Offering Last-modified header for feeds is only half the work, even if
we bail out early on HEAD requests. We should also check that same date
against If-modified-since, and bail out early with 304 Not Modified if
that's the case.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitweb: last-modified time should be commiter, not author
The last-modified time header added by RSS to increase cache hits from
readers should be set to the date the repository was last modified. The
author time in this respect is not a good guess because the last commit
might come from a oldish patch.
Use the committer time for the last-modified header to ensure a more
correct guess of the last time the repository was modified.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The RSS 2.0 specifications defines not one but _two_ dates for its
channel element! Woohoo! Luckily, it seems that consensus seems to be
that if both are present they should be equal, except for some very
obscure and discouraged cases. Since lastBuildDate would make more sense
for us and pubDate seems to be the most commonly used, we defined both
and make them equal.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The RSS 2.0 specification allows an optional managingEditor tag for the
channel, containing the "email address for person responsible for editorial
content", which is basically the project owner.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add <generator> tag to RSS and Atom feed. Versioning info (gitweb/git
core versions, separated by a literal slash) is stored in the
appropriate attribute for the Atom feed, and in the tag content for the
RSS feed.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Define the channel image for the rss feed when the logo or favicon are
defined, preferring the former to the latter. As suggested in the RSS
2.0 specifications, the image's title and link as set to the same as the
channel's.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git: use run_command() to execute dashed externals
We used to simply try calling execvp(); if it succeeded, then we were done
and the new program was running. If it didn't, then we knew that it wasn't
a valid command.
Unfortunately, this interacted badly with the new pager handling. Now that
git remains the parent process and the pager is spawned, git has to hang
around until the pager is finished. We install an atexit handler to do
this, but that handler never gets called if we successfully run execvp.
You could see this behavior by running any dashed external using a pager
(e.g., "git -p stash list"). The command finishes running, but the pager
is still going. In the case of less, it then gets an error reading from
the terminal and exits, potentially leaving the terminal in a broken state
(and not showing the output).
This patch just uses run_command() to try running the dashed external. The
parent git process then waits for the external process to complete and
then handles the pager cleanup as it would for an internal command.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
run_command() returns a single integer specifying either an
error code or the exit status of the spawned program. The
only way to tell the difference is that the error codes are
outside of the allowed range of exit status values.
Rather than make each caller implement the test against a
magic limit, let's provide a macro.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
run_command(): handle missing command errors more gracefully
When run_command() was asked to run a non-existant command, its behavior
varied depending on the platform:
- on POSIX systems, we would fork, and then after the execvp call
failed, we could call die(), which prints a message to stderr and
exits with code 128.
- on Windows, we do a PATH lookup, realize the program isn't there, and
then return ERR_RUN_COMMAND_FORK
The goal of this patch is to make it clear to callers that the specific
error was a missing command. To do this, we will return the error code
ERR_RUN_COMMAND_EXEC, which is already defined in run-command.h, checked
for in several places, but never actually gets set.
The new behavior is:
- on POSIX systems, we exit the forked process with code 127 (the same
as the shell uses to report missing commands). The parent process
recognizes this code and returns an EXEC error. The stderr message is
silenced, since the caller may be speculatively trying to run a
command. Instead, we use trace_printf so that somebody interested in
debugging can see the error that occured.
- on Windows, we check errno, which is already set correctly by
mingw_spawnvpe, and report an EXEC error instead of a FORK error
Thus it is safe to speculatively run a command:
int r = run_command_v_opt(argv, 0);
if (r == -ERR_RUN_COMMAND_EXEC)
/* oops, it wasn't found; try something else */
else
/* we failed for some other reason, error is in r */
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Makefile: Make 'configure --with-expat=path' actually work
While the configure script sets the EXPATDIR environment variable to
whatever value was passed to its option --with-expat as the prefix of
the location of the expat library and headers, the Makefile ignored it.
This patch fixes this bug.
Signed-off-by: Serge van den Boom <svdb@stack.nl> Signed-off-by: Junio C Hamano <gitster@pobox.com>
There is a static function called run_command which
conflicts with the library function in run-command.c; this
isn't a problem currently, but prevents including
run-command.h in git.c.
This patch just renames the static function to something
more specific and non-conflicting.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Move has_commit() from branch to a common location, in preparation for
using it in "git-tag". Rename it to is_descendant_of() to make it more
unique and descriptive.
Signed-off-by: Jake Goulding <goulding@vivisimo.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Moving opt_parse_with_commit() from branch to a common location, in
preparation for using it in tag. Rename it to match naming convention
of other option parsing functions.
Signed-off-by: Jake Goulding <goulding@vivisimo.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
revision traversal: allow UNINTERESTING objects to be missing
Most of the existing codepaths were meant to treat missing uninteresting
objects to be a silently ignored non-error, but there were a few places
in handle_commit() and add_parents_to_list(), which are two key functions
in the revision traversal machinery, that cared:
- When a tag refers to an object that we do not have, we barfed. We
ignore such a tag if it is painted as UNINTERESTING with this change.
- When digging deeper into the ancestry chain of a commit that is already
painted as UNINTERESTING, in order to paint its parents UNINTERESTING,
we barfed if parse_parent() for a parent commit object failed. We can
ignore such a parent commit object.
Update usage statement to remove a no-longer supported option, and to hide two
options (one a no-op, one internal) unless --help-all is used.
Use "test -t 0" instead of "tty -s" to detect when stdin is a terminal. (test
-t 0 is used elsewhere in git-am and in other git shell scripts, tty -s is
not, and appears to be deprecated by POSIX)
Use "test ..." instead of "[ ... ]" and "die <msg>" instead of "echo <msg>
>&2; exit 1" to be consistent with rest of script.
Signed-off-by: Jay Soffian <jaysoffian@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The last test case checks whether unpacked objects receive the time stamp
of the pack file. Due to different implementations of stat(2) by MSYS and
our version in compat/mingw.c, the test fails in about half of the test
runs.
Note the following facts:
- The test uses perl's -M operator to compare the time stamps. Since we
depend on MSYS perl, the result of this operator is based on MSYS's
implementation of the stat(2) call.
- NTFS on Windows records fractional seconds.
- The MSYS implementation of stat(2) *rounds* fractional seconds to full
seconds instead of truncating them. This becomes obvious by comparing the
modification times reported by 'ls --full-time $f' and 'stat $f' for
various files $f.
- Our implementation of stat(2) in compat/mingw.c *truncates* to full
seconds.
The consequence of this is that
- add_packed_git() picks up a truncated whole second modification time
from the pack file time stamp, which is then used for the loose objects,
while the pack file retains its time stamp in fractional seconds;
- but the test case compared the pack file's rounded modification times
to the loose objects' truncated modification times.
And half of the time the rounded modification time is not the same as its
truncated modification time.
The fix is that we replace perl by 'test-chmtime -v +0', which prints the
truncated whole-second mtime without modifying it.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
send-pack: do not send unknown object name from ".have" to pack-objects
test-path-utils: Fix off by one, found by valgrind
get_sha1_basic(): fix invalid memory access, found by valgrind
send-pack: do not send unknown object name from ".have" to pack-objects
v1.6.1 introduced ".have" extension to the protocol to allow the receiving
side to advertise objects that are reachable from refs in the repositories
it borrows from. This was meant to be used by the sending side to avoid
sending such objects; they are already available through the alternates
mechanism.
The client side implementation in v1.6.1, which was introduced with 40c155f (push: prepare sender to receive extended ref information from the
receiver, 2008-09-09) aka v1.6.1-rc1~203^2~1, were faulty in that it did
not consider the possiblity that the repository receiver borrows from
might have objects it does not know about.
This fixes it by refraining from passing missing commits to underlying
pack-objects. Revision machinery may need to be tightened further to
treat missing uninteresting objects as non-error events, but this is an
obvious and safe fix for a maintenance release that is almost good enough.
mergetool: respect autocrlf by using checkout-index
Previously, git mergetool used cat-file which does not perform git to
worktree conversion. This changes mergetool to use git checkout-index
instead which means that the temporary files used for mergetool use the
correct line endings for the platform.
Signed-off-by: Charles Bailey <charles@hashpling.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git mergetool listed some candidates for mergetools twice, depending on
the environment.
This slightly changes the behavior when both KDE_FULL_SESSION and
GNOME_DESKTOP_SESSION_ID are set at the same time; in such a case
meld is used in favor of kdiff3 (the old code favored kdiff3 in such a
case), but it should not matter in practice.
Signed-off-by: Johannes Gilger <heipei@hackvalue.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff.c: output correct index lines for a split diff
A patch that changes the filetype (e.g. regular file to symlink) of a path
must be split into a deletion event followed by a creation event, which
means that we need to have two independent metainfo lines for each.
However, the code reused the single set of metainfo lines.
As the blob object names recorded on the index lines are usually not used
nor validated on the receiving end, this is not an issue with normal use
of the resulting patch. However, when accepting a binary patch to delete
a blob, git-apply verified that the postimage blob object name on the
index line is 0{40}, hence a patch that deletes a regular file blob that
records binary contents to create a blob with different filetype (e.g. a
symbolic link) failed to apply. "git am -3" also uses the blob object
names recorded on the index line, so it would also misbehave when
synthesizing a preimage tree.
This moves the code to generate metainfo lines around, so that two
independent sets of metainfo lines are used for the split halves.
Additional tests by Jeff King.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
tests: Avoid single-shot environment export for shell function invocation
Some shells have issues with a single-shot environment variable export
when invoking a shell function. This fixes the ones I found that invoke
test_must_fail that way.
rebase -i: correctly remember --root flag across --continue
d911d14 (rebase -i: learn to rebase root commit, 2009-01-02) tried to
remember the --root flag across a merge conflict in a broken way.
Introduce a flag file $DOTEST/rebase-root to fix and clarify.
While at it, also make sure $UPSTREAM is always initialized to guard
against existing values in the environment.
[tr: added tests]
Signed-off-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
the Mac OS X C standard library is always thread safe and always
includes the pthread library. So explicitly using -pthread causes an
'unrecognized option' compiler warning.
This patch clears PTHREAD_LIBS if Darwin is detected.
Signed-off-by: Ted Pavlic <ted@tedpavlic.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Windows: Revert to default paths and convert them by RUNTIME_PREFIX
The RUNTIME_PREFIX mechanism allows us to use the default paths on
Windows too. Defining RUNTIME_PREFIX explicitly requests for
translation of paths relative to the executable at runtime.
Signed-off-by: Steffen Prohaska <prohaska@zib.de> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Compute prefix at runtime if RUNTIME_PREFIX is set
This commit adds support for relocatable binaries (called
RUNTIME_PREFIX). Such binaries can be moved together with the
system configuration files to a different directory, as long as the
relative paths from the binary to the configuration files is
preserved. This functionality is essential on Windows where we
deliver git binaries with an installer that allows to freely choose
the installation location.
If RUNTIME_PREFIX is unset we use the static prefix. This will be
the default on Unix. Thus, the behavior on Unix will remain
identical to the old implementation, which used to add the prefix
in the Makefile.
If RUNTIME_PREFIX is set the prefix is computed from the location
of the executable. In this case, system_path() tries to strip
known directories that executables can be located in from the path
of the executable. If the path is successfully stripped it is used
as the prefix. For example, if the executable is
"/msysgit/bin/git" and BINDIR is "bin", then the prefix computed is
"/msysgit".
If the runtime prefix computation fails, we fall back to the static
prefix specified in the makefile. This can be the case if the
executable is not installed at a known location. Note that our
test system sets GIT_CONFIG_NOSYSTEM to tell git to ignore global
configuration files during testing. Hence testing does not trigger
the fall back.
Note that RUNTIME_PREFIX only works on Windows, though adding
support on Unix should not be too hard. The implementation
requires argv0_path to be set to an absolute path. argv0_path must
point to the directory of the executable. We use assert() to
verify this in debug builds. On Windows, the wrapper for main()
(see compat/mingw.h) guarantees that argv0_path is correctly
initialized. On Unix, further work is required before
RUNTIME_PREFIX can be enabled.
Signed-off-by: Steffen Prohaska <prohaska@zib.de> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Modify setup_path() to only add git_exec_path() to PATH
Searching git programs only in the highest priority location is
sufficient. It does not make sense that some of the required
programs are located at the highest priority location but other
programs are picked up from a lower priority exec-path. If
exec-path is overridden a complete set of commands should be
provided, otherwise several different versions could get mixed,
which is likely to cause confusion.
If a user explicitly overrides the default location (by --exec-path
or GIT_EXEC_PATH), we now expect that all the required programs are
found there. Instead of adding the directories "argv_exec_path",
"getenv(EXEC_PATH_ENVIRONMENT)", and "system_path(GIT_EXEC_PATH)"
to PATH, we now rely on git_exec_path(), which implements the same
order, but only returns the highest priority location to search for
executables.
Accessing only the location with highest priority is also required
for testing executables built with RUNTIME_PREFIX. The call to
system_path() should be avoided if RUNTIME_PREFIX is set and the
executable is not installed at its final destination. Because we
test before installing, we want to avoid calling system_path()
during tests. The modifications in this commit avoid calling
system_path(GIT_EXEC_PATH) if a higher-priority location is
provided, which is the case when running the tests.
Signed-off-by: Steffen Prohaska <prohaska@zib.de> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add calls to git_extract_argv0_path() in programs that call git_config_*
Programs that use git_config need to find the global configuration.
When runtime prefix computation is enabled, this requires that
git_extract_argv0_path() is called early in the program's main().
This commit adds the necessary calls.
Signed-off-by: Steffen Prohaska <prohaska@zib.de> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Refactor git_set_argv0_path() to git_extract_argv0_path()
This commit moves the code that computes the dirname of argv[0]
from git.c's main() to git_set_argv0_path() and renames the function
to git_extract_argv0_path(). This makes the code in git.c's main
less cluttered, and we can use the dirname computation from other
main() functions too.
[ spr:
- split Steve's original commit and wrote new commit message.
- Integrated Johannes Schindelin's cca1704897e7fdb182f68d4c48a437c5d7bc5203 while rebasing onto master.
]
Signed-off-by: Steve Haslam <shaslam@lastminute.com> Signed-off-by: Steffen Prohaska <prohaska@zib.de> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Move computation of absolute paths from Makefile to runtime (in preparation for RUNTIME_PREFIX)
This commit prepares the Makefile for relocatable binaries (called
RUNTIME_PREFIX). Such binaries will be able to be moved together
with the system configuration files to a different directory,
requiring to compute the prefix at runtime.
In a first step, we make all paths relative in the Makefile and
teach system_path() to add the prefix instead. We used to compute
absolute paths in the Makefile and passed them to C as defines. We
now pass relative paths to C and call system_path() to add the
prefix at runtime.
Signed-off-by: Steffen Prohaska <prohaska@zib.de> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Include examples of using HEAD. The order of examples
introduces new concepts one by one. This pushes the
example of deleting a ref to the end of the list.
Documentation: simplify refspec format description
The refspec format description was a mix of regexp and BNF, making it
very difficult to read. The format was also wrong: it did not show
that each part of a refspec is optional in different situations.
Rather than having a confusing grammar, just present the format in
informal prose.
Signed-off-by: Anders Melchiorsen <mail@cup.kalibalik.dk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
rebase -i --root: fix check for number of arguments
If we are not rebasing with --root, then $# can only be either 1 (base)
or 2 (base and the name of the branch to be rebased).
If we are rebasing with --root, then it is Ok if $# is 0 (rebase the
current branch down to everything) or 1 (rebase the named branch down to
everything).
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Acked-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This new option tells 'git-am' to ignore the date header field
recorded in the format-patch output. The commits will have the
timestamp when they are created instead.
You can work a lot in one day to accumulate many changes, but
apply and push to the public repository only some of them at
the end of the first day. Then next day you can spend all your
working hours reading comics or chatting with your coworkers,
and apply your remaining patches from the previous day using
this option to pretend that you have been working at the end
of the day.
Signed-off-by: しらいしななこ <nanako3@lavabit.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* kb/lstat-cache:
lstat_cache(): introduce clear_lstat_cache() function
lstat_cache(): introduce invalidate_lstat_cache() function
lstat_cache(): introduce has_dirs_only_path() function
lstat_cache(): introduce has_symlink_or_noent_leading_path() function
lstat_cache(): more cache effective symlink/directory detection
* js/diff-color-words:
Change the spelling of "wordregex".
color-words: Support diff.wordregex config option
color-words: make regex configurable via attributes
color-words: expand docs with precise semantics
color-words: enable REG_NEWLINE to help user
color-words: take an optional regular expression describing words
color-words: change algorithm to allow for 0-character word boundaries
color-words: refactor word splitting and use ALLOC_GROW()
Add color_fwrite_lines(), a function coloring each line individually
Added a test for this option, similar to (and based on) t9133 about
ignorance of .git directories
Signed-off-by: Vitaly "_Vi" Shukela <public_vi@tut.by> Acked-by: Eric Wong <normalperson@yhbt.net>
[ew: replaced 'echo -e' with printf so it works on POSIX shells]
[ew: added Vitaly to copyright even though it's based on my test]
This will be useful when somebody want to checkout something partially from
repository with some non-standart layout or exclude some files from it.
Example: repository has structure /module-{a,b,c}/{trunk,branches,tags}/...
Modules are interdependent, and you want it to be single repostory (to commit
to all modules simultaneously and view complete history), but do not want
branches and tags be checked out into working copy.
Other use case is excluding some large blobs.
The quirk for now is that user must specify this option every fetch/rebase;
in other case he may get extra files or "file not found" errors. It may be
will be resolved by adding regular expression to .git/config into
[svn-remote ...] to make it persistent.
Signed-off-by: Vitaly "_Vi" Shukela <public_vi@tut.by> Acked-by: Eric Wong <normalperson@yhbt.net>
[ew: replaced 4-space indent with tabs]
[ew: prefixed $ignore_regex with an underscore to be consistent
with other globals in git-svn]
[ew: rearranged functions to minimize diff and removed prototype
usage to be consistent with the rest of git-svn (and other
Perl code in git (and they're ugly to me)]
Remove a call to git-log that I introduced for debugging and that
accidentally made it into d18ba22 (sha1_name: support @{-N} syntax in
get_sha1(), 2009-01-17).
Signed-off-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
http-push: refactor lock-related headers creation for curl requests
DAV-related headers (more specifically, headers related to the lock token,
namely, If, Lock-Token, and Timeout) for curl requests are created and
allocated individually, eg a "if_header" variable for the "If: " header, a
"timeout_header" variable for the "Timeout: " header.
This patch provides a new function ("get_dav_token_headers") that creates
these header, saving methods from allocating memory, and from issuing a
"curl_slist_append()" call. The temporary string storage given to
curl_slist_append() is freed much earlier than the previous code with this
patch, but this change is safe, because curl_slist_append() keeps a copy
of the given string.
In part, this patch also addresses the fact that commit 753bc91 (Remove
the requirement opaquelocktoken uri scheme) did not update memory
allocations for DAV-related headers.
Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* tr/maint-no-index-fixes:
diff --no-index -q: fix endless loop
diff --no-index: test for pager after option parsing
diff: accept -- when using --no-index
* am/maint-push-doc:
Documentation: avoid using undefined parameters
Documentation: mention branches rather than heads
Documentation: remove a redundant elaboration
Documentation: git push repository can also be a remote
Cloning an empty repository manually (that is, doing 'git init' and
then doing all configuration by hand) can be a lot of work. Save the
user this work by allowing the cloning of empty repositories.
Signed-off-by: Sverre Rabbelier <srabbelier@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The documentation for git-describe says the default abbreviation is 8
hexadecimal digits while cache.c clearly shows DEFAULT_ABBREV set to 7.
This patch corrects the documentation.
Signed-off-by: Boyd Stephen Smith Jr <bss@iguanasuicide.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since ea27a18 (spawn pager via run_command interface), the
original git process actually does git work, and the pager
is a child process (actually, on Windows it has always been
that way, since Windows lacks fork). After spawning the
pager, we register an atexit() handler that waits for the
pager to finish.
Unfortunately, that handler does not always run. In
particular, if git is killed by a signal, then we exit
immediately. The calling shell then thinks that git is done;
however, the pager is still trying to run and impact the
terminal. The result can be seen by running a long git
process with a pager (e.g., "git log -p") and hitting ^C.
Depending on your config, you should see the shell prompt,
but pressing a key causes the pager to do any terminal
de-initialization sequence.
This patch just intercepts any death-dealing signals and
waits for the pager before dying. Under typical less
configuration, that means hitting ^C will cause git to stop
generating output, but the pager will keep running.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The current code is very inconsistent about which signals
are caught for doing cleanup of temporary files and lock
files. Some callsites checked only SIGINT, while others
checked a variety of death-dealing signals.
This patch factors out those signals to a single function,
and then calls it everywhere. For some sites, that means
this is a simple clean up. For others, it is an improvement
in that they will now properly clean themselves up after a
larger variety of signals.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a piece of code wanted to do some cleanup before exiting
(e.g., cleaning up a lockfile or a tempfile), our usual
strategy was to install a signal handler that did something
like this:
do_cleanup(); /* actual work */
signal(signo, SIG_DFL); /* restore previous behavior */
raise(signo); /* deliver signal, killing ourselves */
For a single handler, this works fine. However, if we want
to clean up two _different_ things, we run into a problem.
The most recently installed handler will run, but when it
removes itself as a handler, it doesn't put back the first
handler.
This patch introduces sigchain, a tiny library for handling
a stack of signal handlers. You sigchain_push each handler,
and use sigchain_pop to restore whoever was before you in
the stack.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are two pieces of code that create tempfiles for diff:
run_external_diff and run_textconv. The former cleans up its
tempfiles in the face of premature death (i.e., by die() or
by signal), but the latter does not. After this patch, they
will both use the same cleanup routines.
To make clear what the change is, let me first explain what
happens now:
- run_external_diff uses a static global array of 2
diff_tempfile structs (since it knows it will always
need exactly 2 tempfiles). It calls prepare_temp_file
(which doesn't know anything about the global array) on
each of the structs, creating the tempfiles that need to
be cleaned up. It then registers atexit and signal
handlers to look through the global array and remove the
tempfiles. If it succeeds, it calls the handler manually
(which marks the tempfile structs as unused).
- textconv has its own tempfile struct, which it allocates
using prepare_temp_file and cleans up manually. No
signal or atexit handlers.
The new code moves the installation of cleanup handlers into
the prepare_temp_file function. Which means that that
function now has to understand that there is static tempfile
storage. So what happens now is:
- run_external_diff calls prepare_temp_file
- prepare_temp_file calls claim_diff_tempfile, which
allocates an unused slot from our global array
- prepare_temp_file installs (if they have not already
been installed) atexit and signal handlers for cleanup
- prepare_temp_file sets up the tempfile as usual
- prepare_temp_file returns a pointer to the allocated
tempfile
The advantage being that run_external_diff no longer has to
care about setting up cleanup handlers. Now by virtue of
calling prepare_temp_file, run_textconv gets the same
benefit, as will any future users of prepare_temp_file.
There are also a few side benefits to the specific
implementation:
- we now install cleanup handlers _before_ allocating the
tempfile, closing a race which could leave temp cruft
- when allocating a slot in the global array, we will now
detect a situation where the old slots were not properly
vacated (i.e., somebody forgot to call remove upon
leaving the function). In the old code, such a situation
would silently overwrite the tempfile names, meaning we
would forget to clean them up. The new code dies with a
bug warning.
- we make sure only to install the signal handler once.
This isn't a big deal, since we are just overwriting the
old handler, but will become an issue when a later patch
converts the code to use sigchain
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We had defined some SIG_FOO macros that appear in the code, but that are
not supported on Windows, in order to make the code compile. But a
subsequent change will assert that a signal number is non-zero. We now
use the signal numbers that are commonly used on POSIX systems.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* am/maint-push-doc:
Documentation: avoid using undefined parameters
Documentation: mention branches rather than heads
Documentation: remove a redundant elaboration
Documentation: git push repository can also be a remote
* sb/hook-cleanup:
run_hook(): allow more than 9 hook arguments
run_hook(): check the executability of the hook before filling argv
api-run-command.txt: talk about run_hook()
Move run_hook() from builtin-commit.c into run-command.c (libgit)
checkout: don't crash on file checkout before running post-checkout hook
bash completion: add 'rename' subcommand to git-remote
Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de> Acked-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When diff is invoked with --color-words (w/o =regex), use the regular
expression the user has configured as diff.wordregex.
diff drivers configured via attributes take precedence over the
diff.wordregex-words setting. If the user wants to change them, they have
their own configuration variables.
Signed-off-by: Boyd Stephen Smith Jr <bss@iguanasuicide.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Simplify parsing branch switching events in reflog
We only accept "checkout: moving from A to B" newer style reflog entries,
in order to pick up A. There is no point computing where B begins at
after running strstr to locate " to ", nor adding 4 and then subtracting 4
from the same pointer.
tutorial-2: Update with the new "git commit" ouput
An earlier commit c5ee71f (commit: more compact summary and without extra
quotes, 2009-01-19) changed the "git commit" output when creating a
commit. This patch updates the example session in the tutorial to match
the new output.
Signed-off-by: Santi Béjar <santi@agolina.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>