| Age | Commit message (Collapse) | Author |
|---|
|
* maint-2.29:
Git 2.29.3
Git 2.28.1
Git 2.27.1
Git 2.26.3
Git 2.25.5
Git 2.24.4
Git 2.23.4
Git 2.22.5
Git 2.21.4
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.28:
Git 2.28.1
Git 2.27.1
Git 2.26.3
Git 2.25.5
Git 2.24.4
Git 2.23.4
Git 2.22.5
Git 2.21.4
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.27:
Git 2.27.1
Git 2.26.3
Git 2.25.5
Git 2.24.4
Git 2.23.4
Git 2.22.5
Git 2.21.4
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.26:
Git 2.26.3
Git 2.25.5
Git 2.24.4
Git 2.23.4
Git 2.22.5
Git 2.21.4
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.25:
Git 2.25.5
Git 2.24.4
Git 2.23.4
Git 2.22.5
Git 2.21.4
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.24:
Git 2.24.4
Git 2.23.4
Git 2.22.5
Git 2.21.4
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.23:
Git 2.23.4
Git 2.22.5
Git 2.21.4
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.22:
Git 2.22.5
Git 2.21.4
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.21:
Git 2.21.4
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.20:
Git 2.20.5
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.19:
Git 2.19.6
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.18:
Git 2.18.5
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
* maint-2.17:
Git 2.17.6
unpack_trees(): start with a fresh lstat cache
run-command: invalidate lstat cache after a command finished
checkout: fix bug that makes checkout follow symlinks in leading path
|
|
Before checking out a file, we have to confirm that all of its leading
components are real existing directories. And to reduce the number of
lstat() calls in this process, we cache the last leading path known to
contain only directories. However, when a path collision occurs (e.g.
when checking out case-sensitive files in case-insensitive file
systems), a cached path might have its file type changed on disk,
leaving the cache on an invalid state. Normally, this doesn't bring
any bad consequences as we usually check out files in index order, and
therefore, by the time the cached path becomes outdated, we no longer
need it anyway (because all files in that directory would have already
been written).
But, there are some users of the checkout machinery that do not always
follow the index order. In particular: checkout-index writes the paths
in the same order that they appear on the CLI (or stdin); and the
delayed checkout feature -- used when a long-running filter process
replies with "status=delayed" -- postpones the checkout of some entries,
thus modifying the checkout order.
When we have to check out an out-of-order entry and the lstat() cache is
invalid (due to a previous path collision), checkout_entry() may end up
using the invalid data and thrusting that the leading components are
real directories when, in reality, they are not. In the best case
scenario, where the directory was replaced by a regular file, the user
will get an error: "fatal: unable to create file 'foo/bar': Not a
directory". But if the directory was replaced by a symlink, checkout
could actually end up following the symlink and writing the file at a
wrong place, even outside the repository. Since delayed checkout is
affected by this bug, it could be used by an attacker to write
arbitrary files during the clone of a maliciously crafted repository.
Some candidate solutions considered were to disable the lstat() cache
during unordered checkouts or sort the entries before passing them to
the checkout machinery. But both ideas include some performance penalty
and they don't future-proof the code against new unordered use cases.
Instead, we now manually reset the lstat cache whenever we successfully
remove a directory. Note: We are not even checking whether the directory
was the same as the lstat cache points to because we might face a
scenario where the paths refer to the same location but differ due to
case folding, precomposed UTF-8 issues, or the presence of `..`
components in the path. Two regression tests, with case-collisions and
utf8-collisions, are also added for both checkout-index and delayed
checkout.
Note: to make the previously mentioned clone attack unfeasible, it would
be sufficient to reset the lstat cache only after the remove_subtree()
call inside checkout_entry(). This is the place where we would remove a
directory whose path collides with the path of another entry that we are
currently trying to check out (possibly a symlink). However, in the
interest of a thorough fix that does not leave Git open to
similar-but-not-identical attack vectors, we decided to intercept
all `rmdir()` calls in one fell swoop.
This addresses CVE-2021-21300.
Co-authored-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br>
|
|
Build update.
* da/vs-build-iconv-fix:
ci(vs-build): stop passing the iconv library location explicitly
|
|
Something changed in `vcpkg` (which we use in our Visual C++ build to
provide the dependencies such as libcurl) and our `vs-build` job started
failing in CI. The reason is that we had a work-around in place to help
CMake find iconv, and this work-around is neither needed nor does it
work anymore.
For the full discussion with the vcpkg project, see this comment:
https://github.com/microsoft/vcpkg/issues/14780#issuecomment-735368280
Signed-off-by: Dennis Ameling <dennis@dennisameling.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
In 1af265f0 (compat/bswap.h: simplify MSVC endianness
detection, 2020-11-08) we attempted to simplify code by assuming MSVC
builds will be for little-endian machines, since only unusably old
versions of MSVC supported big-endian MIPS and m68k architectures.
However, it's possible that MSVC could be ported to build for a
big-endian architecture again, so the simplification wasn't as
future-proof as hoped.
So let's go back to the old way of detecting MSVC, and then checking
architecture from a list of little-endian architecture macros.
Note that MSVC does not treat ARM64 as bi-endian, so we can safely treat
it as little-endian.
Helped-by: brian m. carlson <sandals@crustytoothpaste.net>
Helped-by: Jeff King <peff@peff.net>
Helped-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Helped-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Daniel Gurney <dgurney99@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Modern MSVC or Windows versions don't support big-endian, so it's
unnecessary to consider architectures when using it.
This also makes ARM64 MSVC builds succeed.
Helped-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Daniel Gurney <dgurney99@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
In 554544276a (*.[ch]: remove extern from function declarations using
spatch, 2019-04-29), `extern` on function declarations were declared to
be redundant and thus removed from the codebase. An `extern` was
accidentally reintroduced in 08809c09aa (mingw: add a helper function to
attach GDB to the current process, 2020-02-13).
Remove this spurious `extern`.
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Compilation fix around type punning.
* jk/drop-unaligned-loads:
Revert "fast-export: use local array to store anonymized oid"
bswap.h: drop unaligned loads
|
|
Our put_be32() routine and its variants (get_be32(), put_be64(), etc)
has two implementations: on some platforms we cast memory in place and
use nothl()/htonl(), which can cause unaligned memory access. And on
others, we pick out the individual bytes using bitshifts.
This introduces extra complexity, and sometimes causes compilers to
generate warnings about type-punning. And it's not clear there's any
performance advantage.
This split goes back to 660231aa97 (block-sha1: support for
architectures with memory alignment restrictions, 2009-08-12). The
unaligned versions were part of the original block-sha1 code in
d7c208a92e (Add new optimized C 'block-sha1' routines, 2009-08-05),
which says it is:
Based on the mozilla SHA1 routine, but doing the input data accesses a
word at a time and with 'htonl()' instead of loading bytes and shifting.
Back then, Linus provided timings versus the mozilla code which showed a
27% improvement:
https://lore.kernel.org/git/alpine.LFD.2.01.0908051545000.3390@localhost.localdomain/
However, the unaligned loads were either not the useful part of that
speedup, or perhaps compilers and processors have changed since then.
Here are times for computing the sha1 of 4GB of random data, with and
without -DNO_UNALIGNED_LOADS (and BLK_SHA1=1, of course). This is with
gcc 10, -O2, and the processor is a Core i9-9880H.
[stock]
Benchmark #1: t/helper/test-tool sha1 <foo.rand
Time (mean ± σ): 6.638 s ± 0.081 s [User: 6.269 s, System: 0.368 s]
Range (min … max): 6.550 s … 6.841 s 10 runs
[-DNO_UNALIGNED_LOADS]
Benchmark #1: t/helper/test-tool sha1 <foo.rand
Time (mean ± σ): 6.418 s ± 0.015 s [User: 6.058 s, System: 0.360 s]
Range (min … max): 6.394 s … 6.447 s 10 runs
And here's the same test run on an AMD A8-7600, using gcc 8.
[stock]
Benchmark #1: t/helper/test-tool sha1 <foo.rand
Time (mean ± σ): 11.721 s ± 0.113 s [User: 10.761 s, System: 0.951 s]
Range (min … max): 11.509 s … 11.861 s 10 runs
[-DNO_UNALIGNED_LOADS]
Benchmark #1: t/helper/test-tool sha1 <foo.rand
Time (mean ± σ): 11.744 s ± 0.066 s [User: 10.807 s, System: 0.928 s]
Range (min … max): 11.637 s … 11.863 s 10 runs
So the unaligned loads don't seem to help much, and actually make things
worse. It's possible there are platforms where they provide more
benefit, but:
- the non-x86 platforms for which we use this code are old and obscure
(powerpc and s390).
- the main caller that cares about performance is block-sha1. But
these days it is rarely used anyway, in favor of sha1dc (which is
already much slower, and nobody seems to have cared that much).
Let's just drop unaligned versions entirely in the name of simplicity.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Signed-off-by: Orgad Shaneh <orgads@gmail.com>
Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Signed-off-by: Orgad Shaneh <orgads@gmail.com>
Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
"unlink" emulation on MinGW has been optimized.
* jh/mingw-unlink:
mingw: improve performance of mingw_unlink()
|
|
Update mingw_unlink() to first try to delete the file with existing
permissions before trying to force it.
Windows throws an error when trying to delete a read-only file. The
mingw_unlink() compatibility wrapper always tries to _wchmod(666) the
file before calling _wunlink() to avoid that error. However, since
most files in the worktree are already writable, this is usually
wasted effort.
Update mingw_unlink() to just call DeleteFileW() directly and if that
succeeds return. If that fails, fall back into the existing code path
to update the permissions and use _wunlink() to get the existing
error code mapping.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
We eventually want to drop the argv_array name and just use strvec
consistently. There's no particular reason we have to do it all at once,
or care about interactions between converted and unconverted bits.
Because of our preprocessor compat layer, the names are interchangeable
to the compiler (so even a definition and declaration using different
names is OK).
This patch converts remaining files from the first half of the alphabet,
to keep the diff to a manageable size.
The conversion was done purely mechanically with:
git ls-files '*.c' '*.h' |
xargs perl -i -pe '
s/ARGV_ARRAY/STRVEC/g;
s/argv_array/strvec/g;
'
and then selectively staging files with "git add '[abcdefghjkl]*'".
We'll deal with any indentation/style fallouts separately.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Workaround breakage in MSVC build, where "curl-config --cflags"
gives settings appropriate for GCC build.
* js/msvc-build-fix:
msvc: fix "REG_STARTEND" issue
|
|
In 897d68e7af82 (Makefile: use curl-config --cflags, 2020-03-26), we
taught the build process to use `curl-config --cflags` to make sure that
it can find cURL's headers.
In the MSVC build, this is completely bogus because we're running in a
Git for Windows SDK whose `curl-config` supports the _GCC_ build.
Let's just ignore each and every `-I<path>` option where `<path>` points
to GCC/Clang specific headers.
Reported by Jeff Hostetler in
https://github.com/microsoft/git/issues/275.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Code clean-up by removing a compatibility implementation of a
function we no longer use.
* cb/no-more-gmtime:
compat: remove gmtime
|
|
ccd469450a (date.c: switch to reentrant {gm,local}time_r, 2019-11-28)
removes the only gmtime() call we had and moves to gmtime_r() which
doesn't have the same portability problems.
Remove the compat gmtime code since it is no longer needed, and confirm
by successfull running t4212 in FreeBSD 9.3 amd64 (the oldest I could
get a hold off).
Further work might be needed to ensure 32bit time_t systems (like FreeBSD
i386) will handle correctly the overflows tested in t4212, but that is
orthogonal to this change, and it doesn't change the current behaviour
as neither gmtime() or gmtime_r() will ever return NULL on those systems
because time_t is unsigned.
Signed-off-by: Carlo Marcelo Arenas Belón <carenas@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
The "bugreport" tool.
* es/bugreport:
bugreport: drop extraneous includes
bugreport: add compiler info
bugreport: add uname info
bugreport: gather git version and build info
bugreport: add tool to generate debugging info
help: move list_config_help to builtin/help
|
|
In Linux with musl libc, we have this inclusion chain:
compat/regex/regex.c:69
`-> compat/regex/regex_internal.h
`-> /usr/include/stdlib.h
`-> /usr/include/features.h
`-> /usr/include/alloca.h
In that inclusion chain, `<features.h>` claims it's _BSD_SOURCE
compatible when it's NOT asked to be either
{_POSIX,_GNU,_XOPEN,_BSD}_SOURCE, or __STRICT_ANSI__.
And, `<stdlib.h>` will include `<alloca.h>` to be compatible with
software written for GNU and BSD. Thus, redefine `alloca` macro,
which was defined before at compat/regex/regex.c:66.
Considering this is only compat code, we've taken from other project,
it's not our business to decide which source should we adhere to.
Include `<stdlib.h>` early to prevent the redefinition of alloca.
This also remove a potential warning about alloca not defined on:
#undef alloca
Helped-by: Ramsay Jones <ramsay@ramsayjones.plus.com>
Signed-off-by: Đoàn Trần Công Danh <congdanhqx@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
* js/mingw-isilon-nfs:
mingw: cope with the Isilon network file system
|
|
Misc fixes for Windows.
* js/mingw-fixes:
mingw: help debugging by optionally executing bash with strace
mingw: do not treat `COM0` as a reserved file name
mingw: use modern strftime implementation if possible
|
|
To help pinpoint the source of a regression, it is useful to know some
info about the compiler which the user's Git client was built with. By
adding a generic get_compiler_info() in 'compat/' we can choose which
relevant information to share per compiler; to get started, let's
demonstrate the version of glibc if the user built with 'gcc'.
Signed-off-by: Emily Shaffer <emilyshaffer@google.com>
Helped-by: Đoàn Trần Công Danh <congdanhqx@gmail.com>
Helped-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Starting in 3ac68a93fd2, help.o began to depend on builtin/branch.o,
builtin/clean.o, and builtin/config.o. This meant that help.o was
unusable outside of the context of the main Git executable.
To make help.o usable by other commands again, move list_config_help()
into builtin/help.c (where it makes sense to assume other builtin libraries
are present).
When command-list.h is included but a member is not used, we start to
hear a compiler warning. Since the config list is generated in a fairly
different way than the command list, and since commands and config
options are semantically different, move the config list into its own
header and move the generator into its own script and build rule.
For reasons explained in 976aaedc (msvc: add a Makefile target to
pre-generate the Visual Studio solution, 2019-07-29), some build
artifacts we consider non-source files cannot be generated in the
Visual Studio environment, and we already have some Makefile tweaks
to help Visual Studio to use generated command-list.h header file.
Do the same to a new generated file, config-list.h, introduced by
this change.
Helped-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Emily Shaffer <emilyshaffer@google.com>
|
|
On certain network filesystems (currently encountered with Isilon, but
in theory more network storage solutions could be causing the same
issue), when the directory in question is missing,
`raceproof_create_file()` fails with an `ERROR_INVALID_PARAMETER`
instead of an `ERROR_PATH_NOT_FOUND`.
Since it is highly unlikely that we produce such an error by mistake
(the parameters we pass are fairly benign), we can be relatively certain
that the directory is missing in this instance. So let's just translate
that error automagically.
This fixes https://github.com/git-for-windows/git/issues/1345.
Signed-off-by: Nathan Sanders <spekbukkem@gmail.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
MSYS2's strace facility is very useful for debugging... With this patch,
the bash will be executed through strace if the environment variable
GIT_STRACE_COMMANDS is set, which comes in real handy when investigating
issues in the test suite.
Also support passing a path to a log file via GIT_STRACE_COMMANDS to
force Git to call strace.exe with the `-o <path>` argument, i.e. to log
into a file rather than print the log directly.
That comes in handy when the output would otherwise misinterpreted by a
calling process as part of Git's output.
Note: the values "1", "yes" or "true" are *not* specifying paths, but
tell Git to let strace.exe log directly to the console.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
In 4dc42c6c186 (mingw: refuse paths containing reserved names,
2019-12-21), we started disallowing file names that are reserved, e.g.
`NUL`, `CONOUT$`, etc.
This included `COM<n>` where `<n>` is a digit. Unfortunately, this
includes `COM0` but only `COM1`, ..., `COM9` are reserved, according to
the official documentation, `COM0` is mentioned in the "NT Namespaces"
section but it is explicitly _omitted_ from the list of reserved names:
https://docs.microsoft.com/en-us/windows/win32/fileio/naming-a-file#naming-conventions
Tests corroborate this: it is totally possible to write a file called
`com0.c` on Windows 10, but not `com1.c`.
So let's tighten the code to disallow only the reserved `COM<n>` file
names, but to allow `COM0` again.
This fixes https://github.com/git-for-windows/git/issues/2470.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Microsoft introduced a new "Universal C Runtime Library" (UCRT) with
Visual Studio 2015. The UCRT comes with a new strftime() implementation
that supports more date formats. We link git against the older
"Microsoft Visual C Runtime Library" (MSVCRT), so to use the UCRT
strftime() we need to load it from ucrtbase.dll using
DECLARE_PROC_ADDR()/INIT_PROC_ADDR().
Most supported Windows systems should have recieved the UCRT via Windows
update, but in some cases only MSVCRT might be available. In that case
we fall back to using that implementation.
With this change, it is possible to use e.g. the `%g` and `%V` date
format specifiers, e.g.
git show -s --format=%cd --date=format:‘%g.%V’ HEAD
Without this change, the user would see this error message on Windows:
fatal: invalid strftime format: '‘%g.%V’'
This fixes https://github.com/git-for-windows/git/issues/2495
Signed-off-by: Matthias Aßhauer <mha1993@live.de>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
On Cygwin, the codepath for POSIX-like systems is taken in
run-command.c::start_command(). The prepare_cmd() helper
function is called to decide if the command needs to be looked
up in the PATH. The logic there is to do the PATH-lookup if
and only if it does not have any slash '/' in it. If this test
passes we end up attempting to run the command by appending the
string after each colon-separated component of PATH.
The Cygwin environment supports both Windows and POSIX style
paths, so both forwardslahes '/' and back slashes '\' can be
used as directory separators for any external program the user
supplies.
Examples for path strings which are being incorrectly searched
for in the PATH instead of being executed as is:
- "C:\Program Files\some-program.exe"
- "a\b\c.exe"
To handle these, the PATH lookup detection logic in prepare_cmd()
is taught to know about this Cygwin quirk, by introducing
has_dir_sep(path) helper function to abstract away the difference
between true POSIX and Cygwin systems.
Signed-off-by: Andras Kucsma <r0maikx02b@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Dev support.
* js/mingw-open-in-gdb:
mingw: add a helper function to attach GDB to the current process
|
|
MinGW's poll() emulation has been improved.
* am/mingw-poll-fix:
mingw: workaround for hangs when sending STDIN
|
|
C pedantry ;-) fix.
* jk/clang-sanitizer-fixes:
obstack: avoid computing offsets from NULL pointer
xdiff: avoid computing non-zero offset from NULL pointer
avoid computing zero offsets from NULL pointer
merge-recursive: use subtraction to flip stage
merge-recursive: silence -Wxor-used-as-pow warning
|
|
MinGW's poll() emulation has been improved.
* am/mingw-poll-fix:
mingw: workaround for hangs when sending STDIN
|
|
Code cleanup.
* rs/micro-cleanups:
use strpbrk(3) to search for characters from a given set
quote: use isalnum() to check for alphanumeric characters
|
|
Explanation
-----------
The problem here is flawed `poll()` implementation. When it tries to
see if pipe can be written without blocking, it eventually calls
`NtQueryInformationFile()` and tests `WriteQuotaAvailable`. However,
the meaning of quota was misunderstood. The value of quota is reduced
when either some data was written to a pipe, *or* there is a pending
read on the pipe. Therefore, if there is a pending read of size >= than
the pipe's buffer size, poll() will think that pipe is not writable and
will hang forever, usually that means deadlocking both pipe users.
I have studied the problem and found that Windows pipes track two values:
`QuotaUsed` and `BytesInQueue`. The code in `poll()` apparently wants to
know `BytesInQueue` instead of quota. Unfortunately, `BytesInQueue` can
only be requested from read end of the pipe, while `poll()` receives
write end.
The git's implementation of `poll()` was copied from gnulib, which also
contains a flawed implementation up to today.
I also had a look at implementation in cygwin, which is also broken in a
subtle way. It uses this code in `pipe_data_available()`:
fpli.WriteQuotaAvailable = (fpli.OutboundQuota - fpli.ReadDataAvailable)
However, `ReadDataAvailable` always returns 0 for the write end of the pipe,
turning the code into an obfuscated version of returning pipe's total
buffer size, which I guess will in turn have `poll()` always say that pipe
is writable. The commit that introduced the code doesn't say anything about
this change, so it could be some debugging code that slipped in.
These are the typical sizes used in git:
0x2000 - default read size in `strbuf_read()`
0x1000 - default read size in CRT, used by `strbuf_getwholeline()`
0x2000 - pipe buffer size in compat\mingw.c
As a consequence, as soon as child process uses `strbuf_read()`,
`poll()` in parent process will hang forever, deadlocking both
processes.
This results in two observable behaviors:
1) If parent process begins sending STDIN quickly (and usually that's
the case), then first `poll()` will succeed and first block will go
through. MAX_IO_SIZE_DEFAULT is 8MB, so if STDIN exceeds 8MB, then
it will deadlock.
2) If parent process waits a little bit for any reason (including OS
scheduler) and child is first to issue `strbuf_read()`, then it will
deadlock immediately even on small STDINs.
The problem is illustrated by `git stash push`, which will currently
read the entire patch into memory and then send it to `git apply` via
STDIN. If patch exceeds 8MB, git hangs on Windows.
Possible solutions
------------------
1) Somehow obtain `BytesInQueue` instead of `QuotaUsed`
I did a pretty thorough search and didn't find any ways to obtain
the value from write end of the pipe.
2) Also give read end of the pipe to `poll()`
That can be done, but it will probably invite some dirty code,
because `poll()`
* can accept multiple pipes at once
* can accept things that are not pipes
* is expected to have a well known signature.
3) Make `poll()` always reply "writable" for write end of the pipe
Afterall it seems that cygwin (accidentally?) does that for years.
Also, it should be noted that `pump_io_round()` writes 8MB blocks,
completely ignoring the fact that pipe's buffer size is only 8KB,
which means that pipe gets clogged many times during that single
write. This may invite a deadlock, if child's STDERR/STDOUT gets
clogged while it's trying to deal with 8MB of STDIN. Such deadlocks
could be defeated with writing less than pipe's buffer size per
round, and always reading everything from STDOUT/STDERR before
starting next round. Therefore, making `poll()` always reply
"writable" shouldn't cause any new issues or block any future
solutions.
4) Increase the size of the pipe's buffer
The difference between `BytesInQueue` and `QuotaUsed` is the size
of pending reads. Therefore, if buffer is bigger than size of reads,
`poll()` won't hang so easily. However, I found that for example
`strbuf_read()` will get more and more hungry as it reads large inputs,
eventually surpassing any reasonable pipe buffer size.
Chosen solution
---------------
Make `poll()` always reply "writable" for write end of the pipe.
Hopefully one day someone will find a way to implement it properly.
Reproduction
------------
printf "%8388608s" X >large_file.txt
git stash push --include-untracked -- large_file.txt
I have decided not to include this as test to avoid slowing down the
test suite. I don't expect the specific problem to come back, and
chances are that `git stash push` will be reworked to avoid sending the
entire patch via STDIN.
Signed-off-by: Alexandr Miloslavskiy <alexandr.miloslavskiy@syntevo.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
We can check if certain characters are present in a string by calling
strchr(3) on each of them, or we can pass them all to a single
strpbrk(3) call. The latter is shorter, less repetitive and slightly
more efficient, so let's do that instead.
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
|
Dev support.
* js/mingw-open-in-gdb:
mingw: add a helper function to attach GDB to the current process
|
|
Work around test breakages caused by custom regex engine used in
libasan, when address sanitizer is used with more recent versions
of gcc and clang.
* jk/asan-build-fix:
Makefile: use compat regex with SANITIZE=address
|
