The main readdir loop in read_directory_recursive() is replaced with a
new one that checks if cached results of a directory is still valid.
If a file is added or removed from the index, the containing directory
is invalidated (but not its subdirs). If directory's mtime is changed,
the same happens. If a .gitignore is updated, the containing directory
and all subdirs are invalidated recursively. If dir_struct#flags or
other conditions change, the cache is ignored.
If a directory is invalidated, we opendir/readdir/closedir and run the
exclude machinery on that directory listing as usual. If untracked
cache is also enabled, we'll update the cache along the way. If a
directory is validated, we simply pull the untracked listing out from
the cache. The cache also records the list of direct subdirs that we
have to recurse in. Fully excluded directories are seen as "untracked
files".
In the best case when no dirs are invalidated, read_directory()
becomes a series of
stat(dir), open(.gitignore), fstat(), read(), close() and optionally
hash_sha1_file()
For comparison, standard read_directory() is a sequence of
opendir(), readdir(), open(.gitignore), fstat(), read(), close(), the
expensive last_exclude_matching() and closedir().
We already try not to open(.gitignore) if we know it does not exist,
so open/fstat/read/close sequence does not apply to every
directory. The sequence could be reduced further, as noted in
prep_exclude() in another patch. So in theory, the entire best-case
read_directory sequence could be reduced to a series of stat() and
nothing else.
This is not a silver bullet approach. When you compile a C file, for
example, the old .o file is removed and a new one with the same name
created, effectively invalidating the containing directory's cache
(but not its subdirectories). If your build process touches every
directory, this cache adds extra overhead for nothing, so it's a good
idea to separate generated files from tracked files.. Editors may use
the same strategy for saving files. And of course you're out of luck
running your repo on an unsupported filesystem and/or operating system.
Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This allows us to feed different info to read_directory_recursive()
based on untracked cache in the next patch.
Helped-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It's easy to see that if an existing .gitignore changes, its SHA-1
would be different and invalidate_gitignore() is called.
If .gitignore is removed, add_excludes() will treat it like an empty
.gitignore, which again should invalidate the cached directory data.
if .gitignore is added, lookup_untracked() already fills initial
.gitignore SHA-1 as "empty file", so again invalidate_gitignore() is
called.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Make sure the starting conditions and all global exclude files are
good to go. If not, either disable untracked cache completely, or wipe
out the cache and start fresh.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The idea is if we can capture all input and (non-rescursive) output of
read_directory_recursive(), and can verify later that all the input is
the same, then the second r_d_r() should produce the same output as in
the first run.
The requirement for this to work is stat info of a directory MUST
change if an entry is added to or removed from that directory (and
should not change often otherwise). If your OS and filesystem do not
meet this requirement, untracked cache is not for you. Most file
systems on *nix should be fine. On Windows, NTFS is fine while FAT may
not be [1] even though FAT on Linux seems to be fine.
The list of input of r_d_r() is in the big comment block in dir.h. In
short, the output of a directory (not counting subdirs) mainly depends
on stat info of the directory in question, all .gitignore leading to
it and the check_only flag when r_d_r() is called recursively. This
patch records all this info (and the output) as r_d_r() runs.
Two hash_sha1_file() are required for $GIT_DIR/info/exclude and
core.excludesfile unless their stat data matches. hash_sha1_file() is
only needed when .gitignore files in the worktree are modified,
otherwise their SHA-1 in index is used (see the previous patch).
We could store stat data for .gitignore files so we don't have to
rehash them if their content is different from index, but I think
.gitignore files are rarely modified, so not worth extra cache data
(and hashing penalty read-cache.c:verify_hdr(), as we will be storing
this as an index extension).
The implication is, if you change .gitignore, you better add it to the
index soon or you lose all the benefit of untracked cache because a
modified .gitignore invalidates all subdirs recursively. This is
especially bad for .gitignore at root.
This cached output is about untracked files only, not ignored files
because the number of tracked files is usually small, so small cache
overhead, while the number of ignored files could go really high
(e.g. *.o files mixing with source code).
[1] "Description of NTFS date and time stamps for files and folders"
http://support.microsoft.com/kb/299648
Helped-by: Torsten Bögershausen <tboegi@web.de>
Helped-by: David Turner <dturner@twopensource.com>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is not used anywhere yet. But the goal is to compare quickly if a
.gitignore file has changed when we have the SHA-1 of both old (cached
somewhere) and new (from index or a tree) versions.
Helped-by: Junio C Hamano <gitster@pobox.com>
Helped-by: Torsten Bögershausen <tboegi@web.de>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Various issues around "reflog expire", e.g. using --updateref when
expiring a reflog for a symbolic reference, have been corrected
and/or made saner.
* mh/expire-updateref-fixes:
reflog_expire(): never update a reference to null_sha1
reflog_expire(): ignore --updateref for symbolic references
reflog: improve and update documentation
struct ref_lock: delete the force_write member
lock_ref_sha1_basic(): do not set force_write for missing references
write_ref_sha1(): move write elision test to callers
write_ref_sha1(): remove check for lock == NULL
A corrupt input to "git diff -M" can cause us to segfault.
* jk/diffcore-rename-duplicate:
diffcore-rename: avoid processing duplicate destinations
diffcore-rename: split locate_rename_dst into two functions
The borrowed code in kwset API did not follow our usual convention
to use "unsigned char" to store values that range from 0-255.
* bw/kwset-use-unsigned:
kwset: use unsigned char to store values with high-bit set
The interaction between "git submodule update" and the
submodule.*.update configuration was not clearly documented.
* ms/submodule-update-config-doc:
submodule: improve documentation of update subcommand
The prompt string "remove?" used when "git clean -i" asks the user
if a path should be removed was localizable, but the code always
expects a substring of "yes" to tell it to go ahead. Always show
[y/N] as part of this prompt to hint that the answer is not (yet)
localized.
* ja/clean-confirm-i18n:
Add hint interactive cleaning
"git diff --shortstat --dirstat=changes" showed a dirstat based on
lines that was never asked by the end user in addition to the
dirstat that the user asked for.
* mk/diff-shortstat-dirstat-fix:
diff --shortstat --dirstat: remove duplicate output
"git remote add" mentioned "--tags" and "--no-tags" and was not
clear that fetch from the remote in the future will use the default
behaviour when neither is given to override it.
* mg/doc-remote-tags-or-not:
git-remote.txt: describe behavior without --tags and --no-tags
Description given by "grep -h" for its --exclude-standard option
was phrased poorly.
* nd/grep-exclude-standard-help-fix:
grep: correct help string for --exclude-standard
Code cleanups.
* rs/simple-cleanups:
sha1_name: use strlcpy() to copy strings
pretty: use starts_with() to check for a prefix
for-each-ref: use skip_prefix() to avoid duplicate string comparison
connect: use strcmp() for string comparison
The configuration variable 'mailinfo.scissors' was hard to
discover in the documentation.
* mm/am-c-doc:
Documentation/git-am.txt: mention mailinfo.scissors config variable
Documentation/config.txt: document mailinfo.scissors
Correct a breakage to git-svn around v2.2 era that triggers
premature closing of FileHandle.
* ew/svn-maint-fixes:
Git::SVN::*: avoid premature FileHandle closure
git-svn: fix localtime=true on non-glibc environments
Even though we officially haven't dropped Perl 5.8 support, the
Getopt::Long package that came with it does not support "--no-"
prefix to negate a boolean option; manually add support to help
people with older Getopt::Long package.
* km/send-email-getopt-long-workarounds:
git-send-email.perl: support no- prefix with older GetOptions
"update-index --refresh" used to leak when an entry cannot be
refreshed for whatever reason.
* sb/plug-leak-in-make-cache-entry:
read-cache.c: free cache entry when refreshing fails
"git fast-import" used to crash when it could not close and
conclude the resulting packfile cleanly.
* jk/fast-import-die-nicely-fix:
fast-import: avoid running end_packfile recursively
In v2.2.0, we broke "git prune" that runs in a repository that
borrows from an alternate object store.
* jk/prune-mtime:
sha1_file: fix iterating loose alternate objects
for_each_loose_file_in_objdir: take an optional strbuf path
Certain older vintages of cURL give irregular output from
"curl-config --vernum", which confused our build system.
* tc/curl-vernum-output-broken-in-7.11:
Makefile: handle broken curl version number in version check
An earlier workaround to squelch unhelpful deprecation warnings
from the complier on Mac OSX unnecessarily set minimum required
version of the OS, which the user might want to raise (or lower)
for other reasons.
* es/squelch-openssl-warnings-on-macosx:
git-compat-util: do not step on MAC_OS_X_VERSION_MIN_REQUIRED
Longstanding configuration variable naming rules has been added to
the documentation.
* jc/conf-var-doc:
CodingGuidelines: describe naming rules for configuration variables
config.txt: mark deprecated variables more prominently
config.txt: clarify that add.ignore-errors is deprecated
The credential helper for Windows (in contrib/) used to mishandle
a user name with an at-sign in it.
* av/wincred-with-at-in-username-fix:
wincred: fix get credential if username has "@"
Older GnuPG implementations may not correctly import the keyring
material we prepare for the tests to use.
* ch/new-gpg-drops-rfc-1991:
t/lib-gpg: sanity-check that we can actually sign
t/lib-gpg: include separate public keys in keyring.gpg
Clarify in the documentation that "remote.<nick>.pushURL" and
"remote.<nick>.URL" are there to name the same repository accessed
via different transports, not two separate repositories.
* jc/remote-set-url-doc:
Documentation/git-remote.txt: stress that set-url is not for triangular
Reading configuration from a blob object, when it ends with a lone
CR, use to confuse the configuration parser.
* jk/config-no-ungetc-eof:
config_buf_ungetc: warn when pushing back a random character
config: do not ungetc EOF
We didn't format an integer that wouldn't fit in "int" but in
"uintmax_t" correctly.
* jk/decimal-width-for-uintmax:
decimal_width: avoid integer overflow
"git push --signed" gave an incorrectly worded error message when
the other side did not support the capability.
* jc/push-cert:
transport-helper: fix typo in error message when --signed is not supported
"git fetch" over a remote-helper that cannot respond to "list"
command could not fetch from a symbolic reference e.g. HEAD.
* mh/deref-symref-over-helper-transport:
transport-helper: do not request symbolic refs to remote helpers
The insn sheet "git rebase -i" creates did not fully honor
core.abbrev settings.
* ks/rebase-i-abbrev:
rebase -i: use full object name internally throughout the script
The tests that wanted to see that file becomes unreadable after
running "chmod a-r file", and the tests that wanted to make sure it
is not run as root, we used "can we write into the / directory?" as
a cheap substitute, but on some platforms that is not a good
heuristics. The tests and their prerequisites have been updated to
check what they really require.
* jk/sanity:
test-lib.sh: set prerequisite SANITY by testing what we really need
tests: correct misuses of POSIXPERM
t/lib-httpd: switch SANITY check for NOT_ROOT
