Commit eed6ca7 caused a minor regression when it switched to using
tempfile() to generate the temporary compose file. Since tempfile()
creates the file at the time it generates the filename, zero-length
temporary files are being left behind unless --compose is used (in which
case the file is cleaned up).
This patch fixes the regression by not calling tempfile() to generate
the compose filename unless --compose is in use.
Signed-off-by: Jay Soffian <jaysoffian@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* git://git.bogomips.org/git-svn:
git-svn: read the dcommit url from the config file on a per remote basis
git-svn: fix delete+add branch tracking with empty files
git-svn: Create leading directories in create-ignore
Since this timestamp is used to check for racy-clean files, it is
important to keep it uptodate.
For the 'git checkout' command without the '-q' option, this make a
huge difference. Before, each and every file which was updated, was
racy-clean after the call to unpack_trees() and write_index() but
before the GIT process ended.
And because of the call to show_local_changes() in builtin-checkout.c,
we ended up reading those files back into memory, doing a SHA1 to
check if the files was really different from the index. And, of
course, no file was different.
With this fix, 'git checkout' without the '-q' option should now be
almost as fast as with the '-q' option, but not quite, as we still do
some few lstat(2) calls more without the '-q' option.
Below is some average numbers for 10 checkout's to v2.6.27 and 10 to
v2.6.25 of the Linux kernel, to show the difference:
before (git version 1.6.2.rc1.256.g58a87):
7.860 user 2.427 sys 19.465 real 52.8% CPU faults: 0 major 95331 minor
after:
6.184 user 2.160 sys 17.619 real 47.4% CPU faults: 0 major 38994 minor
Signed-off-by: Kjetil Barvik <barvik@broadpark.no>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a particular changeset affects multiple depot paths, it
will appear multiple times in the output of "p4 changes".
Filter out the duplicates to avoid the extra empty commits that
this otherwise would create.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Acked-by: Simon Hausmann <simon@lst.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Prefixes the branch name with "BARE:" if you're in a
bare repository.
Signed-off-by: Marius Storm-Olsen <git@storm-olsen.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The commit url for dcommit is determined in the following order:
commandline option --commit-url
svn.commiturl
svn-remote.<name>.commiturl
svn-remote.<name>.url
Signed-off-by: Peter Oberndorfer <kumbayo84@arcor.de>
Acked-by: Eric Wong <normalperson@yhbt.net>
Original bug report and test case by Björn Steinbrink.
Björn Steinbrink <B.Steinbrink@gmx.de> wrote:
> Hi Eric,
>
> seems that the empty symlink stuff gets confused about which revision to
> use when looking for the parent's file.
>
> r3 = f1a6fcf6b0a1c4a373d0b2b65a3d70700084f361 (tags/1.0.1)
> Found possible branch point: file:///home/doener/h/svn/tags/1.0 => file:///home/doener/h/svn/branches/1.0, 4
> Found branch parent: (1.0) 63ae640ba01014ecbb3df590999ed1fa5914545b
> Following parent with do_switch
> Successfully followed parent
> r5 = 26fcfef5bcced97ab74faf1af7341a2ae0d272aa (1.0)
> Found possible branch point: file:///home/doener/h/svn/branches/1.0 => file:///home/doener/h/svn/tags/1.0.1, 5
> Found branch parent: (tags/1.0.1) 26fcfef5bcced97ab74faf1af7341a2ae0d272aa
> Following parent with do_switch
> Scanning for empty symlinks, this may take a while if you have many empty files
> You may disable this with `git config svn.brokenSymlinkWorkaround false'.
> This may be done in a different terminal without restarting git svn
> Filesystem has no item: File not found: revision 3, path '/branches/1.0/file' at /usr/local/libexec/git-core/git-svn line 3318
>
> Note how it tries to look at revision 3 instead of revision 5 (which it
> correctly detected as the parent). The import succeeds when
> svn.brokenSymlinkWorkaround is set to false. Testcase below.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Since SVN tracks empty directories and git does not, we can not assume
that the directory exists when creating .gitignore files.
Signed-off-by: Brian Gernhardt <benji@silverinsanity.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Doing so would be incoherent since --get-color would pick a color slot
and ignore the variable type option (e.g. --bool), and the type would
require a variable name.
Suggested by Junio C Hamano.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Only --bool, --int, or --bool-or-int can be used, but not any
combination of them.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Either --global, --system, or --file can be used, but not any
combination.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch has benefited from comments by Johannes
Schindelin and Junio C Hamano.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In preparation for parseopt.
Also remove some unecessary comments since the usage is described in the
documentation.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Essentially this replaces 'file' with 'prefix' in the cases where the
variable is used as a prefix, which is consistent with other git
commands.
When using the --list option general errors where not properly reported,
only errors related with the 'file'. Now they are reported, and 'file'
is irrelevant.
That reduces the rest of 'file' usage to nothing, therefore now only
'prefix' remains.
Suggested by Johannes Schindelin.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently git_config() returns an error if there is no repo config file
available (cwd is not a git repo); it will correctly parse the system
and global config files, but still return an error.
It doesn't affect anything else since almost nobody is checking for the
return code (with the exception of 'git remote update').
A reorganization in 'git config' would benefit from being able to
properly detect errors in git_config() without the noise generated when
cwd is not a git repo.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
For deep threading mode, i.e., the mode that gives a thread structured
like
+ [PATCH 0/n] Cover letter
`-+ [PATCH 1/n] First patch
`-+ [PATCH 2/n] Second patch
`-+ ...
we currently have to use 'git send-email --thread' (the default). On
the other hand, format-patch also has a --thread option which gives
shallow mode, i.e.,
+ [PATCH 0/n] Cover letter
|-+ [PATCH 1/n] First patch
|-+ [PATCH 2/n] Second patch
...
To reduce the confusion resulting from having two indentically named
features in different tools giving different results, let format-patch
take an optional argument '--thread=deep' that gives the same output
as 'send-mail --thread'. With no argument, or 'shallow', behave as
before. Also add a configuration variable format.thread with the same
semantics.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, format-patch --thread --cover-letter --in-reply-to $parent
makes all mails, including the cover letter, a reply to $parent.
However, we would want the reader to consider the cover letter above
all the patches.
This changes the semantics so that only the cover letter is a reply to
$parent, while all the patches are formatted as replies to the cover
letter.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, format-patch can only track a single reference (the
In-Reply-To:) for each mail. To ensure proper threading, we should
list all known references for every mail.
Change the rev_info.ref_message_id field to a string_list, so that we
can append references at will, and change the output formatting
routines to print all of them in the References: header. The last
entry in the list is implicitly assumed to be the In-Reply-To:, which
gives output consistent with RFC 2822:
The "References:" field will contain the contents of the parent's
"References:" field (if any) followed by the contents of the
parent's "Message-ID:" field (if any).
Note that this is just preparatory work; nothing uses it yet, so all
"References:" fields in the output are still only one deep.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* 'for-junio' of git://source.winehq.org/~julliard/git/git:
Add a README in the contrib/emacs directory.
git.el: Improve the confirmation message on remove and revert.
git.el: Make sure that file lists are sorted as they are created.
This avoids a possibly redundant sort in git-update-status-files and
git-status-filenames-map, and allows callers to continue using the
list without having to copy it.
It also fixes the confusing success messages reported by Brent
Goodrick.
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Talking about --date, one thing I wanted for the 1234567890 date was to
get things in the raw format. Sure, you get them with --pretty=raw, but it
felt a bit sad that you couldn't just ask for the date in raw format.
So here's a throw-away patch (meaning: I won't be re-sending it, because I
really don't think it's a big deal) to add "--date=raw". It just prints
out the internal raw git format - seconds since epoch plus timezone (put
another way: 'date +"%s %z"' format)
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
So far, git-patch-id was untested. Add some simple checks for output
format and patch (in)equality.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-fsck, of all tools, has very few tests. This adds some more:
* a corrupted object;
* a branch pointing to a non-commit;
* a tag pointing to a nonexistent object;
* and a tag pointing to an object of a type other than what the tag
itself claims.
Only the first two are caught. At least the third probably should,
too, but currently slips through.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
So far there were no tests checking that log --graph actually works.
Note that the tests strip trailing whitespace, as the current --graph
emits trailing whitespace on lines that do not contain anything but
graph lines.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is only a very rudimentary test, but it was untested before.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With gcc's --coverage option, we can perform automatic coverage data
collection for the test suite.
Add a new Makefile target 'coverage' that scraps all previous coverage
results, recompiles git with the required compiler/linker flags (in
addition to any flags you specify manually), then runs the test suite
and compiles a report.
The compilation must be done with all optimizations disabled, since
inlined functions (and for line-by-line coverage, also optimized
branches/loops) break coverage tracking.
The tests are run serially (with -j1). The coverage code should
theoretically allow concurrent access to its data files, but the
author saw random test failures. Obviously this could be improved.
The report currently consists of a list of functions that were never
executed during the tests, which is written to
'coverage-untested-functions'. Once this list becomes reasonably
short, we would also want to look at branches that were never taken.
Currently only toplevel *.c files are considered. It would be nice to
at least include xdiff, but --coverage did not save data to
subdirectories on the system used to write this (gcc 4.3.2).
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Just saying that index.lock exists doesn't tell the user _what_ to do
to fix the problem. We should give an indication that it's normally
safe to delete index.lock after making sure git isn't running here.
Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This was introduced in 85af7929ee but
not documented outside the commit message.
Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The current implementation only hyperlinks the first hash on
a given line of the commit message. It seems sensible to
highlight all of them if there are multiple, and it seems
plausible that there would be multiple even with a tidy line
length limit, because they can be abbreviated as short as 8
characters.
Benchmark:
I wanted to make sure that using the 'e' switch to the Perl regex
wasn't going to kill performance, since this is called once per commit
message line displayed.
In all three A/B scenarios I tried, the A and B yielded the same
results within 2%, where A is the version of code before this patch
and B is the version after.
1: View a commit message containing the last 1000 commit hashes
2: View a commit message containing 1000 lines of 40 dots to avoid
hyperlinking at the same message length
3: View a short merge commit message with a few lines of text and
no hashes
All were run in CGI mode on my sub-production hardware on a recent
clone of git.git. Numbers are the average of 10 reqests per second
with the first request discarded, since I expect this change to affect
primarily CPU usage. Measured with ApacheBench.
Note that the web page rendered was the same; while the new code
supports multiple hashes per line, there was at most one per line.
The primary purpose of scenarios 2 and 3 were to verify that the
addition of 1000 commit messages had an impact on how much of the time
was spent rendering commit messages. They were all within 2% of 0.80
requests per second (much faster).
So I think the patch has no noticeable effect on performance.
Signed-off-by: Marcel M. Cary <marcel@oak.homeunix.org>
Acked-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
At least for the author of this patch, the logic in system_path() was
too hard to understand. Using the function strip_path_suffix() documents
the idea of the code better.
The real change is to add the suffix "git", so that a runtime prefix will
be computed correctly even when the executable was called in /git/ as is
the case in msysGit (Windows insists to search the current directory
before the PATH when looking for an executable).
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function strip_path_suffix() will try to strip a given suffix from
a given path. The suffix must start at a directory boundary (i.e. "core"
is not a path suffix of "libexec/git-core", but "git-core" is).
Arbitrary runs of directory separators ("slashes") are assumed identical.
Example:
strip_path_suffix("C:\\msysgit/\\libexec\\git-core",
"libexec///git-core", &prefix)
will set prefix to "C:\\msysgit" and return 0.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
t4014 tests format-patch --thread since 7d812145, but the tests were
ineffective right from the start at least for bash and dash. The
loops of the form
for ...; do something || break; done
introduced by 7d812145 and 5d02294 always exit with status 0, even if
'something' failed, because 'break' returns 0 unless there was no loop
to break.
We take a rather different approach that uses an admittedly heinous
inline Perl script to mangle all interesting information into a format
that is invariant between runs. We can then test the full patch
sequence in one go (with --stdout), doing away with the loop problem.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If we inside verify_uptodate() can already tell from the ce entry that
it is already uptodate by testing it with ce_uptodate(ce), there is no
need to call lstat(2) and ie_match_stat() afterwards.
And, reading from the commit log message from:
commit eadb583134
Author: Junio C Hamano <gitster@pobox.com>
Date: Fri Jan 18 23:45:24 2008 -0800
Avoid running lstat(2) on the same cache entry.
this also seems to be correct usage of the ce_uptodate() macro
introduced by that patch.
This will avoid lots of lstat(2) calls in some cases, for example
by running the 'git checkout' command.
Signed-off-by: Kjetil Barvik <barvik@broadpark.no>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since the filesystem ext4 is now defined as stable in Linux v2.6.28,
and ext4 supports nanonsecond resolution timestamps natively, it is
time to make USE_NSEC work as expected.
This will make racy git situations less likely to happen. For 'git
checkout' this means it will be less likely that we have to open, read
the contents of the file into RAM, and check if file is really
modified or not. The result sould be a litle less used CPU time, less
pagefaults and a litle faster program, at least for 'git checkout'.
Since the number of possible racy git situations would increase when
disks gets faster, this patch would be more and more helpfull as times
go by. For a fast Solid State Disk, this patch should be helpfull.
Note that, when file operations starts to take less than 1 nanosecond,
one would again start to get more racy git situations.
For more info on racy git, see Documentation/technical/racy-git.txt
For more info on ext4, see http://kernelnewbies.org/Ext4
Signed-off-by: Kjetil Barvik <barvik@broadpark.no>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
'struct cache' does not have a 'usec' member, but a 'unsigned int
nsec' member. Simmilar 'struct stat' does not have a 'st_mtim.usec'
member, and we should instead use 'st_mtim.tv_nsec'.
Signed-off-by: Kjetil Barvik <barvik@broadpark.no>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We were not testing the output of "git branch" anywhere.
Not only does this not protect us against regressions in the
output, but we are not exercising code paths which may have
bugs (such as the one fixed by 45e2b61).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 45e2b61 fixed the initialization of a "len" struct
parameter via strlen. We can use that to clean up what is
now 3 strlens in a 6-line sequence.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Below is oprofile output from GIT command 'git chekcout -q my-v2.6.25'
(move from tag v2.6.27 to tag v2.6.25 of the Linux kernel):
CPU: Core 2, speed 1999.95 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit
mask of 0x00 (Unhalted core cycles) count 20000
Counted INST_RETIRED_ANY_P events (number of instructions retired) with a
unit mask of 0x00 (No unit mask) count 20000
CPU_CLK_UNHALT...|INST_RETIRED:2...|
samples| %| samples| %|
------------------------------------
409247 100.000 342878 100.000 git
CPU_CLK_UNHALT...|INST_RETIRED:2...|
samples| %| samples| %|
------------------------------------
260476 63.6476 257843 75.1996 libz.so.1.2.3
100876 24.6492 64378 18.7758 kernel-2.6.28.4_2.vmlinux
30850 7.5382 7874 2.2964 libc-2.9.so
14775 3.6103 8390 2.4469 git
2020 0.4936 4325 1.2614 libcrypto.so.0.9.8
191 0.0467 32 0.0093 libpthread-2.9.so
58 0.0142 36 0.0105 ld-2.9.so
1 2.4e-04 0 0 libldap-2.3.so.0.2.31
Detail list of the top 20 function entries (libz counted in one blob):
CPU_CLK_UNHALTED INST_RETIRED_ANY_P
samples % samples % image name symbol name
260476 63.6862 257843 75.2725 libz.so.1.2.3 /lib/libz.so.1.2.3
16587 4.0555 3636 1.0615 libc-2.9.so memcpy
7710 1.8851 277 0.0809 libc-2.9.so memmove
3679 0.8995 1108 0.3235 kernel-2.6.28.4_2.vmlinux d_validate
3546 0.8670 2607 0.7611 kernel-2.6.28.4_2.vmlinux __getblk
3174 0.7760 1813 0.5293 libc-2.9.so _int_malloc
2396 0.5858 3681 1.0746 kernel-2.6.28.4_2.vmlinux copy_to_user
2270 0.5550 2528 0.7380 kernel-2.6.28.4_2.vmlinux __link_path_walk
2205 0.5391 1797 0.5246 kernel-2.6.28.4_2.vmlinux ext4_mark_iloc_dirty
2103 0.5142 1203 0.3512 kernel-2.6.28.4_2.vmlinux find_first_zero_bit
2077 0.5078 997 0.2911 kernel-2.6.28.4_2.vmlinux do_get_write_access
2070 0.5061 514 0.1501 git cache_name_compare
2043 0.4995 1501 0.4382 kernel-2.6.28.4_2.vmlinux rcu_irq_exit
2022 0.4944 1732 0.5056 kernel-2.6.28.4_2.vmlinux __ext4_get_inode_loc
2020 0.4939 4325 1.2626 libcrypto.so.0.9.8 /usr/lib/libcrypto.so.0.9.8
1965 0.4804 1384 0.4040 git patch_delta
1708 0.4176 984 0.2873 kernel-2.6.28.4_2.vmlinux rcu_sched_grace_period
1682 0.4112 727 0.2122 kernel-2.6.28.4_2.vmlinux sysfs_slab_alias
1659 0.4056 290 0.0847 git find_pack_entry_one
1480 0.3619 1307 0.3816 kernel-2.6.28.4_2.vmlinux ext4_writepage_trans_blocks
Notice the memmove line, where the CPU did 7710 / 277 = 27.8 cycles
per instruction, and compared to the total cycles spent inside the
source code of GIT for this command, all the memmove() calls
translates to (7710 * 100) / 14775 = 52.2% of this.
Retesting with a GIT program compiled for gcov usage, I found out that
the memmove() calls came from remove_index_entry_at() in read-cache.c,
where we have:
memmove(istate->cache + pos,
istate->cache + pos + 1,
(istate->cache_nr - pos) * sizeof(struct cache_entry *));
remove_index_entry_at() is called 4902 times from check_updates() in
unpack-trees.c, and each time called we move each cache_entry pointers
(from the removed one) one step to the left.
Since we have 28828 entries in the cache this time, and if we on
average move half of them each time, we in total move approximately
4902 * 0.5 * 28828 * 4 = 282 629 712 bytes, or twice this amount if
each pointer is 8 bytes (64 bit).
OK, is seems that the function check_updates() is called 28 times, so
the estimated guess above had been more correct if check_updates() had
been called only once, but the point is: we get lots of bytes moved.
To fix this, and use an O(N) algorithm instead, where N is the number
of cache_entries, we delete/remove all entries in one loop through all
entries.
From a retest, the new remove_marked_cache_entries() from the patch
below, ended up with the following output line from oprofile:
46 0.0105 15 0.0041 git remove_marked_cache_entries
If we can trust the numbers from oprofile in this case, we saved
approximately ((7710 - 46) * 20000) / (2 * 1000 * 1000 * 1000) = 0.077
seconds CPU time with this fix for this particular test. And notice
that now the CPU did only 46 / 15 = 3.1 cycles/instruction.
Signed-off-by: Kjetil Barvik <barvik@broadpark.no>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
tests: fix "export var=val"
Skip timestamp differences for diff --no-index
Documentation/git-push: --all, --mirror, --tags can not be combined
Some shells do not like "export var=val"; the right way to write
it is to do an assignment and then export just the variable name.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
