Skip to content

G
git

李少辉-开发者 / git
与 Fork 源项目一致

从无法访问的项目Fork

通知 2
Star 1
Fork 0
G
git

体验新版 GitCode，发现更多精彩内容 >>

切换分支/标签
  1. 27 3月, 2018 11 次提交
  3. 06 3月, 2018 3 次提交
    • Æ
      perl Git::LoadCPAN: emit better errors under NO_PERL_CPAN_FALLBACKS · 1aca69c0
      由 Ævar Arnfjörð Bjarmason 提交于 
      Before my 20d2a30f ("Makefile: replace perl/Makefile.PL with simple
make rules", 2017-12-10) on an OS package that removed the
private-Error.pm copy we carried around manually removing the OS's
Error.pm would yield:

    $ git add -p
    Can't locate Error.pm in @inc (you may need to install the Error module) [...]

Now, before this change we'll instead emit this more cryptic error:

    $ git add -p
    BUG: '/usr/share/perl5/Git/FromCPAN' should be a directory! at /usr/share/perl5/Git/Error.pm line 36.

This is a confusing error. Now if the new NO_PERL_CPAN_FALLBACKS
option is specified and we can't find the module we'll instead emit:

    $ /tmp/git/bin/git add -p
    BUG: The 'Error' module is not here, but NO_PERL_CPAN_FALLBACKS was set!

    [...]

Where [...] is the lengthy explanation seen in the change below, which
explains what the potential breakage is, and how to fix it.

The reason for checking @@NO_PERL_CPAN_FALLBACKS@@] against the empty
string in Perl is as opposed to checking for a boolean value is that
that's (as far as I can tell) make's idea of a string that's set, and
e.g. NO_PERL_CPAN_FALLBACKS=0 is enough to set NO_PERL_CPAN_FALLBACKS.
Signed-off-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      1aca69c0
    • T
      Makefile: add NO_PERL_CPAN_FALLBACKS knob · 075321c0
      由 Todd Zullinger 提交于 
      We include some perl modules which are not part of the core perl
install, as a convenience.  This allows us to rely on those modules in
our perl-based tools and scripts without requiring users to install the
modules from CPAN or their operating system packages.

Users whose operating system provides these modules and packagers of Git
often don't want to ship or use these bundled modules.  Allow these
users to set NO_PERL_CPAN_FALLBACKS to avoid installing the bundled
modules.
Signed-off-by: NTodd Zullinger <tmz@pobox.com>
Signed-off-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      075321c0
    • Æ
      perl: move the perl/Git/FromCPAN tree to perl/FromCPAN · 382029fc
      由 Ævar Arnfjörð Bjarmason 提交于 
      Move the CPAN modules that have lived under perl/Git/FromCPAN since my
20d2a30f ("Makefile: replace perl/Makefile.PL with simple make
rules", 2017-12-10) to perl/FromCPAN.

A subsequent change will teach the Makefile to only install these
copies of CPAN modules if a flag that distro packagers would like to
set isn't set. Due to how the wildcard globbing is being done it's
much easier to accomplish that if they're moved to their own
directory.
Signed-off-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      382029fc
  5. 18 2月, 2018 1 次提交
  7. 16 2月, 2018 1 次提交
    • S
      Makefile: generate Git(3pm) as dependency of the 'doc' and 'man' targets · 2530afd3
      由 SZEDER Gábor 提交于 
      Since commit 20d2a30f (Makefile: replace perl/Makefile.PL with
simple make rules, 2017-12-10), the Git(3pm) man page is only
generated as an indirect dependency of the 'install-doc' and
'install-man' Makefile targets.  Consequently, if someone runs 'make
man && sudo make install-man' (or their 'doc' counterparts), then
Git(3pm) will be generated as root, and the resulting root-owned files
and directories will in turn cause the next user-run 'make clean' to
fail.  This was not an issue in the past, because Git(3pm) was
generated when 'make all' descended into 'perl/', which is usually not
run as root.

List Git(3pm) as a dependency of the 'doc' and 'man' Makefile targets,
too, so it gets generated by targets that are usually built as
ordinary users.

While at it, add 'install-man-perl' to the list of .PHONY targets.
Signed-off-by: NSZEDER Gábor <szeder.dev@gmail.com>
Acked-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      2530afd3
  9. 13 2月, 2018 1 次提交
    • R
      Makefile: suppress a sparse warning for pack-revindex.c · 54360a19
      由 Ramsay Jones 提交于 
      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: NRamsay Jones <ramsay@ramsayjones.plus.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      54360a19
  11. 25 1月, 2018 1 次提交
  13. 04 1月, 2018 1 次提交
    • J
      perl: treat PERLLIB_EXTRA as an extra path again · 7a7bfc7a
      由 Jonathan Nieder 提交于 
      PERLLIB_EXTRA was introduced in v1.9-rc0~88^2 (2013-11-15) as a way
for packagers to add additional directories such as the location of
Subversion's perl bindings to Git's perl path.  Since 20d2a30f
(Makefile: replace perl/Makefile.PL with simple make rules,
2012-12-10) setting that variable breaks perl-based commands instead:

 $ PATH=$HOME/opt/git/bin:$PATH
 $ make install prefix=$HOME/opt/git PERLLIB_EXTRA=anextralibdir
[...]
 $ head -2 $HOME/opt/git/libexec/git-core/git-add--interactive
 #!/usr/bin/perl
 use lib (split(/:/, $ENV{GITPERLLIB} || ":helloiamanextrainstlibdir" || "/usr/local/google/home/jrn/opt/git/share/perl5"));
 $ git add -p
 Empty compile time value given to use lib at /home/jrn/opt/git/libexec/git-core/git-add--interactive line 2.

Removing the spurious ":" at the beginning of ":$PERLLIB_EXTRA" avoids
the "Empty compile time value" error but with that tweak the problem
still remains: PERLLIB_EXTRA ends up replacing instead of
supplementing the perllibdir that would be passed to 'use lib' if
PERLLIB_EXTRA were not set.

The intent was to simplify, as the commit message to 20d2a30f
explains:

| The scripts themselves will 'use lib' the target directory, but if
| INSTLIBDIR is set it overrides it. It doesn't have to be this way,
| it could be set in addition to INSTLIBDIR, but my reading of
| [v1.9-rc0~88^2] is that this is the desired behavior.

Restore the previous code structure to make PERLLIB_EXTRA work again.

Reproducing this problem requires an invocation of "make install"
instead of running bin-wrappers/git in place, since the latter sets
the GITPERLLIB environment variable, avoiding trouble.
Reported-by: NJonathan Koren <jdkoren@google.com>
Signed-off-by: NJonathan Nieder <jrnieder@gmail.com>
Acked-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      7a7bfc7a
  15. 29 12月, 2017 2 次提交
  17. 15 12月, 2017 2 次提交
  19. 12 12月, 2017 1 次提交
    • Æ
      Makefile: replace perl/Makefile.PL with simple make rules · 20d2a30f
      由 Ævar Arnfjörð Bjarmason 提交于 
      Replace the perl/Makefile.PL and the fallback perl/Makefile used under
NO_PERL_MAKEMAKER=NoThanks with a much simpler implementation heavily
inspired by how the i18n infrastructure's build process works[1].

The reason for having the Makefile.PL in the first place is that it
was initially[2] building a perl C binding to interface with libgit,
this functionality, that was removed[3] before Git.pm ever made it to
the master branch.

We've since since started maintaining a fallback perl/Makefile, as
MakeMaker wouldn't work on some platforms[4]. That's just the tip of
the iceberg. We have the PM.stamp hack in the top-level Makefile[5] to
detect whether we need to regenerate the perl/perl.mak, which I fixed
just recently to deal with issues like the perl version changing from
under us[6].

There is absolutely no reason for why this needs to be so complex
anymore. All we're getting out of this elaborate Rube Goldberg machine
was copying perl/* to perl/blib/* as we do a string-replacement on
the *.pm files to hardcode @@LOCALEDIR@@ in the source, as well as
pod2man-ing Git.pm & friends.

So replace the whole thing with something that's pretty much a copy of
how we generate po/build/**.mo from po/*.po, just with a small sed(1)
command instead of msgfmt. As that's being done rename the files
from *.pm to *.pmc just to indicate that they're generated (see
"perldoc -f require").

While I'm at it, change the fallback for Error.pm from being something
where we'll ship our own Error.pm if one doesn't exist at build time
to one where we just use a Git::Error wrapper that'll always prefer
the system-wide Error.pm, only falling back to our own copy if it
really doesn't exist at runtime. It's now shipped as
Git::FromCPAN::Error, making it easy to add other modules to
Git::FromCPAN::* in the future if that's needed.

Functional changes:

 * This will not always install into perl's idea of its global
   "installsitelib". This only potentially matters for packagers that
   need to expose Git.pm for non-git use, and as explained in the
   INSTALL file there's a trivial workaround.

 * The scripts themselves will 'use lib' the target directory, but if
   INSTLIBDIR is set it overrides it. It doesn't have to be this way,
   it could be set in addition to INSTLIBDIR, but my reading of [7] is
   that this is the desired behavior.

 * We don't build man pages for all of the perl modules as we used to,
   only Git(3pm). As discussed on-list[8] that we were building
   installed manpages for purely internal APIs like Git::I18N or
   private-Error.pm was always a bug anyway, and all the Git::SVN::*
   ones say they're internal APIs.

   There are apparently external users of Git.pm, but I don't expect
   there to be any of the others.

   As a side-effect of these general changes the perl documentation
   now only installed by install-{doc,man}, not a mere "install" as
   before.

1. 5e9637c6 ("i18n: add infrastructure for translating Git with
   gettext", 2011-11-18)

2. b1edc53d ("Introduce Git.pm (v4)", 2006-06-24)

3. 18b0fc1c ("Git.pm: Kill Git.xs for now", 2006-09-23)

4. f848718a ("Make perl/ build procedure ActiveState friendly.",
   2006-12-04)

5. ee9be067 ("perl: detect new files in MakeMaker builds",
   2012-07-27)

6. c59c4939 ("perl: regenerate perl.mak if perl -V changes",
   2017-03-29)

7. 0386dd37 ("Makefile: add PERLLIB_EXTRA variable that adds to
   default perl path", 2013-11-15)

8. 87bmjjv1pu.fsf@evledraar.booking.com ("Re: [PATCH] Makefile:
   replace perl/Makefile.PL with simple make rules"
Signed-off-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      20d2a30f
  21. 09 12月, 2017 4 次提交
    • Æ
      Makefile: under "make dist", include the sha1collisiondetection submodule · bc2ed316
      由 Ævar Arnfjörð Bjarmason 提交于 
      Include the sha1collisiondetection submodule when running "make
dist". Even though we've been shipping the sha1collisiondetection
submodule[1] and using it by default if it's checked out[2] anyone
downloading git as a tarball would just get an empty
sha1collisiondetection/ directory.

Doing this automatically is a feature that's missing from git-archive,
but in the meantime let's bundle this up into the tarball we
ship. This ensures that the DC_SHA1_SUBMODULE flag does what's
intended even in an unpacked tarball, and more importantly means we're
building the exact same code from the same paths from git.git and from
the tarball.

I am not including all the files in the submodule, only the ones git
actually needs (and the licenses), only including some files like this
would be a useful feature if git-archive ever adds the ability to
bundle up submodules.

1. commit 86cfd61e ("sha1dc: optionally use sha1collisiondetection
   as a submodule", 2017-07-01)
2. cac87dc0 ("sha1collisiondetection: automatically enable when
   submodule is populated", 2017-07-01)
Signed-off-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      bc2ed316
    • Æ
      Makefile: don't error out under DC_SHA1_EXTERNAL if DC_SHA1_SUBMODULE=auto · f39e05f2
      由 Ævar Arnfjörð Bjarmason 提交于 
      Fix a logic error in the initial introduction of DC_SHA1_EXTERNAL. If
git.git has a sha1collisiondetection submodule checked out the logic
to set DC_SHA1_SUBMODULE=auto would interact badly with the check for
whether DC_SHA1_SUBMODULE was set.

It would error out, meaning that there's no way to build git with
DC_SHA1_EXTERNAL=YesPlease without deinit-ing the submodule.

Instead, adjust the logic to only fire if the variable is to something
else than "auto" which would mean it's a mistake on the part of
whoever's building git, not just the Makefile tripping over its own
logic.

1. 3964cbbb ("sha1dc: allow building with the external sha1dc
   library", 2017-08-15)
2. cac87dc0 ("sha1collisiondetection: automatically enable when
   submodule is populated", 2017-07-01)
Signed-off-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      f39e05f2
    • J
      decorate: clean up and document API · ddd3e312
      由 Jonathan Tan 提交于 
      Improve the names of the identifiers in decorate.h, document them, and
add an example of how to use these functions.

The example is compiled and run as part of the test suite.
Signed-off-by: NJonathan Tan <jonathantanmy@google.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      ddd3e312
    • J
      t/Makefile: introduce TEST_SHELL_PATH · 3f824e91
      由 Jeff King 提交于 
      You may want to run the test suite with a different shell
than you use to build Git. For instance, you may build with
SHELL_PATH=/bin/sh (because it's faster, or it's what you
expect to exist on systems where the build will be used) but
want to run the test suite with bash (e.g., since that
allows using "-x" reliably across the whole test suite).
There's currently no good way to do this.

You might think that doing two separate make invocations,
like:

  make &&
  make -C t SHELL_PATH=/bin/bash

would work. And it _almost_ does. The second make will see
our bash SHELL_PATH, and we'll use that to run the
individual test scripts (or tell prove to use it to do so).
So far so good.

But this breaks down when "--tee" or "--verbose-log" is
used. Those options cause the test script to actually
re-exec itself using $SHELL_PATH. But wait, wouldn't our
second make invocation have set SHELL_PATH correctly in the
environment?

Yes, but test-lib.sh sources GIT-BUILD-OPTIONS, which we
built during the first "make". And that overrides the
environment, giving us the original SHELL_PATH again.

Let's introduce a new variable that lets you specify a
specific shell to be run for the test scripts. Note that we
have to touch both the main and t/ Makefiles, since we have
to record it in GIT-BUILD-OPTIONS in one, and use it in the
latter.
Signed-off-by: NJeff King <peff@peff.net>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      3f824e91
  23. 06 12月, 2017 1 次提交
    • J
      introduce fetch-object: fetch one promisor object · 88e2f9ed
      由 Jonathan Tan 提交于 
      Introduce fetch-object, providing the ability to fetch one object from a
promisor remote.

This uses fetch-pack. To do this, the transport mechanism has been
updated with 2 flags, "from-promisor" to indicate that the resulting
pack comes from a promisor remote (and thus should be annotated as such
by index-pack), and "no-dependents" to indicate that only the objects
themselves need to be fetched (but fetching additional objects is
nevertheless safe).

Whenever "no-dependents" is used, fetch-pack will refrain from using any
object flags, because it is most likely invoked as part of a dynamic
object fetch by another Git command (which may itself use object flags).
An alternative to this is to leave fetch-pack alone, and instead update
the allocation of flags so that fetch-pack's flags never overlap with
any others, but this will end up shrinking the number of flags available
to nearly every other Git command (that is, every Git command that
accesses objects), so the approach in this commit was used instead.

This will be tested in a subsequent commit.
Signed-off-by: NJonathan Tan <jonathantanmy@google.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      88e2f9ed
  25. 27 11月, 2017 1 次提交
  27. 22 11月, 2017 2 次提交
  29. 01 11月, 2017 1 次提交
  31. 17 10月, 2017 1 次提交
    • B
      protocol: introduce protocol extension mechanisms · 373d70ef
      由 Brandon Williams 提交于 
      Create protocol.{c,h} and provide functions which future servers and
clients can use to determine which protocol to use or is being used.

Also introduce the 'GIT_PROTOCOL' environment variable which will be
used to communicate a colon separated list of keys with optional values
to a server.  Unknown keys and values must be tolerated.  This mechanism
is used to communicate which version of the wire protocol a client would
like to use with a server.
Signed-off-by: NBrandon Williams <bmwill@google.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      373d70ef
  33. 01 10月, 2017 4 次提交
    • B
      fsmonitor: add a performance test · 14527b30
      由 Ben Peart 提交于 
      Add a test utility (test-drop-caches) that flushes all changes to disk
then drops file system cache on Windows, Linux, and OSX.

Add a perf test (p7519-fsmonitor.sh) for fsmonitor.

By default, the performance test will utilize the Watchman file system
monitor if it is installed.  If Watchman is not installed, it will use a
dummy integration script that does not report any new or modified files.
The dummy script has very little overhead which provides optimistic results.

The performance test will also use the untracked cache feature if it is
available as fsmonitor uses it to speed up scanning for untracked files.

There are 4 environment variables that can be used to alter the default
behavior of the performance test:

GIT_PERF_7519_UNTRACKED_CACHE: used to configure core.untrackedCache
GIT_PERF_7519_SPLIT_INDEX: used to configure core.splitIndex
GIT_PERF_7519_FSMONITOR: used to configure core.fsmonitor
GIT_PERF_7519_DROP_CACHE: if set, the OS caches are dropped between tests

The big win for using fsmonitor is the elimination of the need to scan the
working directory looking for changed and untracked files. If the file
information is all cached in RAM, the benefits are reduced.
Signed-off-by: NBen Peart <benpeart@microsoft.com>
Signed-off-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      14527b30
    • B
      fsmonitor: add a test tool to dump the index extension · dd3551f4
      由 Ben Peart 提交于 
      Add a test utility (test-dump-fsmonitor) that will dump the fsmonitor
index extension.
Signed-off-by: NBen Peart <benpeart@microsoft.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      dd3551f4
    • B
      fsmonitor: teach git to optionally utilize a file system monitor to speed up... · 883e248b
      由 Ben Peart 提交于 
      fsmonitor: teach git to optionally utilize a file system monitor to speed up detecting new or changed files.

When the index is read from disk, the fsmonitor index extension is used
to flag the last known potentially dirty index entries. The registered
core.fsmonitor command is called with the time the index was last
updated and returns the list of files changed since that time. This list
is used to flag any additional dirty cache entries and untracked cache
directories.

We can then use this valid state to speed up preload_index(),
ie_match_stat(), and refresh_cache_ent() as they do not need to lstat()
files to detect potential changes for those entries marked
CE_FSMONITOR_VALID.

In addition, if the untracked cache is turned on valid_cached_dir() can
skip checking directories for new or changed files as fsmonitor will
invalidate the cache only for those directories that have been
identified as having potential changes.

To keep the CE_FSMONITOR_VALID state accurate during git operations;
when git updates a cache entry to match the current state on disk,
it will now set the CE_FSMONITOR_VALID bit.

Inversely, anytime git changes a cache entry, the CE_FSMONITOR_VALID bit
is cleared and the corresponding untracked cache directory is marked
invalid.
Signed-off-by: NBen Peart <benpeart@microsoft.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      883e248b
    • J
      oidmap: map with OID as key · 9e6fabde
      由 Jonathan Tan 提交于 
      This is similar to using the hashmap in hashmap.c, but with an
easier-to-use API. In particular, custom entry comparisons no longer
need to be written, and lookups can be done without constructing a
temporary entry structure.

This is implemented as a thin wrapper over the hashmap API. In
particular, this means that there is an additional 4-byte overhead due
to the fact that the first 4 bytes of the hash is redundantly stored.
For now, I'm taking the simpler approach, but if need be, we can
reimplement oidmap without affecting the callers significantly.

oidset has been updated to use oidmap.
Signed-off-by: NJonathan Tan <jonathantanmy@google.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      9e6fabde
  35. 25 9月, 2017 1 次提交
    • M
      packed_ref_cache: keep the `packed-refs` file mmapped if possible · 5b633610
      由 Michael Haggerty 提交于 
      Keep a copy of the `packed-refs` file contents in memory for as long
as a `packed_ref_cache` object is in use:

* If the system allows it, keep the `packed-refs` file mmapped.

* If not (either because the system doesn't support `mmap()` at all,
  or because a file that is currently mmapped cannot be replaced via
  `rename()`), then make a copy of the file's contents in
  heap-allocated space, and keep that around instead.

We base the choice of behavior on a new build-time switch,
`MMAP_PREVENTS_DELETE`. By default, this switch is set for Windows
variants.

After this commit, `MMAP_NONE` and `MMAP_TEMPORARY` are still handled
identically. But the next commit will introduce a difference.

This whole change is still pointless, because we only read the
`packed-refs` file contents immediately after instantiating the
`packed_ref_cache`. But that will soon change.
Signed-off-by: NMichael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      5b633610
  37. 08 9月, 2017 1 次提交
    • J
      add UNLEAK annotation for reducing leak false positives · 0e5bba53
      由 Jeff King 提交于 
      It's a common pattern in git commands to allocate some
memory that should last for the lifetime of the program and
then not bother to free it, relying on the OS to throw it
away.

This keeps the code simple, and it's fast (we don't waste
time traversing structures or calling free at the end of the
program). But it also triggers warnings from memory-leak
checkers like valgrind or LSAN. They know that the memory
was still allocated at program exit, but they don't know
_when_ the leaked memory stopped being useful. If it was
early in the program, then it's probably a real and
important leak. But if it was used right up until program
exit, it's not an interesting leak and we'd like to suppress
it so that we can see the real leaks.

This patch introduces an UNLEAK() macro that lets us do so.
To understand its design, let's first look at some of the
alternatives.

Unfortunately the suppression systems offered by
leak-checking tools don't quite do what we want. A
leak-checker basically knows two things:

  1. Which blocks were allocated via malloc, and the
     callstack during the allocation.

  2. Which blocks were left un-freed at the end of the
     program (and which are unreachable, but more on that
     later).

Their suppressions work by mentioning the function or
callstack of a particular allocation, and marking it as OK
to leak.  So imagine you have code like this:

  int cmd_foo(...)
  {
	/* this allocates some memory */
	char *p = some_function();
	printf("%s", p);
	return 0;
  }

You can say "ignore allocations from some_function(),
they're not leaks". But that's not right. That function may
be called elsewhere, too, and we would potentially want to
know about those leaks.

So you can say "ignore the callstack when main calls
some_function".  That works, but your annotations are
brittle. In this case it's only two functions, but you can
imagine that the actual allocation is much deeper. If any of
the intermediate code changes, you have to update the
suppression.

What we _really_ want to say is that "the value assigned to
p at the end of the function is not a real leak". But
leak-checkers can't understand that; they don't know about
"p" in the first place.

However, we can do something a little bit tricky if we make
some assumptions about how leak-checkers work. They
generally don't just report all un-freed blocks. That would
report even globals which are still accessible when the
leak-check is run.  Instead they take some set of memory
(like BSS) as a root and mark it as "reachable". Then they
scan the reachable blocks for anything that looks like a
pointer to a malloc'd block, and consider that block
reachable. And then they scan those blocks, and so on,
transitively marking anything reachable from a global as
"not leaked" (or at least leaked in a different category).

So we can mark the value of "p" as reachable by putting it
into a variable with program lifetime. One way to do that is
to just mark "p" as static. But that actually affects the
run-time behavior if the function is called twice (you
aren't likely to call main() twice, but some of our cmd_*()
functions are called from other commands).

Instead, we can trick the leak-checker by putting the value
into _any_ reachable bytes. This patch keeps a global
linked-list of bytes copied from "unleaked" variables. That
list is reachable even at program exit, which confers
recursive reachability on whatever values we unleak.

In other words, you can do:

  int cmd_foo(...)
  {
	char *p = some_function();
	printf("%s", p);
	UNLEAK(p);
	return 0;
  }

to annotate "p" and suppress the leak report.

But wait, couldn't we just say "free(p)"? In this toy
example, yes. But UNLEAK()'s byte-copying strategy has
several advantages over actually freeing the memory:

  1. It's recursive across structures. In many cases our "p"
     is not just a pointer, but a complex struct whose
     fields may have been allocated by a sub-function. And
     in some cases (e.g., dir_struct) we don't even have a
     function which knows how to free all of the struct
     members.

     By marking the struct itself as reachable, that confers
     reachability on any pointers it contains (including those
     found in embedded structs, or reachable by walking
     heap blocks recursively.

  2. It works on cases where we're not sure if the value is
     allocated or not. For example:

       char *p = argc > 1 ? argv[1] : some_function();

     It's safe to use UNLEAK(p) here, because it's not
     freeing any memory. In the case that we're pointing to
     argv here, the reachability checker will just ignore
     our bytes.

  3. Likewise, it works even if the variable has _already_
     been freed. We're just copying the pointer bytes. If
     the block has been freed, the leak-checker will skip
     over those bytes as uninteresting.

  4. Because it's not actually freeing memory, you can
     UNLEAK() before we are finished accessing the variable.
     This is helpful in cases like this:

       char *p = some_function();
       return another_function(p);

     Writing this with free() requires:

       int ret;
       char *p = some_function();
       ret = another_function(p);
       free(p);
       return ret;

     But with unleak we can just write:

       char *p = some_function();
       UNLEAK(p);
       return another_function(p);

This patch adds the UNLEAK() macro and enables it
automatically when Git is compiled with SANITIZE=leak.  In
normal builds it's a noop, so we pay no runtime cost.

It also adds some UNLEAK() annotations to show off how the
feature works. On top of other recent leak fixes, these are
enough to get t0000 and t0001 to pass when compiled with
LSAN.

Note the case in commit.c which actually converts a
strbuf_release() into an UNLEAK. This code was already
non-leaky, but the free didn't do anything useful, since
we're exiting. Converting it to an annotation means that
non-leak-checking builds pay no runtime cost. The cost is
minimal enough that it's probably not worth going on a
crusade to convert these kinds of frees to UNLEAKS. I did it
here for consistency with the "sb" leak (though it would
have been equally correct to go the other way, and turn them
both into strbuf_release() calls).
Signed-off-by: NJeff King <peff@peff.net>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>
      0e5bba53