A broken or badly formatted commit might not record author or
committer lines or we may not find a valid name on them. The
function record_person() returned after calling get_commit_buffer()
without calling unuse_commit_buffer() on the memory it obtained in
such cases, potentially leaking it.
Helped-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Each of these sites assumes that commit->buffer is valid.
Since they would segfault if this was not the case, they are
likely to be correct in practice. However, we can
future-proof them by using get_commit_buffer.
And as a side effect, we abstract away the final bare uses
of commit->buffer.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Like the callsites in the previous commit, logmsg_reencode
already falls back to read_sha1_file when necessary.
However, I split its conversion out into its own commit
because it's a bit more complex.
We return either:
1. The original commit->buffer
2. A newly allocated buffer from read_sha1_file
3. A reencoded buffer (based on either 1 or 2 above).
while trying to do as few extra reads/allocations as
possible. Callers currently free the result with
logmsg_free, but we can simplify this by pointing them
straight to unuse_commit_buffer. This is a slight layering
violation, in that we may be passing a buffer from (3).
However, since the end result is to free() anything except
(1), which is unlikely to change, and because this makes the
interface much simpler, it's a reasonable bending of the
rules.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
For both of these sites, we already do the "fallback to
read_sha1_file" trick. But we can shorten the code by just
using get_commit_buffer.
Note that the error cases are slightly different when
read_sha1_file fails. get_commit_buffer will die() if the
object cannot be loaded, or is a non-commit.
For get_sha1_oneline, this will almost certainly never
happen, as we will have just called parse_object (and if it
does, it's probably worth complaining about).
For record_author_date, the new behavior is probably better;
we notify the user of the error instead of silently ignoring
it. And because it's used only for sorting by author-date,
somebody examining a corrupt repo can fallback to the
regular traversal order.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some call sites check commit->buffer to see whether we have
a cached buffer, and if so, do some work with it. In the
long run we may want to switch these code paths to make
their decision on a different boolean flag (because checking
the cache may get a little more expensive in the future).
But for now, we can easily support them by converting the
calls to use get_cached_commit_buffer.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Many sites look at commit->buffer to get more detailed
information than what is in the parsed commit struct.
However, we sometimes drop commit->buffer to save memory,
in which case the caller would need to read the object
afresh. Some callers do this (leading to duplicated code),
and others do not (which opens the possibility of a segfault
if somebody else frees the buffer).
Let's provide a pair of helpers, "get" and "unuse", that let
callers easily get the buffer. They will use the cached
buffer when possible, and otherwise load from disk using
read_sha1_file.
Note that we also need to add a "get_cached" variant which
returns NULL when we do not have a cached buffer. At first
glance this seems to defeat the purpose of "get", which is
to always provide a return value. However, some log code
paths actually use the NULL-ness of commit->buffer as a
boolean flag to decide whether to try printing the
commit. At least for now, we want to continue supporting
that use.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This converts two lines into one at each caller. But more
importantly, it abstracts the concept of freeing the buffer,
which will make it easier to change later.
Note that we also need to provide a "detach" mechanism for a
tricky case in index-pack. We are passed a buffer for the
object generated by processing the incoming pack. If we are
not using --strict, we just calculate the sha1 on that
buffer and return, leaving the caller to free it. But if we
are using --strict, we actually attach that buffer to an
object, pass the object to the fsck functions, and then
detach the buffer from the object again (so that the caller
can free it as usual). In this case, we don't want to free
the buffer ourselves, but just make sure it is no longer
associated with the commit.
Note that we are making the assumption here that the
attach/detach process does not impact the buffer at all
(e.g., it is never reallocated or modified). That holds true
now, and we have no plans to change that. However, as we
abstract the commit_buffer code, this dependency becomes
less obvious. So when we detach, let's also make sure that
we get back the same buffer that we gave to the
commit_buffer code.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This simplifies the code, as logmsg_reencode handles the
reencoding for us in a single call. It also means we learn
logmsg_reencode's trick of pulling the buffer from disk when
commit->buffer is NULL (we currently just silently return!).
It is doubtful this matters in practice, though, as
sequencer operations would not generally turn off
save_commit_buffer.
Note that we may be fixing a bug here. The existing code
does:
if (same_encoding(to, from))
reencode_string(buf, to, from);
That probably should have been "!same_encoding".
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The return value from logmsg_reencode may be either a newly
allocated buffer or a pointer to the existing commit->buffer.
We would not want the caller to accidentally free() or
modify the latter, so let's mark it as const. We can cast
away the constness in logmsg_free, but only once we have
determined that it is a free-able buffer.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In both blame and merge-recursive, we sometimes create a
"fake" commit struct for convenience (e.g., to represent the
HEAD state as if we would commit it). By allocating
ourselves rather than using alloc_commit_node, we do not
properly set the "index" field of the commit. This can
produce subtle bugs if we then use commit-slab on the
resulting commit, as we will share the "0" index with
another commit.
We can fix this by using alloc_commit_node() to allocate.
Note that we cannot free the result, as it is part of our
commit allocator. However, both cases were already leaking
the allocated commit anyway, so there's nothing to fix up.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
commit: push commit_index update into alloc_commit_node
Whenever we create a commit object via lookup_commit, we
give it a unique index to be used with the commit-slab API.
The theory is that any "struct commit" we create would
follow this code path, so any such struct would get an
index. However, callers could use alloc_commit_node()
directly (and get multiple commits with index 0).
Let's push the indexing into alloc_commit_node so that it's
hard for callers to get it wrong.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
alloc: include any-object allocations in alloc_report
When 2c1cbec (Use proper object allocators for unknown
object nodes too, 2007-04-16), added a special "any_object"
allocator, it never taught alloc_report to report on it. To
do so we need to add an extra type argument to the REPORT
macro, as that commit did for DEFINE_ALLOCATOR.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
It is not a good idea to strbuf_attach an arbitrary pointer
just because a function you are calling wants a strbuf.
Attaching implies a transfer of memory ownership; if anyone
were to modify or release the resulting strbuf, we would
free() the pointer, leading to possible problems:
1. Other users of the original pointer might access freed
memory.
2. The pointer might not be the start of a malloc'd
area, so calling free() on it in the first place would
be wrong.
In the two cases modified here, we are fortunate that nobody
touches the strbuf once it is attached, but it is an
accident waiting to happen. Since the previous commit,
commit_tree and friends take a pointer/buf pair, so we can
just do away with the strbufs entirely.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
commit_tree: take a pointer/len pair rather than a const strbuf
While strbufs are pretty common throughout our code, it is
more flexible for functions to take a pointer/len pair than
a strbuf. It's easy to turn a strbuf into such a pair (by
dereferencing its members), but less easy to go the other
way (you can strbuf_attach, but that has implications about
memory ownership).
This patch teaches commit_tree (and its associated callers
and sub-functions) to take such a pair for the commit
message rather than a strbuf. This makes passing the buffer
around slightly more verbose, but means we can get rid of
some dangerous strbuf_attach calls in the next patch.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
These were originally removed by 0232852 (t5537: move
http tests out to t5539, 2014-02-13). However, they were
accidentally re-added in 1ddb4d7 (Merge branch
'nd/upload-pack-shallow', 2014-03-21).
This looks like an error in manual conflict resolution.
Here's what happened:
1. v1.9.0 shipped with the http tests in t5537.
2. We realized that this caused problems, and built 0232852 on top to move the tests to their own file.
This fix made it into v1.9.1.
3. We later had another fix in nd/upload-pack-shallow that
also touched t5537. It was built directly on v1.9.0.
When we merged nd/upload-pack-shallow to master, we got a
conflict; it was built on a version with the http tests, but
we had since removed them. The correct resolution was to
drop the http tests and keep the new ones, but instead we
kept everything.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* mw/symlinks:
setup: fix windows path buffer over-stepping
setup: don't dereference in-tree symlinks for absolute paths
setup: add abspath_part_inside_repo() function
t0060: add tests for prefix_path when path begins with work tree
t0060: add test for prefix_path when path == work tree
t0060: add test for prefix_path on symlinks via absolute paths
t3004: add test for ls-files on symlinks via absolute paths
Instead of showing a warning and working as before, fail and show
the message and force immediate upgrade from their upstream
repositories when these tools are run, per request from their
primary author.
The author of the original topic says he broke the upcoming 2.0
release with something that relates to "synchronization crash
regression" while refusing to give further specifics, so this would
unfortunately be the safest option for the upcoming release.
git-prompt.sh: don't assume the shell expands the value of PS1
Not all shells subject the prompt string to parameter expansion. Test
whether the shell will expand the value of PS1, and use the result to
control whether raw ref names are included directly in PS1.
This fixes a regression introduced in commit 8976500 ("git-prompt.sh:
don't put unsanitized branch names in $PS1"): zsh does not expand PS1
by default, but that commit assumed it did. The bug resulted in
prompts containing the literal string '${__git_ps1_branch_name}'
instead of the actual branch name.
Reported-by: Caleb Thompson <caleb@calebthompson.io> Signed-off-by: Richard Hansen <rhansen@bbn.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-gui: tolerate major version changes when comparing the git version
Since git 2.0.0 starting git gui in a submodule using a gitfile fails with
the following error:
No working directory ../../../<path>
couldn't change working directory
to "../../../<path>": no such file or
directory
This is because "git rev-parse --show-toplevel" is only run when git gui
sees a git version of at least 1.7.0 (which is the version in which the
--show-toplevel option was introduced). But "package vsatisfies" returns
false when the major version changes, which is not what we want here.
Fix that for both places where the git version is checked using vsatisfies
by appending a '-' to the version number. This tells vsatisfies that a
change of the major version is not considered to be a problem, as long as
the new major version is larger. This is done for both the place that
caused the reported bug and another spot where the git version is tested
for another feature.
Reported-by: Chris Packham <judge.packham@gmail.com> Reported-by: Yann Dirson <ydirson@free.fr> Helped-by: Pat Thoyts <patthoyts@users.sourceforge.net> Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de> Tested-by: Chris Packham <judge.packham@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Otherwise it might collide with a function of the same name in the
user's environment.
Suggested-by: SZEDER Gábor <szeder@ira.uka.de> Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The stray "+" has been there ever since the example was added in
v1.8.3-rc0~210^2 (shell: new no-interactive-login command to print a
custom message, 2013-03-09). The "+" sign between paragraphs is
needed in asciidoc to attach extra paragraphs to a list item but here
it is not needed and ends up rendered as a literal "+". Remove it.
A quick search with "grep -e '<p>+' /usr/share/doc/git/html/*.html"
doesn't find any other instances of this problem.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Merge branch 'km/avoid-non-function-return-in-rebase' into maint
"git rebase" used a POSIX shell construct FreeBSD /bin/sh does not
work well with.
* km/avoid-non-function-return-in-rebase:
Revert "rebase: fix run_specific_rebase's use of "return" on FreeBSD"
rebase: avoid non-function use of "return" on FreeBSD
Describe one last minute one-liner fix for regression introduced in
1.9, and fix a grave mischaracterization on a recent remote-hg/bzr
change, pointed out by Felipe.
git-p4: format-patch to diff-tree change breaks binary patches
When applying binary patches a full index is required. format-patch
already handles this, but diff-tree needs '--full-index' argument
to always output full index. When git-p4 runs git-apply to test
the patch, git-apply rejects the patch due to abbreviated blob
object names. This is the error message git-apply emits in this
case:
error: cannot apply binary patch to '<filename>' without full index line
error: <filename>: patch does not apply
Signed-off-by: Tolga Ceylan <tolga.ceylan@gmail.com> Acked-by: Pete Wyckoff <pw@padd.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Makefile: default to -lcurl when no CURL_CONFIG or CURLDIR
The original implementation of CURL_CONFIG support did not match the
original behavior of using -lcurl when CURLDIR was not set. This broke
implementations that were lacking curl-config but did have libcurl
installed along system libraries, such as MSysGit. In other words, the
assumption that curl-config is always installed was incorrect.
Instead, if CURL_CONFIG is empty or returns an empty result (e.g. due
to curl-config being missing), use the old behavior of falling back to
-lcurl.
Signed-off-by: Dave Borowitz <dborowitz@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix a buffer over-stepping issue triggered by providing an absolute path
that is similar to the work tree path.
abspath_part_inside_repo() may currently increment the path pointer by
offset_1st_component() + wtlen, which is too much, since
offset_1st_component() is a subset of wtlen.
For the *nix-style prefix '/', this does (by luck) not cause any issues,
since offset_1st_component() is 1 and there will always be a '/' or '\0'
that can "absorb" this.
In the case of DOS-style prefixes though, the offset_1st_component() is
3 and this can potentially over-step the string buffer. For example if
work_tree = "c:/r"
path = "c:/rl"
Then wtlen is 4, and incrementing the path pointer by (3 + 4) would
end up 2 bytes outside a string buffer of length 6.
Similarly if
work_tree = "c:/r"
path = "c:/rl/d/a"
Then (since the loop starts by also incrementing the pointer one step),
this would mean that the function would miss checking if "c:/rl/d" could
be the work_tree, arguably this is unlikely though, since it would only
be possible with symlinks on windows.
Fix this by simply avoiding to increment by offset_1st_component() and
wtlen at the same time.
Signed-off-by: Martin Erik Werner <martinerikwerner@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Make sure the marks are not written out when the transport helper
did not finish happily, to avoid leaving a marks file that is out of
sync with the reality.
* fc/transport-helper-sync-error-fix:
t5801 (remote-helpers): cleanup environment sets
transport-helper: fix sync issue on crashes
transport-helper: trivial cleanup
transport-helper: propagate recvline() error pushing
remote-helpers: make recvline return an error
transport-helper: remove barely used xchgline()
Versions of Perl's Getopt::Long module before 2.37 do not contain
this fix that first appeared in Getopt::Long version 2.37:
* Bugfix: With gnu_compat, --foo= will no longer trigger "Option
requires an argument" but return the empty string.
Instead of using --prefix="" use --prefix "" when testing an
explictly empty prefix string in order to work with older versions
of Perl's Getopt::Long module.
Also add a paragraph on this workaround to the documentation of
git-svn itself.
Signed-off-by: Kyle J. McKay <mackyle@gmail.com> Acked-by: Johan Herland <johan@herland.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
ewah_bitmap.c: do not assume size_t and eword_t are the same size
When buffer_grow changes the size of the buffer using realloc,
it first computes and saves the rlw pointer's offset into the
buffer using (uint8_t *) math before the realloc but then
restores it using (eword_t *) math.
In order to do this it's necessary to convert the (uint8_t *)
offset into an (eword_t *) offset. It was doing this by
dividing by the sizeof(size_t). Unfortunately sizeof(size_t)
is not same as sizeof(eword_t) on all platforms.
This causes illegal memory accesses and other bad things to
happen when attempting to use bitmaps on those platforms.
Fix this by dividing by the sizeof(eword_t) instead which
will always be correct for all platforms.
Signed-off-by: Kyle J. McKay <mackyle@gmail.com> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-prompt.sh: don't put unsanitized branch names in $PS1
Both bash and zsh subject the value of PS1 to parameter expansion,
command substitution, and arithmetic expansion. Rather than include
the raw, unescaped branch name in PS1 when running in two- or
three-argument mode, construct PS1 to reference a variable that holds
the branch name. Because the shells do not recursively expand, this
avoids arbitrary code execution by specially-crafted branch names such
as '$(IFS=_;cmd=sudo_rm_-rf_/;$cmd)'.
Signed-off-by: Richard Hansen <rhansen@bbn.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* jx/i18n:
i18n: mention "TRANSLATORS:" marker in Documentation/CodingGuidelines
i18n: only extract comments marked with "TRANSLATORS:"
i18n: remove obsolete comments for translators in diffstat generation
i18n: fix uncatchable comments for translators in date.c
Work around /bin/sh that does not like "return" at the top-level
of a file that is dot-sourced from inside a function definition.
* km/avoid-non-function-return-in-rebase:
Revert "rebase: fix run_specific_rebase's use of "return" on FreeBSD"
rebase: avoid non-function use of "return" on FreeBSD
* ep/shell-command-substitution:
t9362-mw-to-git-utf8.sh: use the $( ... ) construct for command substitution
t9360-mw-to-git-clone.sh: use the $( ... ) construct for command substitution
git-tag.sh: use the $( ... ) construct for command substitution
git-revert.sh: use the $( ... ) construct for command substitution
git-resolve.sh: use the $( ... ) construct for command substitution
git-repack.sh: use the $( ... ) construct for command substitution
git-merge.sh: use the $( ... ) construct for command substitution
git-ls-remote.sh: use the $( ... ) construct for command substitution
git-fetch.sh: use the $( ... ) construct for command substitution
git-commit.sh: use the $( ... ) construct for command substitution
git-clone.sh: use the $( ... ) construct for command substitution
git-checkout.sh: use the $( ... ) construct for command substitution
install-webdoc.sh: use the $( ... ) construct for command substitution
howto-index.sh: use the $( ... ) construct for command substitution
Commit 512477b (tests: use "env" to run commands with temporary env-var
settings) missed some variables in the remote-helpers test. Also
standardize these.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
run_external_diff: use an argv_array for the command line
We currently generate the command-line for the external
command using a fixed-length array of size 10. But if there
is a rename, we actually need 11 elements (10 items, plus a
NULL), and end up writing a random NULL onto the stack.
Rather than bump the limit, let's just use an argv_array, which
makes this sort of error impossible.
Noticed-by: Max L <infthi.inbox@gmail.com> Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Git 2.0: git svn: Set default --prefix='origin/' if --prefix is not given
git-svn by default puts its Subversion-tracking refs directly in
refs/remotes/*. This runs counter to Git's convention of using
refs/remotes/$remote/* for storing remote-tracking branches.
Furthermore, combining git-svn with regular git remotes run the risk of
clobbering refs under refs/remotes (e.g. if you have a git remote
called "tags" with a "v1" branch, it will overlap with the git-svn's
tracking branch for the "v1" tag from Subversion.
Even though the git-svn refs stored in refs/remotes/* are not "proper"
remote-tracking branches (since they are not covered by a proper git
remote's refspec), they clearly represent a similar concept, and would
benefit from following the same convention.
For example, if git-svn tracks Subversion branch "foo" at
refs/remotes/foo, and you create a local branch refs/heads/foo to add
some commits to be pushed back to Subversion (using "git svn dcommit),
then it is clearly unhelpful of Git to throw
warning: refname 'foo' is ambiguous.
every time you checkout, rebase, or otherwise interact with the branch.
The existing workaround for this is to supply the --prefix=quux/ to
git svn init/clone, so that git-svn's tracking branches end up in
refs/remotes/quux/* instead of refs/remotes/*. However, encouraging
users to specify --prefix to work around a design flaw in git-svn is
suboptimal, and not a long term solution to the problem. Instead,
git-svn should default to use a non-empty prefix that saves
unsuspecting users from the inconveniences described above.
This patch will only affect newly created git-svn setups, as the
--prefix option only applies to git svn init (and git svn clone).
Existing git-svn setups will continue with their existing (lack of)
prefix. Also, if anyone somehow prefers git-svn's old layout, they
can recreate that by explicitly passing an empty prefix (--prefix "")
on the git svn init/clone command line.
The patch changes the default value for --prefix from "" to "origin/",
updates the git-svn manual page, and fixes the fallout in the git-svn
testcases.
(Note that this patch might be easier to review using the --word-diff
and --word-diff-regex=. diff options.)
[ew: squashed description of <= 1.9 behavior into manpage]
Suggested-by: Thomas Ferris Nicolaisen <tfnico@gmail.com> Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Eric Wong <normalperson@yhbt.net>
* fc/remote-helper-fixes:
remote-bzr: trivial test fix
remote-bzr: include authors field in pushed commits
remote-bzr: add support for older versions
remote-hg: always normalize paths
remote-helpers: allow all tests running from any dir