<rev>:<path>` syntax without the `--source` flag, similar to how it is used in `git show` because any non-flag parameter before `--` is treated as an attribute and any parameter after `--` is treated as a pathname. The change involves creating a new function `read_attr_from_blob`, which given the path reads the blob for the path against the provided source and parses the attributes line by line. This function is plugged into `read_attr()` function wherein we go through the stack of attributes files. Signed-off-by: Karthik Nayak <karthik.188@gmail.com> Signed-off-by: Toon Claes <toon@iotcl.com> Co-authored-by: toon@iotcl.com Signed-off-by: Junio C Hamano <gitster@pobox.com>
2023-01-14 08:30:38 +00:00
git_check_attr(istate, NULL, path, check);
ccheck = check->items;
ca->crlf_action = git_path_check_crlf(ccheck + 4);
if (ca->crlf_action == CRLF_UNDEFINED)
ca->crlf_action = git_path_check_crlf(ccheck + 0);
ca->ident = git_path_check_ident(ccheck + 1);
ca->drv = git_path_check_convert(ccheck + 2);
if (ca->crlf_action != CRLF_BINARY) {
enum eol eol_attr = git_path_check_eol(ccheck + 3);
if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
ca->crlf_action = CRLF_AUTO_INPUT;
else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
ca->crlf_action = CRLF_AUTO_CRLF;
else if (eol_attr == EOL_LF)
ca->crlf_action = CRLF_TEXT_INPUT;
else if (eol_attr == EOL_CRLF)
ca->crlf_action = CRLF_TEXT_CRLF;
}
ca->working_tree_encoding = git_path_check_encoding(ccheck + 5);
/* Save attr and make a decision for action */
ca->attr_action = ca->crlf_action;
if (ca->crlf_action == CRLF_TEXT)
ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
ca->crlf_action = CRLF_BINARY;
if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
ca->crlf_action = CRLF_AUTO_CRLF;
if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
ca->crlf_action = CRLF_AUTO_INPUT;
}
void reset_parsed_attributes(void)
{
struct convert_driver *drv, *next;
attr_check_free(check);
check = NULL;
reset_merge_attributes();
for (drv = user_convert; drv; drv = next) {
next = drv->next;
free((void *)drv->name);
free(drv);
}
user_convert = NULL;
user_convert_tail = NULL;
}
int would_convert_to_git_filter_fd(struct index_state *istate, const char *path)
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
{
struct conv_attrs ca;
convert_attrs(istate, &ca, path);
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
if (!ca.drv)
return 0;
/*
* Apply a filter to an fd only if the filter is required to succeed.
* We must die if the filter fails, because the original data before
* filtering is not available.
*/
if (!ca.drv->required)
return 0;
convert: permit passing additional metadata to filter processes There are a variety of situations where a filter process can make use of some additional metadata. For example, some people find the ident filter too limiting and would like to include the commit or the branch in their smudged files. This information isn't available during checkout as HEAD hasn't been updated at that point, and it wouldn't be available in archives either. Let's add a way to pass this metadata down to the filter. We pass the blob we're operating on, the treeish (preferring the commit over the tree if one exists), and the ref we're operating on. Note that we won't pass this information in all cases, such as when renormalizing or when we're performing diffs, since it doesn't make sense in those cases. The data we currently get from the filter process looks like the following: command=smudge pathname=git.c 0000 With this change, we'll get data more like this: command=smudge pathname=git.c refname=refs/tags/v2.25.1 treeish=c522f061d551c9bb8684a7c3859b2ece4499b56b blob=7be7ad34bd053884ec48923706e70c81719a8660 0000 There are a couple things to note about this approach. For operations like checkout, treeish will always be a commit, since we cannot check out individual trees, but for other operations, like archive, we can end up operating on only a particular tree, so we'll provide only a tree as the treeish. Similar comments apply for refname, since there are a variety of cases in which we won't have a ref. This commit wires up the code to print this information, but doesn't pass any of it at this point. In a future commit, we'll have various code paths pass the actual useful data down. Signed-off-by: brian m. carlson <bk2204@github.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-03-16 18:05:02 +00:00
return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN, NULL, NULL);
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
}
const char *get_convert_attr_ascii(struct index_state *istate, const char *path)
{
struct conv_attrs ca;
convert_attrs(istate, &ca, path);
switch (ca.attr_action) {
case CRLF_UNDEFINED:
return "";
case CRLF_BINARY:
return "-text";
case CRLF_TEXT:
return "text";
case CRLF_TEXT_INPUT:
return "text eol=lf";
case CRLF_TEXT_CRLF:
return "text eol=crlf";
case CRLF_AUTO:
return "text=auto";
case CRLF_AUTO_CRLF:
return "text=auto eol=crlf";
case CRLF_AUTO_INPUT:
return "text=auto eol=lf";
}
return "";
}
int convert_to_git(struct index_state *istate,
const char *path, const char *src, size_t len,
struct strbuf *dst, int conv_flags)
{
int ret = 0;
struct conv_attrs ca;
convert_attrs(istate, &ca, path);
convert: permit passing additional metadata to filter processes There are a variety of situations where a filter process can make use of some additional metadata. For example, some people find the ident filter too limiting and would like to include the commit or the branch in their smudged files. This information isn't available during checkout as HEAD hasn't been updated at that point, and it wouldn't be available in archives either. Let's add a way to pass this metadata down to the filter. We pass the blob we're operating on, the treeish (preferring the commit over the tree if one exists), and the ref we're operating on. Note that we won't pass this information in all cases, such as when renormalizing or when we're performing diffs, since it doesn't make sense in those cases. The data we currently get from the filter process looks like the following: command=smudge pathname=git.c 0000 With this change, we'll get data more like this: command=smudge pathname=git.c refname=refs/tags/v2.25.1 treeish=c522f061d551c9bb8684a7c3859b2ece4499b56b blob=7be7ad34bd053884ec48923706e70c81719a8660 0000 There are a couple things to note about this approach. For operations like checkout, treeish will always be a commit, since we cannot check out individual trees, but for other operations, like archive, we can end up operating on only a particular tree, so we'll provide only a tree as the treeish. Similar comments apply for refname, since there are a variety of cases in which we won't have a ref. This commit wires up the code to print this information, but doesn't pass any of it at this point. In a future commit, we'll have various code paths pass the actual useful data down. Signed-off-by: brian m. carlson <bk2204@github.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-03-16 18:05:02 +00:00
ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN, NULL, NULL);
if (!ret && ca.drv && ca.drv->required)
die(_("%s: clean filter '%s' failed"), path, ca.drv->name);
if (ret && dst) {
src = dst->buf;
len = dst->len;
}
ret |= encode_to_git(path, src, len, dst, ca.working_tree_encoding, conv_flags);
if (ret && dst) {
src = dst->buf;
len = dst->len;
}
if (!(conv_flags & CONV_EOL_KEEP_CRLF)) {
ret |= crlf_to_git(istate, path, src, len, dst, ca.crlf_action, conv_flags);
if (ret && dst) {
src = dst->buf;
len = dst->len;
}
}
return ret | ident_to_git(src, len, dst, ca.ident);
}
void convert_to_git_filter_fd(struct index_state *istate,
const char *path, int fd, struct strbuf *dst,
int conv_flags)
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
{
struct conv_attrs ca;
convert_attrs(istate, &ca, path);
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
assert(ca.drv);
convert: permit passing additional metadata to filter processes There are a variety of situations where a filter process can make use of some additional metadata. For example, some people find the ident filter too limiting and would like to include the commit or the branch in their smudged files. This information isn't available during checkout as HEAD hasn't been updated at that point, and it wouldn't be available in archives either. Let's add a way to pass this metadata down to the filter. We pass the blob we're operating on, the treeish (preferring the commit over the tree if one exists), and the ref we're operating on. Note that we won't pass this information in all cases, such as when renormalizing or when we're performing diffs, since it doesn't make sense in those cases. The data we currently get from the filter process looks like the following: command=smudge pathname=git.c 0000 With this change, we'll get data more like this: command=smudge pathname=git.c refname=refs/tags/v2.25.1 treeish=c522f061d551c9bb8684a7c3859b2ece4499b56b blob=7be7ad34bd053884ec48923706e70c81719a8660 0000 There are a couple things to note about this approach. For operations like checkout, treeish will always be a commit, since we cannot check out individual trees, but for other operations, like archive, we can end up operating on only a particular tree, so we'll provide only a tree as the treeish. Similar comments apply for refname, since there are a variety of cases in which we won't have a ref. This commit wires up the code to print this information, but doesn't pass any of it at this point. In a future commit, we'll have various code paths pass the actual useful data down. Signed-off-by: brian m. carlson <bk2204@github.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-03-16 18:05:02 +00:00
if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN, NULL, NULL))
die(_("%s: clean filter '%s' failed"), path, ca.drv->name);
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
encode_to_git(path, dst->buf, dst->len, dst, ca.working_tree_encoding, conv_flags);
crlf_to_git(istate, path, dst->buf, dst->len, dst, ca.crlf_action, conv_flags);
ident_to_git(dst->buf, dst->len, dst, ca.ident);
convert: stream from fd to required clean filter to reduce used address space The data is streamed to the filter process anyway. Better avoid mapping the file if possible. This is especially useful if a clean filter reduces the size, for example if it computes a sha1 for binary data, like git media. The file size that the previous implementation could handle was limited by the available address space; large files for example could not be handled with (32-bit) msysgit. The new implementation can filter files of any size as long as the filter output is small enough. The new code path is only taken if the filter is required. The filter consumes data directly from the fd. If it fails, the original data is not immediately available. The condition can easily be handled as a fatal error, which is expected for a required filter anyway. If the filter was not required, the condition would need to be handled in a different way, like seeking to 0 and reading the data. But this would require more restructuring of the code and is probably not worth it. The obvious approach of falling back to reading all data would not help achieving the main purpose of this patch, which is to handle large files with limited address space. If reading all data is an option, we can simply take the old code path right away and mmap the entire file. The environment variable GIT_MMAP_LIMIT, which has been introduced in a previous commit is used to test that the expected code path is taken. A related test that exercises required filters is modified to verify that the data actually has been modified on its way from the file system to the object store. Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-26 15:23:25 +00:00
}
static int convert_to_working_tree_ca_internal(const struct conv_attrs *ca,
const char *path, const char *src,
size_t len, struct strbuf *dst,
int normalizing,
const struct checkout_metadata *meta,
struct delayed_checkout *dco)
{
int ret = 0, ret_filter = 0;
ret |= ident_to_worktree(src, len, dst, ca->ident);
if (ret) {
src = dst->buf;
len = dst->len;
}
/*
* CRLF conversion can be skipped if normalizing, unless there
convert: add filter.<driver>.process option Git's clean/smudge mechanism invokes an external filter process for every single blob that is affected by a filter. If Git filters a lot of blobs then the startup time of the external filter processes can become a significant part of the overall Git execution time. In a preliminary performance test this developer used a clean/smudge filter written in golang to filter 12,000 files. This process took 364s with the existing filter mechanism and 5s with the new mechanism. See details here: https://github.com/github/git-lfs/pull/1382 This patch adds the `filter.<driver>.process` string option which, if used, keeps the external filter process running and processes all blobs with the packet format (pkt-line) based protocol over standard input and standard output. The full protocol is explained in detail in `Documentation/gitattributes.txt`. A few key decisions: * The long running filter process is referred to as filter protocol version 2 because the existing single shot filter invocation is considered version 1. * Git sends a welcome message and expects a response right after the external filter process has started. This ensures that Git will not hang if a version 1 filter is incorrectly used with the filter.<driver>.process option for version 2 filters. In addition, Git can detect this kind of error and warn the user. * The status of a filter operation (e.g. "success" or "error) is set before the actual response and (if necessary!) re-set after the response. The advantage of this two step status response is that if the filter detects an error early, then the filter can communicate this and Git does not even need to create structures to read the response. * All status responses are pkt-line lists terminated with a flush packet. This allows us to send other status fields with the same protocol in the future. Helped-by: Martin-Louis Bright <mlbright@gmail.com> Reviewed-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-10-16 23:20:37 +00:00
* is a smudge or process filter (even if the process filter doesn't
* support smudge). The filters might expect CRLFs.
*/
if ((ca->drv && (ca->drv->smudge || ca->drv->process)) || !normalizing) {
ret |= crlf_to_worktree(src, len, dst, ca->crlf_action);
if (ret) {
src = dst->buf;
len = dst->len;
}
}
ret |= encode_to_worktree(path, src, len, dst, ca->working_tree_encoding);
if (ret) {
src = dst->buf;
len = dst->len;
}
ret_filter = apply_filter(
path, src, len, -1, dst, ca->drv, CAP_SMUDGE, meta, dco);
if (!ret_filter && ca->drv && ca->drv->required)
die(_("%s: smudge filter %s failed"), path, ca->drv->name);
return ret | ret_filter;
}
int async_convert_to_working_tree_ca(const struct conv_attrs *ca,
const char *path, const char *src,
size_t len, struct strbuf *dst,
const struct checkout_metadata *meta,
void *dco)
{
return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0,
meta, dco);
}
int convert_to_working_tree_ca(const struct conv_attrs *ca,
const char *path, const char *src,
size_t len, struct strbuf *dst,
const struct checkout_metadata *meta)
{
return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0,
meta, NULL);
}
int renormalize_buffer(struct index_state *istate, const char *path,
const char *src, size_t len, struct strbuf *dst)
{
struct conv_attrs ca;
int ret;
convert_attrs(istate, &ca, path);
ret = convert_to_working_tree_ca_internal(&ca, path, src, len, dst, 1,
NULL, NULL);
if (ret) {
src = dst->buf;
len = dst->len;
}
return ret | convert_to_git(istate, path, src, len, dst, CONV_EOL_RENORMALIZE);
}
/*****************************************************************
*
* Streaming conversion support
*
*****************************************************************/
typedef int (*filter_fn)(struct stream_filter *,
const char *input, size_t *isize_p,
char *output, size_t *osize_p);
typedef void (*free_fn)(struct stream_filter *);
struct stream_filter_vtbl {
filter_fn filter;
free_fn free;
};
struct stream_filter {
struct stream_filter_vtbl *vtbl;
};
static int null_filter_fn(struct stream_filter *filter UNUSED,
const char *input, size_t *isize_p,
char *output, size_t *osize_p)
{
size_t count;
if (!input)
return 0; /* we do not keep any states */
count = *isize_p;
if (*osize_p < count)
count = *osize_p;
if (count) {
memmove(output, input, count);
*isize_p -= count;
*osize_p -= count;
}
return 0;
}
static void null_free_fn(struct stream_filter *filter UNUSED)
{
; /* nothing -- null instances are shared */
}
static struct stream_filter_vtbl null_vtbl = {
.filter = null_filter_fn,
.free = null_free_fn,
};
static struct stream_filter null_filter_singleton = {
.vtbl = &null_vtbl,
};
int is_null_stream_filter(struct stream_filter *filter)
{
return filter == &null_filter_singleton;
}
/*
* LF-to-CRLF filter
*/
struct lf_to_crlf_filter {
struct stream_filter filter;
unsigned has_held:1;
char held;
};
static int lf_to_crlf_filter_fn(struct stream_filter *filter,
const char *input, size_t *isize_p,
char *output, size_t *osize_p)
{
size_t count, o = 0;
struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
/*
* We may be holding onto the CR to see if it is followed by a
* LF, in which case we would need to go to the main loop.
* Otherwise, just emit it to the output stream.
*/
if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
output[o++] = lf_to_crlf->held;
lf_to_crlf->has_held = 0;
}
/* We are told to drain */
if (!input) {
*osize_p -= o;
return 0;
}
count = *isize_p;
if (count || lf_to_crlf->has_held) {
size_t i;
int was_cr = 0;
if (lf_to_crlf->has_held) {
was_cr = 1;
lf_to_crlf->has_held = 0;
}
for (i = 0; o < *osize_p && i < count; i++) {
char ch = input[i];
if (ch == '\n') {
output[o++] = '\r';
} else if (was_cr) {
/*
* Previous round saw CR and it is not followed
* by a LF; emit the CR before processing the
* current character.
*/
output[o++] = '\r';
}
/*
* We may have consumed the last output slot,
* in which case we need to break out of this
* loop; hold the current character before
* returning.
*/
if (*osize_p <= o) {
lf_to_crlf->has_held = 1;
lf_to_crlf->held = ch;
continue; /* break but increment i */
}
if (ch == '\r') {
was_cr = 1;
continue;
}
was_cr = 0;
output[o++] = ch;
}
*osize_p -= o;
*isize_p -= i;
if (!lf_to_crlf->has_held && was_cr) {
lf_to_crlf->has_held = 1;
lf_to_crlf->held = '\r';
}
}
return 0;
}
static void lf_to_crlf_free_fn(struct stream_filter *filter)
{
free(filter);
}
static struct stream_filter_vtbl lf_to_crlf_vtbl = {
.filter = lf_to_crlf_filter_fn,
.free = lf_to_crlf_free_fn,
};
static struct stream_filter *lf_to_crlf_filter(void)
{
struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
return (struct stream_filter *)lf_to_crlf;
}
/*
* Cascade filter
*/
#define FILTER_BUFFER 1024
struct cascade_filter {
struct stream_filter filter;
struct stream_filter *one;
struct stream_filter *two;
char buf[FILTER_BUFFER];
int end, ptr;
};
static int cascade_filter_fn(struct stream_filter *filter,
const char *input, size_t *isize_p,
char *output, size_t *osize_p)
{
struct cascade_filter *cas = (struct cascade_filter *) filter;
size_t filled = 0;
size_t sz = *osize_p;
size_t to_feed, remaining;
/*
* input -- (one) --> buf -- (two) --> output
*/
while (filled < sz) {
remaining = sz - filled;
/* do we already have something to feed two with? */
if (cas->ptr < cas->end) {
to_feed = cas->end - cas->ptr;
if (stream_filter(cas->two,
cas->buf + cas->ptr, &to_feed,
output + filled, &remaining))
return -1;
cas->ptr += (cas->end - cas->ptr) - to_feed;
filled = sz - remaining;
continue;
}
/* feed one from upstream and have it emit into our buffer */
to_feed = input ? *isize_p : 0;
if (input && !to_feed)
break;
remaining = sizeof(cas->buf);
if (stream_filter(cas->one,
input, &to_feed,
cas->buf, &remaining))
return -1;
cas->end = sizeof(cas->buf) - remaining;
cas->ptr = 0;
if (input) {
size_t fed = *isize_p - to_feed;
*isize_p -= fed;
input += fed;
}
/* do we know that we drained one completely? */
if (input || cas->end)
continue;
/* tell two to drain; we have nothing more to give it */
to_feed = 0;
remaining = sz - filled;
if (stream_filter(cas->two,
NULL, &to_feed,
output + filled, &remaining))
return -1;
if (remaining == (sz - filled))
break; /* completely drained two */
filled = sz - remaining;
}
*osize_p -= filled;
return 0;
}
static void cascade_free_fn(struct stream_filter *filter)
{
struct cascade_filter *cas = (struct cascade_filter *)filter;
free_stream_filter(cas->one);
free_stream_filter(cas->two);
free(filter);
}
static struct stream_filter_vtbl cascade_vtbl = {
.filter = cascade_filter_fn,
.free = cascade_free_fn,
};
static struct stream_filter *cascade_filter(struct stream_filter *one,
struct stream_filter *two)
{
struct cascade_filter *cascade;
if (!one || is_null_stream_filter(one))
return two;
if (!two || is_null_stream_filter(two))
return one;
cascade = xmalloc(sizeof(*cascade));
cascade->one = one;
cascade->two = two;
cascade->end = cascade->ptr = 0;
cascade->filter.vtbl = &cascade_vtbl;
return (struct stream_filter *)cascade;
}
/*
* ident filter
*/
#define IDENT_DRAINING (-1)
#define IDENT_SKIPPING (-2)
struct ident_filter {
struct stream_filter filter;
struct strbuf left;
int state;
char ident[GIT_MAX_HEXSZ + 5]; /* ": x40 $" */
};
static int is_foreign_ident(const char *str)
{
int i;
if (!skip_prefix(str, "$Id: ", &str))
return 0;
for (i = 0; str[i]; i++) {
if (isspace(str[i]) && str[i+1] != '$')
return 1;
}
return 0;
}
static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
{
size_t to_drain = ident->left.len;
if (*osize_p < to_drain)
to_drain = *osize_p;
if (to_drain) {
memcpy(*output_p, ident->left.buf, to_drain);
strbuf_remove(&ident->left, 0, to_drain);
*output_p += to_drain;
*osize_p -= to_drain;
}
if (!ident->left.len)
ident->state = 0;
}
static int ident_filter_fn(struct stream_filter *filter,
const char *input, size_t *isize_p,
char *output, size_t *osize_p)
{
struct ident_filter *ident = (struct ident_filter *)filter;
static const char head[] = "$Id";
if (!input) {
/* drain upon eof */
switch (ident->state) {
default:
strbuf_add(&ident->left, head, ident->state);
consistently use "fallthrough" comments in switches Gcc 7 adds -Wimplicit-fallthrough, which can warn when a switch case falls through to the next case. The general idea is that the compiler can't tell if this was intentional or not, so you should annotate any intentional fall-throughs as such, leaving it to complain about any unannotated ones. There's a GNU __attribute__ which can be used for annotation, but of course we'd have to #ifdef it away on non-gcc compilers. Gcc will also recognize specially-formatted comments, which matches our current practice. Let's extend that practice to all of the unannotated sites (which I did look over and verify that they were behaving as intended). Ideally in each case we'd actually give some reasons in the comment about why we're falling through, or what we're falling through to. And gcc does support that with -Wimplicit-fallthrough=2, which relaxes the comment pattern matching to anything that contains "fallthrough" (or a variety of spelling variants). However, this isn't the default for -Wimplicit-fallthrough, nor for -Wextra. In the name of simplicity, it's probably better for us to support the default level, which requires "fallthrough" to be the only thing in the comment (modulo some window dressing like "else" and some punctuation; see the gcc manual for the complete set of patterns). This patch suppresses all warnings due to -Wimplicit-fallthrough. We might eventually want to add that to the DEVELOPER Makefile knob, but we should probably wait until gcc 7 is more widely adopted (since earlier versions will complain about the unknown warning type). Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-21 06:25:41 +00:00
/* fallthrough */
case IDENT_SKIPPING:
consistently use "fallthrough" comments in switches Gcc 7 adds -Wimplicit-fallthrough, which can warn when a switch case falls through to the next case. The general idea is that the compiler can't tell if this was intentional or not, so you should annotate any intentional fall-throughs as such, leaving it to complain about any unannotated ones. There's a GNU __attribute__ which can be used for annotation, but of course we'd have to #ifdef it away on non-gcc compilers. Gcc will also recognize specially-formatted comments, which matches our current practice. Let's extend that practice to all of the unannotated sites (which I did look over and verify that they were behaving as intended). Ideally in each case we'd actually give some reasons in the comment about why we're falling through, or what we're falling through to. And gcc does support that with -Wimplicit-fallthrough=2, which relaxes the comment pattern matching to anything that contains "fallthrough" (or a variety of spelling variants). However, this isn't the default for -Wimplicit-fallthrough, nor for -Wextra. In the name of simplicity, it's probably better for us to support the default level, which requires "fallthrough" to be the only thing in the comment (modulo some window dressing like "else" and some punctuation; see the gcc manual for the complete set of patterns). This patch suppresses all warnings due to -Wimplicit-fallthrough. We might eventually want to add that to the DEVELOPER Makefile knob, but we should probably wait until gcc 7 is more widely adopted (since earlier versions will complain about the unknown warning type). Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-21 06:25:41 +00:00
/* fallthrough */
case IDENT_DRAINING:
ident_drain(ident, &output, osize_p);
}
return 0;
}
while (*isize_p || (ident->state == IDENT_DRAINING)) {
int ch;
if (ident->state == IDENT_DRAINING) {
ident_drain(ident, &output, osize_p);
if (!*osize_p)
break;
continue;
}
ch = *(input++);
(*isize_p)--;
if (ident->state == IDENT_SKIPPING) {
/*
* Skipping until '$' or LF, but keeping them
* in case it is a foreign ident.
*/
strbuf_addch(&ident->left, ch);
if (ch != '\n' && ch != '$')
continue;
if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
strbuf_setlen(&ident->left, sizeof(head) - 1);
strbuf_addstr(&ident->left, ident->ident);
}
ident->state = IDENT_DRAINING;
continue;
}
if (ident->state < sizeof(head) &&
head[ident->state] == ch) {
ident->state++;
continue;
}
if (ident->state)
strbuf_add(&ident->left, head, ident->state);
if (ident->state == sizeof(head) - 1) {
if (ch != ':' && ch != '$') {
strbuf_addch(&ident->left, ch);
ident->state = 0;
continue;
}
if (ch == ':') {
strbuf_addch(&ident->left, ch);
ident->state = IDENT_SKIPPING;
} else {
strbuf_addstr(&ident->left, ident->ident);
ident->state = IDENT_DRAINING;
}
continue;
}
strbuf_addch(&ident->left, ch);
ident->state = IDENT_DRAINING;
}
return 0;
}
static void ident_free_fn(struct stream_filter *filter)
{
struct ident_filter *ident = (struct ident_filter *)filter;
strbuf_release(&ident->left);
free(filter);
}
static struct stream_filter_vtbl ident_vtbl = {
.filter = ident_filter_fn,
.free = ident_free_fn,
};
static struct stream_filter *ident_filter(const struct object_id *oid)
{
struct ident_filter *ident = xmalloc(sizeof(*ident));
xsnprintf(ident->ident, sizeof(ident->ident),
": %s $", oid_to_hex(oid));
strbuf_init(&ident->left, 0);
ident->filter.vtbl = &ident_vtbl;
ident->state = 0;
return (struct stream_filter *)ident;
}
/*
* Return an appropriately constructed filter for the given ca, or NULL if
* the contents cannot be filtered without reading the whole thing
* in-core.
*
* Note that you would be crazy to set CRLF, smudge/clean or ident to a
* large binary blob you would want us not to slurp into the memory!
*/
struct stream_filter *get_stream_filter_ca(const struct conv_attrs *ca,
const struct object_id *oid)
{
struct stream_filter *filter = NULL;
if (classify_conv_attrs(ca) != CA_CLASS_STREAMABLE)
return NULL;
if (ca->ident)
filter = ident_filter(oid);
if (output_eol(ca->crlf_action) == EOL_CRLF)
filter = cascade_filter(filter, lf_to_crlf_filter());
else
filter = cascade_filter(filter, &null_filter_singleton);
return filter;
}
struct stream_filter *get_stream_filter(struct index_state *istate,
const char *path,
const struct object_id *oid)
{
struct conv_attrs ca;
convert_attrs(istate, &ca, path);
return get_stream_filter_ca(&ca, oid);
}
void free_stream_filter(struct stream_filter *filter)
{
filter->vtbl->free(filter);
}
int stream_filter(struct stream_filter *filter,
const char *input, size_t *isize_p,
char *output, size_t *osize_p)
{
return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
}
void init_checkout_metadata(struct checkout_metadata *meta, const char *refname,
const struct object_id *treeish,
const struct object_id *blob)
{
memset(meta, 0, sizeof(*meta));
if (refname)
meta->refname = refname;
if (treeish)
oidcpy(&meta->treeish, treeish);
if (blob)
oidcpy(&meta->blob, blob);
}
void clone_checkout_metadata(struct checkout_metadata *dst,
const struct checkout_metadata *src,
const struct object_id *blob)
{
memcpy(dst, src, sizeof(*dst));
if (blob)
oidcpy(&dst->blob, blob);
}
enum conv_attrs_classification classify_conv_attrs(const struct conv_attrs *ca)
{
if (ca->drv) {
if (ca->drv->process)
return CA_CLASS_INCORE_PROCESS;
if (ca->drv->smudge || ca->drv->clean)
return CA_CLASS_INCORE_FILTER;
}
if (ca->working_tree_encoding)
return CA_CLASS_INCORE;
if (ca->crlf_action == CRLF_AUTO || ca->crlf_action == CRLF_AUTO_CRLF)
return CA_CLASS_INCORE;
return CA_CLASS_STREAMABLE;
}