git.git
16 years agoMerge branch 'np/maint-safer-pack'
Junio C Hamano [Wed, 3 Sep 2008 00:46:48 +0000 (17:46 -0700)]
Merge branch 'np/maint-safer-pack'

* np/maint-safer-pack:
  fixup_pack_header_footer(): use nicely aligned buffer sizes
  index-pack: use fixup_pack_header_footer()'s validation mode
  pack-objects: use fixup_pack_header_footer()'s validation mode
  improve reliability of fixup_pack_header_footer()
  pack-objects: improve returned information from write_one()

16 years agoMerge branch 'rf/man-env'
Junio C Hamano [Wed, 3 Sep 2008 00:45:41 +0000 (17:45 -0700)]
Merge branch 'rf/man-env'

* rf/man-env:
  builtin-help: fallback to GIT_MAN_VIEWER before man

16 years agoMerge branch 'jc/author-nickname'
Junio C Hamano [Wed, 3 Sep 2008 00:45:28 +0000 (17:45 -0700)]
Merge branch 'jc/author-nickname'

* jc/author-nickname:
  git commit --author=$name: look $name up in existing commits

16 years agoMerge branch 'dp/pid-uintmax'
Junio C Hamano [Wed, 3 Sep 2008 00:45:25 +0000 (17:45 -0700)]
Merge branch 'dp/pid-uintmax'

* dp/pid-uintmax:
  cast pid_t's to uintmax_t to improve portability

16 years agogitweb: ref markers link to named shortlogs
Giuseppe Bilotta [Tue, 2 Sep 2008 19:47:05 +0000 (21:47 +0200)]
gitweb: ref markers link to named shortlogs

This patch turns ref markers for tags and heads into links to
appropriate views for the ref name, depending on current context.

For annotated tags, we link to the tag view, unless that's the current
view, in which case we switch to shortlog. For other refs, we prefer the
current view if it's history or (short)log, and default to shortlog
otherwise.

Appropriate changes are made in the CSS to prevent ref markers from
being annoyingly blue and underlined, unless hovered. A visual
indication of the target view difference is also implemented by making
annotated tags show up in italic.

Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
Acked-by: Petr Baudis <pasky@suse.cz>
Acked-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'maint'
Junio C Hamano [Wed, 3 Sep 2008 00:10:08 +0000 (17:10 -0700)]
Merge branch 'maint'

* maint:
  Makefile: add merge_recursive.h to LIB_H
  Improve documentation for --dirstat diff option
  Bring local clone's origin URL in line with that of a remote clone
  Documentation: minor cleanup in a use case in 'git stash' manual
  Documentation: fix disappeared lines in 'git stash' manpage
  Documentation: fix reference to a for-each-ref option

16 years agoMakefile: add merge_recursive.h to LIB_H
Miklos Vajna [Tue, 2 Sep 2008 23:49:05 +0000 (01:49 +0200)]
Makefile: add merge_recursive.h to LIB_H

When modifying merge-recursive.h, for example builtin-merge-recursive.c
have to be recompiled which was not true till now, causing various
runtime errors using an incremental build.

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoImprove documentation for --dirstat diff option
Heikki Orsila [Mon, 1 Sep 2008 00:50:28 +0000 (03:50 +0300)]
Improve documentation for --dirstat diff option

Signed-off-by: Heikki Orsila <heikki.orsila@iki.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoBring local clone's origin URL in line with that of a remote clone
Johan Herland [Mon, 1 Sep 2008 19:07:33 +0000 (21:07 +0200)]
Bring local clone's origin URL in line with that of a remote clone

On a local clone, "git clone" would use the fully DWIMmed path as the origin
URL in the resulting repo. This was slightly inconsistent with the case of a
remote clone where the _given_ URL was used as the origin URL (because the
DWIMming was done remotely, and was therefore not available to "git clone").

This behaviour caused problems when cloning a local non-bare repo with
relative submodule URLs, because these submodule URLs would then be resolved
against the DWIMmed URL (e.g. "/repo/.git") instead of the given URL (e.g.
"/repo").

This patch teaches "git clone" to use the _given_ URL - instead of the
DWIMmed path - as the origin URL. This causes relative submodule URLs to be
resolved correctly, as long the _given_ URL indicates the correct directory
against which the submodule URLs should be resolved.

The patch also updates a testcase that contained the old-style origin URLs.

Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: minor cleanup in a use case in 'git stash' manual
SZEDER Gábor [Tue, 2 Sep 2008 01:45:01 +0000 (03:45 +0200)]
Documentation: minor cleanup in a use case in 'git stash' manual

There is no need to explicitly pass the file to be committed to 'git
commit', because it's contents is already in the index.

Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: fix disappeared lines in 'git stash' manpage
SZEDER Gábor [Tue, 2 Sep 2008 01:35:24 +0000 (03:35 +0200)]
Documentation: fix disappeared lines in 'git stash' manpage

Asciidoc removes lines starting with a dot when creating manpages.
Since those lines were comments in use case examples showing shell
commands, preceed those lines with a hash sign.

Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: fix reference to a for-each-ref option
SZEDER Gábor [Mon, 1 Sep 2008 21:02:09 +0000 (23:02 +0200)]
Documentation: fix reference to a for-each-ref option

... to match the synopsis section

Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agocast pid_t's to uintmax_t to improve portability
David Soria Parra [Sun, 31 Aug 2008 12:09:39 +0000 (14:09 +0200)]
cast pid_t's to uintmax_t to improve portability

Some systems (like e.g. OpenSolaris) define pid_t as long,
therefore all our sprintf that use %i/%d cause a compiler warning
beacuse of the implicit long->int cast. To make sure that
we fit the limits, we display pids as PRIuMAX and cast them explicitly
to uintmax_t.

Signed-off-by: David Soria Parra <dsp@php.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'maint'
Junio C Hamano [Sun, 31 Aug 2008 23:52:53 +0000 (16:52 -0700)]
Merge branch 'maint'

* maint:
  Document sendemail.envelopesender configuration
  Document clarification: gitmodules, gitattributes
  config.txt: Add missing colons after option name

16 years agoDocument sendemail.envelopesender configuration
Ask Bjørn Hansen [Sun, 31 Aug 2008 20:32:43 +0000 (13:32 -0700)]
Document sendemail.envelopesender configuration

Signed-off-by: Ask Bjørn Hansen <ask@develooper.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocument clarification: gitmodules, gitattributes
Gustaf Hendeby [Sun, 31 Aug 2008 16:00:27 +0000 (18:00 +0200)]
Document clarification: gitmodules, gitattributes

The SYNOPSIS section of gitattibutes and gitmodule fail to clearly
specify the name of the in tree files used.  This patch brings in the
initial `.' and the fact that the `.gitmodules' file should reside at
the top-level of the working tree.

Signed-off-by: Gustaf Hendeby <hendeby@isy.liu.se>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoconfig.txt: Add missing colons after option name
Teemu Likonen [Fri, 29 Aug 2008 07:29:42 +0000 (10:29 +0300)]
config.txt: Add missing colons after option name

gitcvs.usecrlfattr --> gitcvs.usecrlfattr::

This fixes an asciidoc markup issue.

Signed-off-by: Teemu Likonen <tlikonen@iki.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit commit --author=$name: look $name up in existing commits
Junio C Hamano [Wed, 27 Aug 2008 06:13:13 +0000 (23:13 -0700)]
git commit --author=$name: look $name up in existing commits

This allows "git commit --author=$name" to accept a name that is not in
the required "A U Thor <author@example.xz>" format, and use that to look
up an author name that matches from existing commits.

When using this feature, it is the user's responsibility to give a name
that uniquely matches the name s/he wants, as the logic returns the name
from the first matching commit.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'maint'
Junio C Hamano [Sun, 31 Aug 2008 03:31:39 +0000 (20:31 -0700)]
Merge branch 'maint'

* maint:
  gitattributes: -crlf is not binary
  git-apply: Loosen "match_beginning" logic
  Fix example in git-name-rev documentation
  shell: do not play duplicated definition games to shrink the executable
  Fix use of hardlinks in "make install"
  pack-objects: Allow missing base objects when creating thin packs

16 years agogitattributes: -crlf is not binary
Junio C Hamano [Sat, 30 Aug 2008 21:35:15 +0000 (14:35 -0700)]
gitattributes: -crlf is not binary

The description of crlf attribute incorrectly said that "-crlf" means
binary.  It is true that for binary files you would want "-crlf", but
that is not the same thing.

We also have supported attribute macros and via that mechanism a handy
"binary" to specify "-crlf -diff" at the same time.  It was not documented
anywhere as far as I can tell, even though the support was there from
the very beginning.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-apply: Loosen "match_beginning" logic
Junio C Hamano [Sat, 30 Aug 2008 20:20:31 +0000 (13:20 -0700)]
git-apply: Loosen "match_beginning" logic

Even after a handfle attempts, match_beginning logic still has corner
cases:

    1bf1a85 (apply: treat EOF as proper context., 2006-05-23)
    65aadb9 (apply: force matching at the beginning., 2006-05-24)
    4be6096 (apply --unidiff-zero: loosen sanity checks ..., 2006-09-17)
    ee5a317 (Fix "git apply" to correctly enforce "match ..., 2008-04-06)

This is a tricky piece of code.

We still incorrectly enforce "match_beginning" for -U0 matches.
I noticed this while trying out an example sequence from Clemens Buchacher:

    $ echo a >victim
    $ git add victim
    $ echo b >>victim
    $ git diff -U0 >patch
    $ cat patch
    diff --git i/victim w/victim
    index 7898192..422c2b7 100644
    --- i/victim
    +++ w/victim
    @@ -1,0 +2 @@ a
    +b
    $ git apply --cached --unidiff-zero <patch
    $ git show :victim
    b
    a

The change inserts a new line before the second line, but we insist it to
be applied at the beginning.  As the result, the code refuses to apply it
at the original offset, and we end up adding the line at the beginning.

Updates to the test script are by Clemens Buchacher.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix example in git-name-rev documentation
Junio C Hamano [Sat, 30 Aug 2008 20:08:50 +0000 (13:08 -0700)]
Fix example in git-name-rev documentation

Since 59d3f54 (name-rev: avoid "^0" when unneeded, 2007-02-20), name-rev
stopped showing an unnecessary "^0" to dereference a tag down to a commit.
The patch should have made a matching update to the documentation, but we
forgot.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'sp/missing-thin-base' into maint
Junio C Hamano [Sat, 30 Aug 2008 15:38:19 +0000 (08:38 -0700)]
Merge branch 'sp/missing-thin-base' into maint

* sp/missing-thin-base:
  pack-objects: Allow missing base objects when creating thin packs

16 years agoMerge branch 'sb/daemon'
Junio C Hamano [Sat, 30 Aug 2008 15:17:15 +0000 (08:17 -0700)]
Merge branch 'sb/daemon'

* sb/daemon:
  daemon.c: minor style fixup
  git-daemon: rewrite kindergarden, new option --max-connections
  git-daemon: Simplify dead-children reaping logic
  git-daemon: use LOG_PID, simplify logging code
  git-daemon: call logerror() instead of error()

16 years agoMerge branch 'af/maint-install-no-handlink' into maint
Junio C Hamano [Sat, 30 Aug 2008 05:39:25 +0000 (22:39 -0700)]
Merge branch 'af/maint-install-no-handlink' into maint

* af/maint-install-no-handlink:
  Fix use of hardlinks in "make install"
  Makefile: always provide a fallback when hardlinks fail

16 years agoshell: do not play duplicated definition games to shrink the executable
Junio C Hamano [Wed, 20 Aug 2008 01:05:39 +0000 (18:05 -0700)]
shell: do not play duplicated definition games to shrink the executable

Playing with linker games to shrink git-shell did not go well with various
other platforms and compilers.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix use of hardlinks in "make install"
Alex Riesen [Thu, 28 Aug 2008 13:57:32 +0000 (15:57 +0200)]
Fix use of hardlinks in "make install"

The code failed to filter-out git-add properly on platforms were $X is
not empty (ATM there is only one such a platform).

Than it tried to create a hardlink to the file ($execdir/git-add) it just
removed (because git-add is first in the BUILT_INS), so ln failed (but
because stderr was redirected into /dev/null the error was never seen), and
the whole install ended up using "ln -s" instead.

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agofixup_pack_header_footer(): use nicely aligned buffer sizes
Nicolas Pitre [Fri, 29 Aug 2008 20:08:02 +0000 (16:08 -0400)]
fixup_pack_header_footer(): use nicely aligned buffer sizes

It should be more efficient to use nicely aligned buffer sizes, either
for filesystem operations or SHA1 checksums.  Also, using a relatively
small nominal size might allow for the data to remain in L1 cache
between both SHA1_Update() calls.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoindex-pack: use fixup_pack_header_footer()'s validation mode
Nicolas Pitre [Fri, 29 Aug 2008 20:08:01 +0000 (16:08 -0400)]
index-pack: use fixup_pack_header_footer()'s validation mode

When completing a thin pack, a new header has to be written to
the pack and a new SHA1 computed.  Make sure that the SHA1 of what
is being read back matches the SHA1 of what was written for both:
the original pack and the appended objects.

To do so, a couple write_or_die() calls were converted to sha1write()
which has the advantage of doing some buffering as well as handling
SHA1 and CRC32 checksum already.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agopack-objects: use fixup_pack_header_footer()'s validation mode
Nicolas Pitre [Fri, 29 Aug 2008 20:08:00 +0000 (16:08 -0400)]
pack-objects: use fixup_pack_header_footer()'s validation mode

When limiting the pack size, a new header has to be written to the
pack and a new SHA1 computed.  Make sure that the SHA1 of what is being
read back matches the SHA1 of what was written.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoimprove reliability of fixup_pack_header_footer()
Nicolas Pitre [Fri, 29 Aug 2008 20:07:59 +0000 (16:07 -0400)]
improve reliability of fixup_pack_header_footer()

Currently, this function has the potential to read corrupted pack data
from disk and give it a valid SHA1 checksum.  Let's add the ability to
validate SHA1 checksum of existing data along the way, including before
and after any arbitrary point in the pack.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agopack-objects: improve returned information from write_one()
Nicolas Pitre [Fri, 29 Aug 2008 20:07:58 +0000 (16:07 -0400)]
pack-objects: improve returned information from write_one()

This function returns 0 when the current object couldn't be written
due to the pack size limit, otherwise the current offset in the pack.
There is a problem with this approach however, since current object
could be a delta and its delta base might just have been written in
the same write_one() call, but those successfully written objects are
not accounted in the offset variable tracked by the caller. Currently
this is not an issue but a subsequent patch will need this.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'np/verify-pack' into maint
Junio C Hamano [Sat, 30 Aug 2008 04:48:02 +0000 (21:48 -0700)]
Merge branch 'np/verify-pack' into maint

* np/verify-pack:
  discard revindex data when pack list changes

16 years agobuiltin-help: fallback to GIT_MAN_VIEWER before man
Romain Francoise [Fri, 29 Aug 2008 15:00:43 +0000 (17:00 +0200)]
builtin-help: fallback to GIT_MAN_VIEWER before man

In some situations it is useful to be able to switch viewers via the
environment, e.g. in Emacs shell buffers.  So check the GIT_MAN_VIEWER
environment variable and try it before falling back to "man".

Signed-off-by: Romain Francoise <romain@orebokech.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'maint'
Junio C Hamano [Fri, 29 Aug 2008 07:16:39 +0000 (00:16 -0700)]
Merge branch 'maint'

* maint:
  tutorial: gentler illustration of Alice/Bob workflow using gitk
  pretty=format: respect date format options
  make git-shell paranoid about closed stdin/stdout/stderr
  Document gitk --argscmd flag.
  Fix '--dirstat' with cross-directory renaming
  for-each-ref: Allow a trailing slash in the patterns

16 years agogit-p4: Fix checkout bug when using --import-local.
Tor Arvid Lund [Wed, 27 Aug 2008 22:36:12 +0000 (00:36 +0200)]
git-p4: Fix checkout bug when using --import-local.

When this option is passed to git p4 clone, the checkout at the end would
previously fail. This patch fixes it by optionally creating the master branch
from refs/heads/p4/master, which is the correct one for this option.

Signed-off-by: Tor Arvid Lund <torarvid@gmail.com>
Acked-By: Simon Hausmann <simon@lst.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agotutorial: gentler illustration of Alice/Bob workflow using gitk
Paolo Ciarrocchi [Thu, 28 Aug 2008 12:23:52 +0000 (14:23 +0200)]
tutorial: gentler illustration of Alice/Bob workflow using gitk

Update to gitutorial as discussedin the git mailing list:

http://marc.info/?t=121969390900002&r=1&w=2

Signed-off-by: Paolo Ciarrocchi <paolo.ciarrocchi@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agopretty=format: respect date format options
Jeff King [Fri, 29 Aug 2008 00:54:59 +0000 (20:54 -0400)]
pretty=format: respect date format options

When running a command like:

  git log --pretty=format:%ad --date=short

the date option was ignored. This patch causes it to use whatever
format was specified by --date (or by --relative-date, etc), just
as the non-user formats would do.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomake git-shell paranoid about closed stdin/stdout/stderr
Paolo Bonzini [Wed, 27 Aug 2008 15:20:35 +0000 (17:20 +0200)]
make git-shell paranoid about closed stdin/stdout/stderr

It is in general unsafe to start a program with one or more of file
descriptors 0/1/2 closed.  Karl Chen for example noticed that stat_command
does this in order to rename a pipe file descriptor to 0:

    dup2(from, 0);
    close(from);

... but if stdin was closed (for example) from == 0, so that

    dup2(0, 0);
    close(0);

just ends up closing the pipe.  Another extremely rare but nasty problem
would occur if an "important" file ends up in file descriptor 2, and is
corrupted by a call to die().

Fixing this in git was considered to be overkill, so this patch works
around it only for git-shell.  The fix is simply to open all the "low"
descriptors to /dev/null in main.

Signed-off-by: Paolo Bonzini <bonzini@gnu.org>
Acked-by: Stephen R. van den Berg <srb@cuci.nl>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocument gitk --argscmd flag.
Yann Dirson [Thu, 28 Aug 2008 22:00:28 +0000 (00:00 +0200)]
Document gitk --argscmd flag.

This was part of my original patch, but appears to have been lost.

Signed-off-by: Yann Dirson <ydirson@altern.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix '--dirstat' with cross-directory renaming
Linus Torvalds [Thu, 28 Aug 2008 23:19:08 +0000 (16:19 -0700)]
Fix '--dirstat' with cross-directory renaming

The dirstat code depends on the fact that we always generate diffs with
the names sorted, since it then just does a single-pass walk-over of the
sorted list of names and how many changes there were. The sorting means
that all files are nicely grouped by directory.

That all works fine.

Except when we have rename detection, and suddenly the nicely sorted list
of pathnames isn't all that sorted at all. And now the single-pass dirstat
walk gets all confused, and you can get results like this:

  [torvalds@nehalem linux]$ git diff --dirstat=2 -M v2.6.27-rc4..v2.6.27-rc5
     3.0% arch/powerpc/configs/
     6.8% arch/arm/configs/
     2.7% arch/powerpc/configs/
     4.2% arch/arm/configs/
     5.6% arch/powerpc/configs/
     8.4% arch/arm/configs/
     5.5% arch/powerpc/configs/
    23.3% arch/arm/configs/
     8.6% arch/powerpc/configs/
     4.0% arch/
     4.4% drivers/usb/musb/
     4.0% drivers/watchdog/
     7.6% drivers/
     3.5% fs/

The trivial fix is to add a sorting pass, fixing it to:

  [torvalds@nehalem linux]$ git diff --dirstat=2 -M v2.6.27-rc4..v2.6.27-rc5
    43.0% arch/arm/configs/
    25.5% arch/powerpc/configs/
     5.3% arch/
     4.4% drivers/usb/musb/
     4.0% drivers/watchdog/
     7.6% drivers/
     3.5% fs/

Spot the difference. In case anybody wonders: it's because of a ton of
renames from {include/asm-blackfin => arch/blackfin/include/asm} that just
totally messed up the file ordering in between arch/arm and arch/powerpc.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agofor-each-ref: Allow a trailing slash in the patterns
Björn Steinbrink [Thu, 28 Aug 2008 02:14:02 +0000 (04:14 +0200)]
for-each-ref: Allow a trailing slash in the patterns

More often than not, I end up using something like refs/remotes/ as the
pattern for for-each-ref, but that doesn't work, because it expects to see
the slash in the ref name right after the matched pattern. So teach it to
accept the slash as the final character in the pattern as well.

Signed-off-by: Björn Steinbrink <B.Steinbrink@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agolist_commands: only call git_exec_path if it is needed
Alex Riesen [Thu, 28 Aug 2008 17:19:42 +0000 (19:19 +0200)]
list_commands: only call git_exec_path if it is needed

Even if it always needed

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoAllow git help work without PATH set
Alex Riesen [Thu, 28 Aug 2008 17:19:07 +0000 (19:19 +0200)]
Allow git help work without PATH set

Just because we can

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMake main_cmds and other_cmds local to builtin-help.c
Alex Riesen [Thu, 28 Aug 2008 17:17:46 +0000 (19:17 +0200)]
Make main_cmds and other_cmds local to builtin-help.c

These are not used anywhere else.

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoRemove useless memset of static command name lists in builtin-merge.c
Alex Riesen [Thu, 28 Aug 2008 17:17:13 +0000 (19:17 +0200)]
Remove useless memset of static command name lists in builtin-merge.c

The statics are always initialized with 0

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoRemove calculation of the longest command name from where it is not used
Alex Riesen [Thu, 28 Aug 2008 17:15:33 +0000 (19:15 +0200)]
Remove calculation of the longest command name from where it is not used

Just calculate it where it is needed - it is cheap and trivial,
as all the lengths are already there (stored when creating the
command lists).

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agodir.c: Avoid c99 array initialization
Brandon Casey [Thu, 28 Aug 2008 22:47:22 +0000 (17:47 -0500)]
dir.c: Avoid c99 array initialization

The following syntax:

        char foo[] = {
                [0] = 1,
                [7] = 2,
                [15] = 3
        };

is a c99 construct which some compilers do not support even though they
support other c99 constructs. This construct can be avoided by folding
these 'special' test cases into the sane_ctype array and making use of
the related infrastructure.

Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agobash-completion: Add all submodule subcommands to the completion list
Matthias Kestenholz [Thu, 28 Aug 2008 08:57:55 +0000 (10:57 +0200)]
bash-completion: Add all submodule subcommands to the completion list

Signed-off-by: Matthias Kestenholz <mk@spinlock.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'mv/merge-custom'
Junio C Hamano [Thu, 28 Aug 2008 00:28:31 +0000 (17:28 -0700)]
Merge branch 'mv/merge-custom'

* mv/merge-custom:
  t7606: fix custom merge test
  Fix "git-merge -s bogo" help text
  Update .gitignore to ignore git-help
  Builtin git-help.
  builtin-help: always load_command_list() in cmd_help()
  Add a second testcase for handling invalid strategies in git-merge
  Add a new test for using a custom merge strategy
  builtin-merge: allow using a custom strategy
  builtin-help: make some internal functions available to other builtins

Conflicts:
help.c

16 years agoMerge branch 'ml/submodule'
Junio C Hamano [Thu, 28 Aug 2008 00:16:30 +0000 (17:16 -0700)]
Merge branch 'ml/submodule'

* ml/submodule:
  git-submodule.sh - Remove trailing / from URL if found
  git-submodule.sh - Remove trailing / from URL if found

16 years agoMerge branch 'am/cherry-pick-rerere'
Junio C Hamano [Wed, 27 Aug 2008 23:40:09 +0000 (16:40 -0700)]
Merge branch 'am/cherry-pick-rerere'

* am/cherry-pick-rerere:
  Make cherry-pick use rerere for conflict resolution.

16 years agoMerge branch 'jc/add-addremove'
Junio C Hamano [Wed, 27 Aug 2008 23:39:57 +0000 (16:39 -0700)]
Merge branch 'jc/add-addremove'

* jc/add-addremove:
  builtin-add.c: optimize -A option and "git add ."
  builtin-add.c: restructure the code for maintainability

16 years agoMerge branch 'np/verify-pack'
Junio C Hamano [Wed, 27 Aug 2008 23:39:46 +0000 (16:39 -0700)]
Merge branch 'np/verify-pack'

* np/verify-pack:
  discard revindex data when pack list changes

16 years agoMerge branch 'da/submodule-sync'
Junio C Hamano [Wed, 27 Aug 2008 23:39:19 +0000 (16:39 -0700)]
Merge branch 'da/submodule-sync'

* da/submodule-sync:
  git-submodule: add "sync" command

16 years agoMerge branch 'maint'
Junio C Hamano [Wed, 27 Aug 2008 23:23:54 +0000 (16:23 -0700)]
Merge branch 'maint'

* maint:
  ctype.c: protect tiny C preprocessor constants
  index-pack: be careful after fixing up the header/footer

16 years agoctype.c: protect tiny C preprocessor constants
Junio C Hamano [Wed, 27 Aug 2008 23:14:22 +0000 (16:14 -0700)]
ctype.c: protect tiny C preprocessor constants

Some platforms contaminate the preprocessor token namespace with their own
definition of SS without being asked.  Avoid getting hit by redefinition
warning messages by explicitly undef SS, AA and DD shorthand we use in this
table definition.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoindex-pack: be careful after fixing up the header/footer
Linus Torvalds [Wed, 27 Aug 2008 19:48:00 +0000 (12:48 -0700)]
index-pack: be careful after fixing up the header/footer

The index-pack command, when processing a thin pack, fixed up the pack
after-the-fact.  It forgets to fsync the result, because it only did that
in one path rather in all cases of fixup.

This moves the fsync_or_die() to the fix-up routine itself, rather than
doing it in one of the callers, so that all cases are covered.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMake it possible to abort the submission of a change to Perforce
Simon Hausmann [Wed, 27 Aug 2008 07:30:29 +0000 (09:30 +0200)]
Make it possible to abort the submission of a change to Perforce

Currently it is not possible to skip the submission of a change to Perforce
when running git-p4 submit. This patch compares the modification time before
and after the submit editor invokation and offers a prompt for skipping if
the submit template file was not saved.

Signed-off-by: Simon Hausmann <simon@lst.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'af/maint-install-no-handlink'
Junio C Hamano [Wed, 27 Aug 2008 00:08:25 +0000 (17:08 -0700)]
Merge branch 'af/maint-install-no-handlink'

* af/maint-install-no-handlink:
  Makefile: always provide a fallback when hardlinks fail

16 years agoMerge branch 'jc/no-slim-shell'
Junio C Hamano [Wed, 27 Aug 2008 00:08:22 +0000 (17:08 -0700)]
Merge branch 'jc/no-slim-shell'

* jc/no-slim-shell:
  Revert "Build-in "git-shell""

16 years agoMerge branch 'maint'
Junio C Hamano [Wed, 27 Aug 2008 00:08:19 +0000 (17:08 -0700)]
Merge branch 'maint'

* maint:
  index-pack: setup git repository
  Suppress some bash redirection error messages
  Fix a warning (on cygwin) to allow -Werror
  Fix "git log -i --grep"

16 years agoformat-patch: use default diff format even with patch options
Jeff King [Mon, 25 Aug 2008 02:10:29 +0000 (22:10 -0400)]
format-patch: use default diff format even with patch options

Previously, running "git format-patch -U5" would cause the
low-level diff machinery to change the diff output format
from "not specified" to "patch". This meant that
format-patch thought we explicitly specified a diff output
format, and would not use the default format. The resulting
message lacked both the diffstat and the summary, as well as
the separating "---".

Now format-patch explicitly checks for this condition and
uses the default. That means that "git format-patch -p" will
now have the "-p" ignored.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoindex-pack: setup git repository
Nguyễn Thái Ngọc Duy [Tue, 26 Aug 2008 14:32:42 +0000 (21:32 +0700)]
index-pack: setup git repository

"git index-pack" is an independent command and does not setup git
repository while still need pack.indexversion. It may miss the
info if it is in a subdirectory of the repository.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoSuppress some bash redirection error messages
Ramsay Jones [Tue, 26 Aug 2008 17:52:57 +0000 (18:52 +0100)]
Suppress some bash redirection error messages

In particular, when testing if the filesystem allows tabs in
filenames, bash issues an error something like:

./t4016-diff-quote.sh: pathname with HT: No such file or directory

which is caused by the failure of the (stdout) redirection,
since the file cannot be created. In order to suppress the
error message, you must redirect stderr to /dev/null, *before*
the stdout redirection on the command-line.

Also, remove a redundant filesystem check from the begining of
the t3902-quoted.sh test and standardise the "test skipped"
message to 'say' on exit.

Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix a warning (on cygwin) to allow -Werror
Ramsay Jones [Tue, 26 Aug 2008 17:50:37 +0000 (18:50 +0100)]
Fix a warning (on cygwin) to allow -Werror

Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMakefile: always provide a fallback when hardlinks fail
Andreas Färber [Mon, 25 Aug 2008 15:33:03 +0000 (17:33 +0200)]
Makefile: always provide a fallback when hardlinks fail

We make hardlinks from "git" to "git-<cmd>" built-ins and have been
careful to avoid cross-device links when linking "git-<cmd>" to
gitexecdir.

However, we were not prepared to deal with a build directory that is
incapable of making hard links within itself. This patch corrects it.

Instead of temporarily linking "git" to gitexecdir, directly link "git-
add", falling back to "cp". Try hardlinking that as "git-<cmd>", falling
back to symlinks or "cp" on error.

While at it, avoid 100+ error messages from hardlink failures when we are
going to fall back to symlinks or "cp" by redirecting the standard error
to /dev/null.

Signed-off-by: Andreas Färber <andreas.faerber@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agodaemon.c: minor style fixup
Junio C Hamano [Sun, 24 Aug 2008 20:27:10 +0000 (13:27 -0700)]
daemon.c: minor style fixup

* "else" on the same line as "}" that closes corresponding "if (...) {";

 * multi-line comments begin with "/*\n";

 * sizeof, even it is not a function, is written as "sizeof(...)";

 * no need to check x?alloc() return value -- it would have died;

 * "if (...) { ... }" that covers the whole function body can be dedented
   by returning from the function early with "if (!...) return;";

 * SP on each side of an operator, i.e. "a > 0", not "a>0";

Also removes stale comment describing how remove_child() used to do its
thing.

Signed-off-by: Junio C Hamano <gitster@pobox.com>:
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-submodule: add "sync" command
David Aguilar [Sun, 24 Aug 2008 19:43:37 +0000 (12:43 -0700)]
git-submodule: add "sync" command

When a submodule's URL changes upstream, existing submodules
will be out of sync since their remote."$origin".url will still
be set to the old value.

This adds a "git submodule sync" command that reads submodules'
URLs from .gitmodules and updates them accordingly.

Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoRevert "Build-in "git-shell""
Junio C Hamano [Tue, 26 Aug 2008 05:39:17 +0000 (22:39 -0700)]
Revert "Build-in "git-shell""

This reverts commit daa0cc9a92c9c2c714aa5f7da6d0ff65b93e0698.
It was a stupid idea to do this; when run as a log-in shell,
it is spawned with argv[0] set to "-git-shell", so the usual
name-based dispatch would not work to begin with.

16 years agoFix "git log -i --grep"
Jeff King [Mon, 25 Aug 2008 06:15:05 +0000 (02:15 -0400)]
Fix "git log -i --grep"

This has been broken in v1.6.0 due to the reorganization of
the revision option parsing code. The "-i" is completely
ignored, but works fine in "git log --grep -i".

What happens is that the code for "-i" looks for
revs->grep_filter; if it is NULL, we do nothing, since there
are no grep filters. But that is obviously not correct,
since we want it to influence the later --grep option. Doing
it the other way around works, since "-i" just impacts the
existing grep_filter option.

Instead, we now always initialize the grep_filter member and
just fill in options and patterns as we get them. This means
that we can no longer check grep_filter for NULL, but
instead must check the pattern list to see if we have any
actual patterns.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoInstall git-shell in bindir, too
Tommi Virtanen [Sun, 24 Aug 2008 20:23:25 +0000 (23:23 +0300)]
Install git-shell in bindir, too

/etc/passwd shell field must be something execable, you can't enter
"/usr/bin/git shell" there. git-shell must be present as a separate
executable, or it is useless.

Signed-off-by: Tommi Virtanen <tv@eagain.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'jc/no-slim-shell'
Junio C Hamano [Sun, 24 Aug 2008 23:14:22 +0000 (16:14 -0700)]
Merge branch 'jc/no-slim-shell'

* jc/no-slim-shell:
  Build-in "git-shell"
  shell: do not play duplicated definition games to shrink the executable

16 years agoMerge branch 'maint' to sync with 1.6.0.1
Junio C Hamano [Sun, 24 Aug 2008 21:48:36 +0000 (14:48 -0700)]
Merge branch 'maint' to sync with 1.6.0.1

16 years agoGIT 1.6.0.1 v1.6.0.1
Junio C Hamano [Sun, 24 Aug 2008 21:47:24 +0000 (14:47 -0700)]
GIT 1.6.0.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'ag/maint-combine-diff-fix' into maint
Junio C Hamano [Sun, 24 Aug 2008 21:32:18 +0000 (14:32 -0700)]
Merge branch 'ag/maint-combine-diff-fix' into maint

* ag/maint-combine-diff-fix:
  Respect core.autocrlf in combined diff

16 years agoMerge branch 'mv/maint-merge-fix' into maint
Junio C Hamano [Sun, 24 Aug 2008 21:29:37 +0000 (14:29 -0700)]
Merge branch 'mv/maint-merge-fix' into maint

* mv/maint-merge-fix:
  merge: fix numerus bugs around "trivial merge" area

16 years agogit-submodule - Use "get_default_remote" from git-parse-remote
Mark Levedahl [Sun, 24 Aug 2008 18:46:10 +0000 (14:46 -0400)]
git-submodule - Use "get_default_remote" from git-parse-remote

Resolve_relative_url was using its own code for this function, but
this is duplication with the best result that this continues to work.
Replace with the common function provided by git-parse-remote.

Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: clarify pager configuration
Jonathan Nieder [Sun, 24 Aug 2008 05:28:32 +0000 (00:28 -0500)]
Documentation: clarify pager configuration

The unwary user may not know how to disable the -FRSX options.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: clarify pager.<cmd> configuration
Jonathan Nieder [Sun, 24 Aug 2008 05:38:06 +0000 (00:38 -0500)]
Documentation: clarify pager.<cmd> configuration

It was not obvious from the text that pager.<cmd> is a boolean
setting.

While we're changing the description, make some other
improvements: lest we forget and fret, clarify that -p and
pager.<cmd> do not kick in when stdout is not a tty; point to
related core.pager and GIT_PAGER settings; use renamed --paginate
option.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoClean up the git-p4 documentation
Simon Hausmann [Sun, 24 Aug 2008 14:12:23 +0000 (16:12 +0200)]
Clean up the git-p4 documentation

This patch massages the documentation a bit for improved readability and cleans
it up from outdated options/commands.

Signed-off-by: Simon Hausmann <simon@lst.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoRespect core.autocrlf in combined diff
Alexander Gavrilov [Sat, 23 Aug 2008 19:21:21 +0000 (23:21 +0400)]
Respect core.autocrlf in combined diff

Fix git-diff to make it produce useful 3-way diffs for merge conflicts in
repositories with autocrlf enabled. Otherwise it always reports that the
whole file was changed, because it uses the contents from the working tree
without necessary conversion.

Signed-off-by: Alexander Gavrilov <angavrilov@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMakefile: enable SNPRINTF_RETURNS_BOGUS for HP-UX
Miklos Vajna [Sat, 23 Aug 2008 22:07:55 +0000 (00:07 +0200)]
Makefile: enable SNPRINTF_RETURNS_BOGUS for HP-UX

In 81cc66a, customization has been added to Makefile for supporting
HP-UX, but git commit is still problematic. This should fix the issue.

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Acked-by: Robert Schiele <rschiele@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agot7606: fix custom merge test
Junio C Hamano [Sun, 24 Aug 2008 02:23:22 +0000 (19:23 -0700)]
t7606: fix custom merge test

Custom merge strategy does not even kick in when the merge is truly
trivial.  The test depended on the behaviour in the git-merge rewritten in
C that broke the trivial merge completely.

Make the test to work on a non-trivial merge to make sure the strategy
kicks in.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'maint'
Junio C Hamano [Sun, 24 Aug 2008 01:28:37 +0000 (18:28 -0700)]
Merge branch 'maint'

* maint:
  unpack_trees(): protect the handcrafted in-core index from read_cache()
  git-p4: Fix one-liner in p4_write_pipe function.
  Completion: add missing '=' for 'diff --diff-filter'
  Fix 'git help help'

16 years agomerge: fix numerus bugs around "trivial merge" area
Junio C Hamano [Sat, 23 Aug 2008 19:56:57 +0000 (12:56 -0700)]
merge: fix numerus bugs around "trivial merge" area

The "trivial merge" codepath wants to optimize itself by making an
internal call to the read-tree machinery, but it does not read the index
before doing so, and the codepath is never exercised.  Incidentally, this
failure to read the index upfront means that the safety to refuse doing
anything when the index is unmerged does not kick in, either.

These two problem are fixed by using read_cache_unmerged() that does read
the index before checking if it is unmerged at the beginning of
cmd_merge().

The primary logic of the merge, however, assumes that the process never
reads the index in-core, and the call to write_cache_as_tree() it makes
from write_tree_trivial() will always read from the on-disk index that is
prepared the strategy back-ends.  This assumption is now broken by the
above fix.  To fix this issue, we now call discard_cache() before calling
write_tree_trivial() when it wants to write the on-disk index as a tree.

When multiple strategies are tried, their results are evaluated by reading
the resulting index and inspecting it.  The codepath needs to make a call
to read_cache() for each successful strategy, and for that to work, they
need to discard_cache() the one read by the previous round.

Also the "trivial merge" forgot that the current commit is one of the
parents of the resulting commit.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agounpack_trees(): protect the handcrafted in-core index from read_cache()
Junio C Hamano [Sat, 23 Aug 2008 19:57:30 +0000 (12:57 -0700)]
unpack_trees(): protect the handcrafted in-core index from read_cache()

unpack_trees() rebuilds the in-core index from scratch by allocating a new
structure and finishing it off by copying the built one to the final
index.

The resulting in-core index is Ok for most use, but read_cache() does not
recognize it as such.  The function is meant to be no-op if you already
have loaded the index, until you call discard_cache().

This change the way read_cache() detects an already initialized in-core
index, by introducing an extra bit, and marks the handcrafted in-core
index as initialized, to avoid this problem.

A better fix in the longer term would be to change the read_cache() API so
that it will always discard and re-read from the on-disk index to avoid
confusion.  But there are higher level API that have relied on the current
semantics, and they and their users all need to get converted, which is
outside the scope of 'maint' track.

An example of such a higher level API is write_cache_as_tree(), which is
used by git-write-tree as well as later Porcelains like git-merge, revert
and cherry-pick.  In the longer term, we should remove read_cache() from
there and add one to cmd_write_tree(); other callers expect that the
in-core index they prepared is what gets written as a tree so no other
change is necessary for this particular codepath.

The original version of this patch marked the index by pointing an
otherwise wasted malloc'ed memory with o->result.alloc, but this version
uses Linus's idea to use a new "initialized" bit, which is conceptually
much cleaner.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-p4: Fix one-liner in p4_write_pipe function.
Tor Arvid Lund [Thu, 21 Aug 2008 21:11:40 +0000 (23:11 +0200)]
git-p4: Fix one-liner in p4_write_pipe function.

The function built a p4 command string via the p4_build_cmd function, but
ignored the result.

Signed-off-by: Tor Arvid Lund <torarvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-submodule: replace duplicated code with a module_list function
David Aguilar [Fri, 22 Aug 2008 07:30:50 +0000 (00:30 -0700)]
git-submodule: replace duplicated code with a module_list function

Several call sites in git-submodule.sh used the same idiom for getting
submodule information:

git ls-files --stage -- "$@" | grep '^160000 '

This patch removes this duplication by introducing a module_list function.

Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agodiscard revindex data when pack list changes
Nicolas Pitre [Fri, 22 Aug 2008 19:45:53 +0000 (15:45 -0400)]
discard revindex data when pack list changes

This is needed to fix verify-pack -v with multiple pack arguments.

Also, in theory, revindex data (if any) must be discarded whenever
reprepare_packed_git() is called. In practice this is hard to trigger
though.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-merge documentation: more details about resolving conflicts
Dan Hensgen [Fri, 22 Aug 2008 03:32:00 +0000 (23:32 -0400)]
git-merge documentation: more details about resolving conflicts

Signed-off-by: Dan Hensgen <dan@methodhead.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoExtend "checkout --track" DWIM to support more cases
Alex Riesen [Thu, 21 Aug 2008 17:23:20 +0000 (19:23 +0200)]
Extend "checkout --track" DWIM to support more cases

The code handles additionally "refs/remotes/<something>/name",
"remotes/<something>/name", and "refs/<namespace>/name".

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoRevert "Convert output messages in merge-recursive to past tense."
Jonathan del Strother [Fri, 22 Aug 2008 17:18:44 +0000 (18:18 +0100)]
Revert "Convert output messages in merge-recursive to past tense."

During a conflicting merge, you would typically see:

  Auto-merged foo.txt
  CONFLICT (content): Merge conflict in foo.txt
  Recorded preimage for 'foo.txt'
  Automatic merge failed; fix conflicts and then commit the result.

and left wondering what happened to "foo.txt".  Did it succeed, and then
conflicted, and then what?

This is because historically there was a progress bar displayed before the
auto-merge is mentioned, and it was expected to take long time, before we
can say "Auto-merged foo.txt".  It turns out it was not the case, and the
original wording "Auto-merging foo.txt" we used to have before 89f40be
(Convert output messages in merge-recursive to past tense., 2007-01-14) is
better.

Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoCompletion: add missing '=' for 'diff --diff-filter'
Eric Raible [Fri, 22 Aug 2008 17:25:06 +0000 (10:25 -0700)]
Completion: add missing '=' for 'diff --diff-filter'

Signed-off-by: Eric Raible <raible@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agotemplates/Makefile: install is unnecessary, just use mkdir -p
Junio C Hamano [Fri, 22 Aug 2008 00:31:50 +0000 (19:31 -0500)]
templates/Makefile: install is unnecessary, just use mkdir -p

The native install on some platforms (namely IRIX 6.5) treats non-absolute
paths as being relative to the root directory rather than relative to
the current directory. Work around this by avoiding install in this case
since it is unnecessary, and instead depend on the local umask setting
and use mkdir.

Tested-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoremote.c: add a function for deleting a refspec array and use it (twice)
Brandon Casey [Fri, 22 Aug 2008 00:16:30 +0000 (19:16 -0500)]
remote.c: add a function for deleting a refspec array and use it (twice)

A number of call sites allocate memory for a refspec array, populate
its members with heap memory, and then free only the refspec pointer
while leaking the memory allocated for the member elements. Provide
a function for freeing the elements of a refspec array and the array
itself.

Caution to callers: code paths must be checked to ensure that the
refspec members "src" and "dst" can be passed to free.

Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agofilter-branch: Grok special characters in tag names
Johannes Sixt [Thu, 21 Aug 2008 14:45:11 +0000 (16:45 +0200)]
filter-branch: Grok special characters in tag names

The tag rewriting code used a 'sed' expression to substitute the new tag
name into the corresponding field of the annotated tag object. But this is
problematic if the tag name contains special characters. In particular,
if the tag name contained a slash, then the 'sed' expression had a syntax
error. We now protect against this by using 'printf' to assemble the
tag header.

Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agotest-lib: do not remove trash_directory if called with --debug
Johannes Schindelin [Thu, 21 Aug 2008 07:49:12 +0000 (09:49 +0200)]
test-lib: do not remove trash_directory if called with --debug

Sometimes you want to keep the trash directory, even if all tests
passed.  For example, when extending tests, it comes it quite handy.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoprovide more errors for the "merge into empty head" case
Paolo Bonzini [Thu, 21 Aug 2008 12:14:18 +0000 (14:14 +0200)]
provide more errors for the "merge into empty head" case

A squash merge into an unborn branch could be implemented by building the
index from the merged-from branch, and doing a single commit, but this is
not supported yet.

A non-fast-forward merge into an unborn branch does not make any sense,
because you cannot make a merge commit if you don't have a commit to use
as the parent.

Signed-off-by: Paolo Bonzini <bonzini@gnu.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMake reflog query '@{1219188291}' act as '@{2008.8.19.16:24:51.-0700}'
Shawn O. Pearce [Thu, 21 Aug 2008 15:40:44 +0000 (08:40 -0700)]
Make reflog query '@{1219188291}' act as '@{2008.8.19.16:24:51.-0700}'

As we support seconds-since-epoch in $GIT_COMMITTER_TIME we should
also support it in a reflog @{...} style notation.  We can easily
tell this part from @{nth} style notation by looking to see if the
value is unreasonably large for an @{nth} style notation.

The value 100000000 was chosen as it is already used by date.c to
disambiguate yyyymmdd format from a seconds-since-epoch time value.

A reflog with 100,000,000 record entries is also simply not valid.
Such a reflog would require at least 7.7 GB to store just the old
and new SHA-1 values.  So our randomly chosen upper limit for @{nth}
notation is "big enough".

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>