The split-index mode had a few corner case bugs fixed.
* tg/split-index-fixes:
travis: run tests with GIT_TEST_SPLIT_INDEX
split-index: don't write cache tree with null oid entries
read-cache: fix reading the shared index for other repos
Update Coccinelle rules to catch and optimize strbuf_addf(&buf, "%s", str)
* rs/strbuf-cocci-workaround:
cocci: use format keyword instead of a literal string
Crash fix for a corner case where an error codepath tried to unlock
what it did not acquire lock on.
* mr/packed-ref-store-fix:
files_initial_transaction_commit(): only unlock if locked
The http tracing code, often used to debug connection issues,
learned to redact potentially sensitive information from its output
so that it can be more safely sharable.
* jt/http-redact-cookies:
http: support omitting data from traces
http: support cookie redaction when tracing
The tracing machinery learned to report tweaking of environment
variables as well.
* nd/trace-with-env:
run-command.c: print new cwd in trace_run_command()
run-command.c: print env vars in trace_run_command()
run-command.c: print program 'git' when tracing git_cmd mode
run-command.c: introduce trace_run_command()
trace.c: move strbuf_release() out of print_trace_line()
trace: avoid unnecessary quoting
sq_quote_argv: drop maxlen parameter
Rewrite two more "git submodule" subcommands in C.
* pc/submodule-helper:
submodule: port submodule subcommand 'deinit' from shell to C
submodule: port submodule subcommand 'sync' from shell to C
Avoid showing a warning message in the middle of a line of "git
diff" output.
* nd/diff-flush-before-warning:
diff.c: flush stdout before printing rename warnings
Doc updates.
* ks/submodule-doc-updates:
Doc/git-submodule: improve readability and grammar of a sentence
Doc/gitsubmodules: make some changes to improve readability and syntax
Retire mru API as it does not give enough abstraction over
underlying list API to be worth it.
* gs/retire-mru:
mru: Replace mru.[ch] with list.h implementation
The first step to getting rid of mru API and using the
doubly-linked list API directly instead.
* ot/mru-on-list:
mru: use double-linked list from list.h
The machinery to clone & fetch, which in turn involves packing and
unpacking objects, have been told how to omit certain objects using
the filtering mechanism introduced by the jh/object-filtering
topic, and also mark the resulting pack as a promisor pack to
tolerate missing objects, taking advantage of the mechanism
introduced by the jh/fsck-promisors topic.
* jh/partial-clone:
t5616: test bulk prefetch after partial fetch
fetch: inherit filter-spec from partial clone
t5616: end-to-end tests for partial clone
fetch-pack: restore save_commit_buffer after use
unpack-trees: batch fetching of missing blobs
clone: partial clone
partial-clone: define partial clone settings in config
fetch: support filters
fetch: refactor calculation of remote list
fetch-pack: test support excluding large blobs
fetch-pack: add --no-filter
fetch-pack, index-pack, transport: partial clone
upload-pack: add object filtering for partial clone
In preparation for implementing narrow/partial clone, the machinery
for checking object connectivity used by gc and fsck has been
taught that a missing object is OK when it is referenced by a
packfile specially marked as coming from trusted repository that
promises to make them available on-demand and lazily.
* jh/fsck-promisors:
gc: do not repack promisor packfiles
rev-list: support termination at promisor objects
sha1_file: support lazily fetching missing objects
introduce fetch-object: fetch one promisor object
index-pack: refactor writing of .keep files
fsck: support promisor objects as CLI argument
fsck: support referenced promisor objects
fsck: support refs pointing to promisor objects
fsck: introduce partialclone extension
extension.partialclone: introduce partial clone extension
The build procedure for perl/ part has been greatly simplified by
weaning ourselves off of MakeMaker.
* ab/simplify-perl-makefile:
perl: treat PERLLIB_EXTRA as an extra path again
perl: avoid *.pmc and fix Error.pm further
Makefile: replace perl/Makefile.PL with simple make rules
Small changes in messages to fit the style and typography of rest.
Reuse already translated messages if possible.
Do not translate messages aimed at developers of git.
Fix unit tests depending on the original string.
Use `test_i18ngrep` for tests with translatable strings.
Change and verify rest of tests via `make GETTEXT_POISON=1 test`.
Signed-off-by: Alexander Shopov <ash@kambanaria.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the description of git interpret-trailers, we describe "a group…of
lines" that have certain characteristics. Ensure both options
describing this group use a singular verb for parallelism.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In check_ignore(), the first pathspec item determines the dtype for any
subsequent ones. That means that a pathspec matching a regular file can
prevent following pathspecs from matching directories, which makes no
sense. Fix that by determining the dtype for each pathspec separately,
by passing the value DT_UNKNOWN to last_exclude_matching() each time.
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Prior to 644eb60bd0 (builtin/describe.c: describe a blob,
2017-11-15), we noticed and complained about missing
objects, since they were not valid commits:
$ git describe 0000000000000000000000000000000000000000
fatal: 0000000000000000000000000000000000000000 is not a valid 'commit' object
After that commit, we feed any non-commit to lookup_blob(),
and complain only if it returns NULL. But the lookup_*
functions do not actually look at the on-disk object
database at all. They return an entry from the in-memory
object hash if present (and if it matches the requested
type), and otherwise auto-create a "struct object" of the
requested type.
A missing object would hit that latter case: we create a
bogus blob struct, walk all of history looking for it, and
then exit successfully having produced no output.
One reason nobody may have noticed this is that some related
cases do still work OK:
1. If we ask for a tree by sha1, then the call to
lookup_commit_referecne_gently() would have parsed it,
and we would have its true type in the in-memory object
hash.
2. If we ask for a name that doesn't exist but isn't a
40-hex sha1, then get_oid() would complain before we
even look at the objects at all.
We can fix this by replacing the lookup_blob() call with a
check of the true type via sha1_object_info(). This is not
quite as efficient as we could possibly make this check. We
know in most cases that the object was already parsed in the
earlier commit lookup, so we could call lookup_object(),
which does auto-create, and check the resulting struct's
type (or NULL). However it's not worth the fragility nor
code complexity to save a single object lookup.
The new tests cover this case, as well as that of a
tree-by-sha1 (which does work as described above, but was
not explicitly tested).
Noticed-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Sparse has, for a long time, been issuing the following warning against
the pack-revindex.c file:
SP pack-revindex.c
pack-revindex.c:64:23: warning: memset with byte count of 262144
This results from a unconditional check, with a hard-coded limit, which
is really only appropriate for the kernel source code. (The check is for
a 'large' byte count in a call to memcpy(), memset(), copy_from_user()
and copy_to_user() functions).
A recent release of sparse (v0.5.1) has introduced some options to allow
this check to be turned off (-Wno-memcpy-max-count) or to specify the
actual limit used (-fmemcpy-max-count=COUNT), rather than a hard-coded
limit of 100000.
In order to suppress the warning, add a target for pack-revindex.sp that
adds the '-Wno-memcpy-max-count' option to the SPARSE_FLAGS variable.
Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since commit f66450ae9 ("cygwin: Remove the Win32 l/stat() implementation",
2013-06-22), the cygwin build has not used the WIN32 API/header files.
This means that the '-isystem /usr/include/w32api' option to sparse is
no longer necessary (to allow sparse to find the WIN32 header files).
In addition, the '-Wno-one-bit-signed-bitfield' option can be removed,
since the warning suppressed by that option was only provoked by a WIN32
header file.
Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This ancient test script does a lot of manual checking of
test conditions with "if" blocks. We can simplify this
by relying on helpers like test_must_fail.
Note that a failing "grep" call here won't produce any
verbose output, but that's OK. These days we rely on "-x" to
tell us about such commands. And in addition, these greps
are soon to be converted to test_i18ngrep (which is itself
soon learning to be more verbose).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 'test_might_fail' is implemented as a thin wrapper around
'test_must_fail', it also accepts the same options. Mention this in
the docs as well.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Note the caveat where 2.17 is stricter about index validation
potentially causing "could not open directory" warnings when git is
upgraded. See the preceding "dir.c: stop ignoring opendir() error in
open_cached_dir()" change.
This caused some mayhem when I upgraded git to a version with this
series at Booking.com, and other users have doubtless enabled the UC
extension and are in for a surprise when they upgrade. Let's give them
a headsup in the docs.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Document the bug tested for in my "status: add a failing test showing
a core.untrackedCache bug" and fixed in Duy's "dir.c: fix missing dir
invalidation in untracked code".
Since this is very likely something others will encounter in the
future on older versions, and it's not obvious how to fix it let's
document both that it exists, and how to "fix" it with a one-off
command.
As noted in that commit, even though this bug gets the untracked cache
into a bad state, we have not yet found a case where this is user
visible, and thus it makes sense for these docs to focus on the
symlink case only.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We moved away from SHA1_HEADER to a preprocessor if chain, but didn't
update the comment discussing the platform defines. Update this comment
so it reflects the current state of our codebase.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since commit dd6fb0053 ("rebase -p: fix quoting when calling `git
merge`"), commit message of the merge commit being rebased is passed to
the merge command using a subshell executing 'git rev-parse --sq-quote'.
Double quotes are needed around this subshell so that, newlines are
kept for the git merge command.
Before this patch, following merge message:
"Merge mybranch into mynewbranch
Awesome commit."
becomes:
"Merge mybranch into mynewbranch Awesome commit."
after a rebase -p.
Fixes: "dd6fb0053 rebase -p: fix quoting when calling `git merge`"
Reported-by: Jamie Iles <jamie.iles@oracle.com>
Suggested-by: Vegard Nossum <vegard.nossum@oracle.com>
Suggested-by: Quentin Casasnovas <quentin.casasnovas@oracle.com>
Signed-off-by: Gregory Herrero <gregory.herrero@oracle.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It perhaps goes without saying that file-local stuff should
be marked static, but it does not hurt to remind people.
Less obvious is that we are settling on "do not include
extern in function declarations". It is already the default
unless the function was previously declared static (but if
you are following a static declaration with an unmarked one,
you should think about why you are declaring the thing
twice). And so it just becomes an extra noise-word in our
header files.
We used to give the opposite advice, so there are quite a
few "extern" markers in early Git code. But this at least
makes a concrete suggestion that we can follow going
forward.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The packet_read_line() function will die if it sees any
protocol or socket errors. But it will return NULL for a
flush packet; some callers which are not expecting this may
dereference NULL if they get an unexpected flush. This would
involve the other side breaking protocol, but we should
flag the error rather than segfault.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The packet_read_line() function dies if it gets an
unexpected EOF. It only returns NULL if we get a flush
packet (or technically, a zero-length "0004" packet, but
nobody is supposed to send those, and they are
indistinguishable from a flush in this interface).
Let's correct error messages which claim an unexpected EOF;
it's really an unexpected flush packet.
While we're here, let's also check "!line" instead of
"!len" in the second case. The two events should always
coincide, but checking "!line" makes it more obvious that we
are not about to dereference NULL.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Correct the pointer arithmetic in adjust_dirname_case() so that it calls
find_dir_entry() with the correct string length. Previously passing in
"dir1/foo" would pass a length of 6 instead of the correct 4. This resulted in
find_dir_entry() never finding the entry and so the subsequent memcpy that would
fold the name to the version with the correct case never executed.
Add a test to validate the corrected behavior with name folding of directories.
Signed-off-by: Ben Peart <benpeart@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When 'test_i18ngrep' can't find the expected pattern, it exits
completely silently; when its negated form does find the pattern that
shouldn't be there, it prints the matching line(s) but otherwise exits
without any error message. This leaves the developer puzzled about
what could have gone wrong.
Make 'test_i18ngrep' more informative on failure by printing an error
message including the invoked 'grep' command and the contents of the
file it had to scan through.
Note that this "dump the scanned file" part is not quite perfect, as
it dumps only the file specified as the function's last positional
parameter, thus assuming that there is only a single file parameter.
I think that's a reasonable assumption to make, one that holds true in
the current code base. And even if someone were to scan multiple
files at once in the future, the worst thing that could happen is that
the verbose error message won't include the contents of all those
files, only the last one. Alas, we can't really do any better than
this, because checking whether the other positional parameters match a
filename can result in false positives: 't3400-rebase.sh' and
't3404-rebase-interactive.sh' contain one test each, where the
'test_i18ngrep's pattern verbatimly matches a file in the trash
directory.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some of the previous patches in this series fixed bogus
'test_i18ngrep' invocations:
- Two invocations where the tested git command's standard output is
directly piped into 'test_i18ngrep'. While convenient, this is an
antipattern, because the pipe hides the git command's exit code,
and the test could continue even if the command exited with error.
- Two invocations that had neither a filename parameter nor anything
piped into their standard input, yet both managed to remain
unnoticed for years. A third similarly bogus invocation is
currently lurking in 'pu' for a couple of weeks now.
Prevent similar mistakes in the future by validating 'test_i18ngrep's
parameters requiring that
- The last parameter names an existing file to be read, effectively
forbidding piping into 'test_i18ngrep'.
Note that this change will also forbid cases where 'test_i18ngrep'
would legitimately read its standard input, e.g. when its standard
input is redirected from a file, or when a git command's standard
output is first written to an intermediate file, which is then
preprocessed by a non-git command before the results are piped
into 'test_i18ngrep'. See two of the previous patches for the
only such cases we had in our test suite. However, reliably
preventing the piping antipattern is arguably more important than
supporting these cases, which can be easily worked around by
opening the file directly or using an intermediate file anyway.
- There are at least two parameters, not including the optional '!'
to negate the pattern. This ought to catch corner cases when
'test_i18ngrep' looks for the name of an existing file on its
standard input; the above check would miss this case becase the
filename as pattern would be the last parameter.
Note that this is not quite perfect, as it doesn't account for any
'grep --options' given as parameters. However, doing so would be
far too complicated, considering that patterns can start with
dashes as well, and in the majority of the cases we don't use any
such options anyway.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Both 'test_i18ncmp' and 'test_i18ngrep' helper functions are supposed
to be called from our test scripts, so they should be in
'test-lib-functions.sh'.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Redirecting 'test_i18ngrep's standard input from a file will interfere
with the linting that will be added in a later patch.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
One of the tests in 't5510-fetch.sh' checks the output of 'git fetch'
using 'test_i18ngrep', and while doing so it prefilters the output
with 'grep' before piping the result into 'test_i18ngrep'.
This prefiltering is unnecessary, with the appropriate pattern
'test_i18ngrep' can do it all by itself. Furthermore, piping data
into 'test_i18ngrep' will interfere with the linting that will be
added in a later patch.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The primary purpose of three tests in 't4001-diff-rename.sh' is to
check rename detection in 'git status', but all three do so by running
'git status' upstream of a pipe, hiding its exit code. Consequently,
the test could continue even if 'git status' exited with error.
Use an intermediate file between 'git status' and 'test_i18ngrep' to
catch a potential failure of the former.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The primary purpose of 't6022-merge-rename.sh' is to test 'git merge',
but one of the tests runs it upstream of a pipe, hiding its exit code.
Consequently, the test could continue even if 'git merge' exited with
error.
Use an intermediate file between 'git merge' and 'test_i18ngrep' to
catch a potential failure of the former.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The second 'test_i18ngrep' invocation in the test 'curl redirects
respect whitelist' is missing its filename parameter. This has
remained unnoticed since its introduction in f4113cac0 (http: limit
redirection to protocol-whitelist, 2015-09-22), because it would only
cause the test to fail if Git was built with a sufficiently old
libcurl version. The test's two ||-chained 'test_i18ngrep'
invocations are supposed to check that either one of the two patterns
is present in 'git clone's error message. As it happens, the first
invocation covers the error message from any reasonably up-to-date
libcurl, thus the second invocation, the one without the filename
parameter, isn't executed at all. Apparently no one has run the test
suite's httpd tests with such an old libcurl in the last 2+ years, or
at least they haven't bothered to notify us about the failed test.
Fix this by consolidating the two patterns into a single extended
regexp, eliminating the need for an ||-chained second 'test_i18ngrep'
invocation.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The test 'push --no-progress silences progress but not status' runs
'test_i18ngrep' without specifying a filename parameter. This has
remained unnoticed since its introduction in e304aeba2 (t5541: test
more combinations of --progress, 2012-05-01), because that
'test_i18ngrep' is supposed to check that the given pattern is not
present in its input, and of course it won't find that pattern if its
input is empty (as it comes from /dev/null). This also means that
this test could miss a potential breakage of 'git push --no-progress'.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Running "make NO_GETTEXT=1 GETTEXT_POISON=1" currently fails
t0205.
While it might seem nonsensical at first glance to both
poison and disable gettext, it's useful to be able to do a
poison test-run on a system that doesn't have gettext at
all. And it works fine for C programs; the problem is only
with the shell code.
The issue is that we check the baked-in USE_GETTEXT_SCHEME
value before GETTEXT_POISON. And when NO_GETTEXT is set, the
Makefile sets USE_GETTEXT_SCHEME to "fallthrough".
So one fix would be to have the Makefile just set
USE_GETTEXT_SCHEME to "poison" if GETTEXT_POISON is set.
But there are two problems with that:
1. USE_GETTEXT_SCHEME is actually a user-facing knob, so
conceivably somebody could override it with:
make USE_GETTEXT_SCHEME=gnu GETTEXT_POISON=1
which would do the wrong thing (though that's much less
likely than them having the variable set in their
config.mak and just overriding GETTEXT_POISON on the
command-line for a one-off test).
2. We don't actually bake GETTEXT_POISON in to the shell
library like we do for the C code. It checks
$GIT_GETTEXT_POISON at runtime, which is set up by the
test suite. So it makes sense to put the fix in the
runtime code, too, which would cover something like:
GIT_GETTEXT_POISON=foo git foo
It's not likely that people use the poison code outside
of running the test suite, but it's easy enough to make
this case work.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We check that a shell variable is non-empty, and then we
check that it's equal to a particular value. Just checking
the latter covers both cases.
I suspect the original was trying to give better output when
the test fails, but using "-x" covers that these days.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The command line parser of "git blame" is prepared to take an
ancient odd argument order "blame <path> <rev>" in addition to the
usual "blame [<rev>] <path>". It has at least two negative
ramifications:
- In order to tell these two apart, it checks if the last command
line argument names a path in the working tree, using
file_exists(). However, "blame <rev> <path>" is a request to
explain each and every line in the contents of <path> stored in
revision <rev> and does not need to have a working tree version
of the file. A check with file_exists() is simply wrong.
- To coerce that mistaken file_exists() check to work, the code
calls setup_work_tree() before doing so, because the path it has
is relative to the top-level of the project tree. However,
"blame <rev> <path>" MUST be usable even in a bare repository,
and there is no reason for letting setup_work_tree() complain
and die with "This operation must be run in a work tree".
To correct the former, switch to check if the last token is a
revision (and if so, parse arguments using "blame <path> <rev>"
rule). Correct the latter by getting rid of setup_work_tree() and
file_exists() check--the only case the call to this function matters
is when we are running "blame <path>" (i.e. no starting revision and
asking to blame the working tree file at <path>, digging through the
HEAD revision), but there is a call in setup_scoreboard() just
before it calls fake_working_tree_commit().
Signed-off-by: Junio C Hamano <gitster@pobox.com>
read_directory() code ignores all paths named ".git" even if it's not
a valid git repository. See treat_path() for details. Since ".git" is
basically invisible to read_directory(), when we are asked to
invalidate a path that contains ".git", we can safely ignore it
because the slow path would not consider it anyway.
This helps when fsmonitor is used and we have a real ".git" repo at
worktree top. Occasionally .git/index will be updated and if the
fsmonitor hook does not filter it, untracked cache is asked to
invalidate the path ".git/index".
Without this patch, we invalidate the root directory unncessarily,
which:
- makes read_directory() fall back to slow path for root directory
(slower)
- makes the index dirty (because UNTR extension is updated). Depending
on the index size, writing it down could also be slow.
A note about the new "safe_path" knob. Since this new check could be
relatively expensive, avoid it when we know it's not needed. If the
path comes from the index, it can't contain ".git". If it does
contain, we may be screwed up at many more levels, not just this one.
Noticed-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
