Code

git.git
15 years agoMerge branch 'jc/maint-1.6.0-split-diff-metainfo' into maint-1.6.0
Junio C Hamano [Fri, 13 Mar 2009 03:01:28 +0000 (20:01 -0700)]
Merge branch 'jc/maint-1.6.0-split-diff-metainfo' into maint-1.6.0

* jc/maint-1.6.0-split-diff-metainfo:
  diff.c: output correct index lines for a split diff

15 years agoadded missing backtick in git-apply.txt
Danijel Tasov [Sat, 28 Feb 2009 20:03:54 +0000 (21:03 +0100)]
added missing backtick in git-apply.txt

Signed-off-by: Danijel Tasov <dt@korn.shell.la>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMake repack less likely to corrupt repository
Junio C Hamano [Tue, 10 Feb 2009 20:16:31 +0000 (12:16 -0800)]
Make repack less likely to corrupt repository

Some platforms refuse to rename a file that is open.  When repacking an
already packed repository without adding any new object, the resulting
pack will contain the same set of objects as an existing pack, and on such
platforms, a newly created packfile cannot replace the existing one.

The logic detected this issue but did not try hard enough to recover from
it.  Especially because the files that needs renaming come in pairs, there
potentially are different failure modes that one can be renamed but the
others cannot.  Asking manual recovery to end users were error prone.

This patch tries to make it more robust by first making sure all the
existing files that need to be renamed have been renamed before
continuing, and attempts to roll back if some failed to rename.

This is based on an initial patch by Robin Rosenberg.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agofast-export: ensure we traverse commits in topological order
Elijah Newren [Wed, 11 Feb 2009 06:03:53 +0000 (23:03 -0700)]
fast-export: ensure we traverse commits in topological order

fast-export will only list as parents those commits which have already
been traversed (making it appear as if merges have been squashed if not
all parents have been traversed).  To avoid this silent squashing of
merge commits, we request commits in topological order.

Signed-off-by: Elijah Newren <newren@gmail.com>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoClear the delta base cache if a pack is rebuilt
Shawn O. Pearce [Wed, 11 Feb 2009 18:15:30 +0000 (10:15 -0800)]
Clear the delta base cache if a pack is rebuilt

There is some risk that re-opening a regenerated pack file with
different offsets could leave stale entries within the delta base
cache that could be matched up against other objects using the same
"struct packed_git*" and pack offset.

Throwing away the entire delta base cache in this case is safer,
as we don't have to worry about a recycled "struct packed_git*"
matching to the wrong base object, resulting in delta apply
errors while unpacking an object.

Suggested-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoClear the delta base cache during fast-import checkpoint
Shawn O. Pearce [Tue, 10 Feb 2009 21:36:12 +0000 (13:36 -0800)]
Clear the delta base cache during fast-import checkpoint

Otherwise we may reuse the same memory address for a totally
different "struct packed_git", and a previously cached object from
the prior occupant might be returned when trying to unpack an object
from the new pack.

Found-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoFixed broken git help -w when installing from RPM
David J. Mellor [Thu, 5 Feb 2009 04:14:29 +0000 (20:14 -0800)]
Fixed broken git help -w when installing from RPM

After the git-core package was renamed to git, git help -w was still looking
for files in /usr/share/doc/git-core-$VERSION instead of
/usr/share/doc/git-$VERSION.

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agobuiltin-mv.c: check for unversionned files before looking at the destination.
Matthieu Moy [Wed, 4 Feb 2009 09:32:08 +0000 (10:32 +0100)]
builtin-mv.c: check for unversionned files before looking at the destination.

The previous code was failing in the case where one moves an
unversionned file to an existing destination, with mv -f: the
"existing destination" was checked first, and the error was cancelled
by the force flag.

We now check the unrecoverable error first, which fixes the bug.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAdd a testcase for "git mv -f" on untracked files.
Matthieu Moy [Wed, 4 Feb 2009 09:32:07 +0000 (10:32 +0100)]
Add a testcase for "git mv -f" on untracked files.

This currently fails with:
git: builtin-mv.c:217: cmd_mv: Assertion `pos >= 0' failed.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMissing && in t/t7001.sh.
Matthieu Moy [Wed, 4 Feb 2009 09:32:06 +0000 (10:32 +0100)]
Missing && in t/t7001.sh.

Without this, the exit status is only the one of the last line.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoUser-manual: "git stash <comment>" form is long gone
William Pursell [Tue, 3 Feb 2009 22:41:14 +0000 (22:41 +0000)]
User-manual: "git stash <comment>" form is long gone

These days you must explicitly say "git stash save <comment>".

Signed-off-by: William Pursell <bill.pursell@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoadd test-dump-cache-tree in Makefile
Guanqun Lu [Wed, 4 Feb 2009 21:00:41 +0000 (05:00 +0800)]
add test-dump-cache-tree in Makefile

5c5ba73 (Makefile: Use generic rule to build test programs,
2007-05-31) tried to use generic rule to build test programs, but it
misses the file 'dump-cache-tree.c', since its name is not prefixed by
'test-'.  This commit solves this little problem by renaming this file
instead of carrying out an explicit rule in Makefile.

Signed-off-by: Guanqun Lu <guanqun.lu@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agofix typo in Documentation
Guanqun Lu [Wed, 4 Feb 2009 21:00:40 +0000 (05:00 +0800)]
fix typo in Documentation

Signed-off-by: Guanqun Lu <guanqun.lu@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoapply: fix access to an uninitialized mode variable, found by valgrind
Johannes Schindelin [Wed, 4 Feb 2009 01:50:15 +0000 (02:50 +0100)]
apply: fix access to an uninitialized mode variable, found by valgrind

When 'tpatch' was initialized successfully, st_mode was already taken
from the previous diff.  We should not try to override it with data
from an lstat() that was never called.

This is a companion patch to 7a07841(git-apply: handle a patch that
touches the same path more than once better).

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogrep: pass -I (ignore binary) down to external grep
Junio C Hamano [Mon, 2 Feb 2009 18:58:20 +0000 (10:58 -0800)]
grep: pass -I (ignore binary) down to external grep

We forgot to pass this option to the external grep process.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agomerge: fix out-of-bounds memory access
René Scharfe [Sat, 31 Jan 2009 14:39:10 +0000 (15:39 +0100)]
merge: fix out-of-bounds memory access

The parameter n of unpack_callback() can have a value of up to
MAX_UNPACK_TREES.  The check at the top of unpack_trees() (its only
(indirect) caller) makes sure it cannot exceed this limit.

unpack_callback() passes it and the array src to unpack_nondirectories(),
which has this loop:

for (i = 0; i < n; i++) {
/* ... */
src[i + o->merge] = o->df_conflict_entry;

o->merge can be 0 or 1, so unpack_nondirectories() potentially accesses
the array src at index MAX_UNPACK_TREES.  This patch makes it big enough.

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: René Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoavoid 31-bit truncation in write_loose_object
Jeff King [Thu, 29 Jan 2009 05:56:34 +0000 (00:56 -0500)]
avoid 31-bit truncation in write_loose_object

The size of the content we are adding may be larger than
2.1G (i.e., "git add gigantic-file"). Most of the code-path
to do so uses size_t or unsigned long to record the size,
but write_loose_object uses a signed int.

On platforms where "int" is 32-bits (which includes x86_64
Linux platforms), we end up passing malloc a negative size.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agotest-path-utils: Fix off by one, found by valgrind
Johannes Schindelin [Tue, 27 Jan 2009 23:07:36 +0000 (00:07 +0100)]
test-path-utils: Fix off by one, found by valgrind

When normalizing an absolute path, we might have to add a slash _and_ a
NUL to the buffer, so the buffer was one too small.

Let's just future proof the code and alloc PATH_MAX + 1 bytes.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoget_sha1_basic(): fix invalid memory access, found by valgrind
Johannes Schindelin [Tue, 27 Jan 2009 23:07:46 +0000 (00:07 +0100)]
get_sha1_basic(): fix invalid memory access, found by valgrind

When get_sha1_basic() is passed a buffer of len 0, it should not
check if buf[len-1] is a curly bracket.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agodiff.c: output correct index lines for a split diff
Junio C Hamano [Mon, 26 Jan 2009 08:08:24 +0000 (00:08 -0800)]
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>
15 years agoFix Documentation for git-describe
Boyd Stephen Smith Jr [Thu, 22 Jan 2009 18:26:25 +0000 (12:26 -0600)]
Fix Documentation for git-describe

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>
15 years agobuiltin-fsck: fix off by one head count
Christian Couder [Sun, 18 Jan 2009 03:46:09 +0000 (04:46 +0100)]
builtin-fsck: fix off by one head count

According to the man page, if "git fsck" is passed one or more heads, it
should verify connectivity and validity of only objects reachable from the
heads it is passed.

However, since 5ac0a20 (Make builtin-fsck.c use parse_options.,
2007-10-15) the command behaved as if no heads were passed, when given
only one argument.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: let asciidoc align related options
Markus Heidelberg [Fri, 16 Jan 2009 21:42:33 +0000 (22:42 +0100)]
Documentation: let asciidoc align related options

Command line options can share the same paragraph of description, if
they are related or synonymous. In these cases they should be written
among each other, so that asciidoc can format them itself.

Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogithooks.txt: add missing word
Stephan Beyer [Fri, 16 Jan 2009 20:36:06 +0000 (21:36 +0100)]
githooks.txt: add missing word

Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agobuiltin-commit.c: do not remove COMMIT_EDITMSG
Stephan Beyer [Fri, 16 Jan 2009 19:40:05 +0000 (20:40 +0100)]
builtin-commit.c: do not remove COMMIT_EDITMSG

git-commit tries to remove the file ./COMMIT_EDITMSG instead of
$GIT_DIR/COMMIT_EDITMSG after commit preparation (e.g. running
hooks, launching editor).
This behavior exists since f5bbc3225c4b07 "Port git commit to C".

Some test cases (e.g. t/t7502-commit.sh) rely on the existence of
$GIT_DIR/COMMIT_EDITMSG after committing and, I guess, many people
are used to it.  So it is best not to remove it.

This patch just removes the removal of COMMIT_EDITMSG.

Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot3404: Add test case for auto-amending only edited commits after "edit"
Stephan Beyer [Thu, 15 Jan 2009 12:56:16 +0000 (13:56 +0100)]
t3404: Add test case for auto-amending only edited commits after "edit"

Add a test case for the bugfix introduced by commit c14c3c82d
"git-rebase--interactive: auto amend only edited commit".

Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot3404: Add test case for aborted --continue after "edit"
Stephan Beyer [Thu, 15 Jan 2009 12:56:15 +0000 (13:56 +0100)]
t3404: Add test case for aborted --continue after "edit"

Add a test case for the bugfix introduced by commit 8beb1f33d
"git-rebase-interactive: do not squash commits on abort".

Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot3501: check that commits are actually done
Stephan Beyer [Thu, 15 Jan 2009 13:03:17 +0000 (14:03 +0100)]
t3501: check that commits are actually done

The basic idea of t3501 is to check whether revert
and cherry-pick works on renamed files.
But as there is no pure cherry-pick/revert test, it is
good to also check if commits are actually done in that
scenario.

Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agofix handling of multiple untracked files for git mv -k
Michael J Gruber [Wed, 14 Jan 2009 17:03:22 +0000 (18:03 +0100)]
fix handling of multiple untracked files for git mv -k

The "-k" option to "git mv" should allow specifying multiple untracked
files. Currently, multiple untracked files raise an assertion if they
appear consecutively as arguments. Fix this by decrementing the loop
index after removing one entry from the array of arguments.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoadd test cases for "git mv -k"
Michael J Gruber [Wed, 14 Jan 2009 17:03:21 +0000 (18:03 +0100)]
add test cases for "git mv -k"

Add test cases for ignoring nonexisting and untracked files using the -k
option to "git mv". There is one known breakage related to multiple
untracked files specfied as consecutive arguments.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agofast-import: Cleanup mode setting.
Felipe Contreras [Wed, 14 Jan 2009 01:37:07 +0000 (03:37 +0200)]
fast-import: Cleanup mode setting.

"S_IFREG | mode" makes only sense for 0644 and 0755.

Even though doing (S_IFREG | mode) may not hurt when mode is any other
supported value, that is only true because S_IFREG mode bit happens to
be already on for S_IFLNK or S_IFGITLINK.

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGit.pm: call Error::Simple() properly
Jay Soffian [Tue, 13 Jan 2009 22:41:35 +0000 (17:41 -0500)]
Git.pm: call Error::Simple() properly

The error message to Error::Simple() must be passed as a single argument.

Signed-off-by: Jay Soffian <jaysoffian@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAvoid spurious error messages on error mistakes.
Pierre Habouzit [Mon, 12 Jan 2009 23:09:36 +0000 (00:09 +0100)]
Avoid spurious error messages on error mistakes.

Prior to that, if the user chose "squash" as a first action, the stderr
looked like:

    grep: /home/madcoder/dev/scm/git/.git/rebase-merge/done: No such file or directory
    Cannot 'squash' without a previous commit

Now the first line is gone.

Signed-off-by: Pierre Habouzit <madcoder@debian.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agocontrib/examples/README: give an explanation of the status of these files
jidanni@jidanni.org [Tue, 13 Jan 2009 01:19:42 +0000 (09:19 +0800)]
contrib/examples/README: give an explanation of the status of these files

We attempt to give an explanation of the status of the files in this
directory.

Signed-off-by: jidanni <jidanni@jidanni.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.5.6' into maint-1.6.0
Junio C Hamano [Wed, 7 Jan 2009 06:12:30 +0000 (22:12 -0800)]
Merge branch 'maint-1.5.6' into maint-1.6.0

* maint-1.5.6:
  README: tutorial.txt is now called gittutorial.txt

15 years agoREADME: tutorial.txt is now called gittutorial.txt
Joey Hess [Wed, 7 Jan 2009 04:23:37 +0000 (23:23 -0500)]
README: tutorial.txt is now called gittutorial.txt

Signed-off-by: Joey Hess <joey@gnu.kitenet.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agodoc/git-fsck: change the way for getting heads' SHA1s
Markus Heidelberg [Sun, 21 Dec 2008 16:30:22 +0000 (17:30 +0100)]
doc/git-fsck: change the way for getting heads' SHA1s

The straightforward way with using 'cat .git/refs/heads/*' doesn't work
with packed refs as well as branches of the form topic/topic1. So let's
use git-for-each-ref for getting the heads' SHA1s in this example.

Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.6.0.6 v1.6.0.6
Junio C Hamano [Sat, 20 Dec 2008 03:27:06 +0000 (19:27 -0800)]
GIT 1.6.0.6

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agofast-import: make tagger information optional
Junio C Hamano [Fri, 19 Dec 2008 22:41:21 +0000 (14:41 -0800)]
fast-import: make tagger information optional

Even though newer Porcelain tools always record the tagger information
when creating new tags, export/import pair should be able to faithfully
reproduce ancient tag objects that lack tagger information.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
15 years agofast-export: deal with tag objects that do not have a tagger
Johannes Schindelin [Sat, 20 Dec 2008 00:00:27 +0000 (01:00 +0100)]
fast-export: deal with tag objects that do not have a tagger

When no tagger was found (old Git produced tags like this),
no "tagger" line is printed (but this is incompatible with the current
git fast-import).

Alternatively, you can pass the option --fake-missing-tagger, forcing
fast-export to fake a tagger

Unspecified Tagger <no-tagger>

with a tag date of the beginning of (Unix) time in the case of a missing
tagger, so that fast-import is still able to import the result.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoSubmittingPatches: mention the usage of real name in Signed-off-by: lines
Miklos Vajna [Sat, 20 Dec 2008 00:52:17 +0000 (01:52 +0100)]
SubmittingPatches: mention the usage of real name in Signed-off-by: lines

Especially with something that is supposed to hopefully have some legal
value down the line if somebody starts making noises, it really would be
nice to have a real person to associate things with. Suggest this in the
SubmittingPatches document.

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-mergetool: properly handle "git mergetool -- filename"
David Aguilar [Sat, 20 Dec 2008 01:01:01 +0000 (17:01 -0800)]
git-mergetool: properly handle "git mergetool -- filename"

Like many git commands, git-mergetool allows "--" to signal
the end of option processing.  This adds a missing "shift"
statement so that this is correctly handled.

Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-send-email: handle email address with quoted comma
Wu Fengguang [Fri, 19 Dec 2008 08:10:10 +0000 (16:10 +0800)]
git-send-email: handle email address with quoted comma

Correctly handle email addresses containing quoted commas, e.g.

    "Zhu, Yi" <yi.zhu@intel.com>, "Li, Shaohua" <shaohua.li@intel.com>

The commas inside the double quotes are not separators.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: fix typos, grammar, asciidoc syntax
Markus Heidelberg [Fri, 19 Dec 2008 12:14:18 +0000 (13:14 +0100)]
Documentation: fix typos, grammar, asciidoc syntax

Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.5.6.6 v1.5.6.6
Junio C Hamano [Wed, 17 Dec 2008 06:13:36 +0000 (22:13 -0800)]
GIT 1.5.6.6

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.5.5.6 v1.5.5.6
Junio C Hamano [Wed, 17 Dec 2008 06:08:22 +0000 (22:08 -0800)]
GIT 1.5.5.6

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.5.4.7 v1.5.4.7
Junio C Hamano [Wed, 17 Dec 2008 06:03:29 +0000 (22:03 -0800)]
GIT 1.5.4.7

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogitweb: do not run "git diff" that is Porcelain
Junio C Hamano [Wed, 17 Dec 2008 03:42:02 +0000 (19:42 -0800)]
gitweb: do not run "git diff" that is Porcelain

Jakub says that legacy-style URI to view two blob differences are never
generated since 1.4.3.  This codepath runs "git diff" Porcelain from the
gitweb, which is a no-no.  It can trigger diff.external command that is
specified in the configuration file of the repository being viewed.

This patch applies to v1.5.4 and later.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agofast-import: close pack before unlinking it
Johannes Schindelin [Mon, 15 Dec 2008 21:11:40 +0000 (22:11 +0100)]
fast-import: close pack before unlinking it

This is sort of a companion patch to 4723ee9(Close files opened by
lock_file() before unlinking.): on Windows, you cannot delete what
is still open.

This makes test 9300-fast-import pass on Windows for me; quite a few
fast-imports leave temporary packs until the test "blank lines not
necessary after other commands" actually tests for the number of files
in .git/objects/pack/, which has a few temporary packs now.

I guess that 8b4eb6b(Do not perform cross-directory renames when
creating packs) was "responsible" for the breakage.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopager: do not dup2 stderr if it is already redirected
Junio C Hamano [Mon, 15 Dec 2008 08:33:34 +0000 (00:33 -0800)]
pager: do not dup2 stderr if it is already redirected

An earlier commit 61b8050 (sending errors to stdout under $PAGER,
2008-02-16) avoided losing the error messages that are sent to the
standard error when $PAGER is in effect by dup2'ing fd 2 to the pager.
his way, showing a tag object that points to a bad object:

    $ git show tag-foo

would give the error message to the pager.  However, it was not quite
right if the user did:

    $ git show 2>error.log tag-foo

i.e. use the pager but store the errors in a separate file.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-show: do not segfault when showing a bad tag
Junio C Hamano [Mon, 15 Dec 2008 08:36:56 +0000 (00:36 -0800)]
git-show: do not segfault when showing a bad tag

When a tag points at a bad or nonexistent object, we should diagnose the
breakage and exit.  An earlier commit 4f3dcc2 (Fix 'git show' on signed
tag of signed tag of commit, 2008-07-01) lost this check and made it
segfault instead; not good.

This fixes it.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-config.txt: fix a typo
Jim Meyering [Fri, 12 Dec 2008 09:00:41 +0000 (10:00 +0100)]
git-config.txt: fix a typo

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agofsck: reduce stack footprint
Linus Torvalds [Thu, 11 Dec 2008 03:44:37 +0000 (19:44 -0800)]
fsck: reduce stack footprint

The logic to mark all objects that are reachable from tips of refs were
implemented as a set of recursive functions.  In a repository with a deep
enough history, this can easily eat up all the available stack space.

Restructure the code to require less stackspace by using an object array
to keep track of the objects that still need to be processed.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agomake sure packs to be replaced are closed beforehand
Nicolas Pitre [Tue, 9 Dec 2008 19:26:52 +0000 (14:26 -0500)]
make sure packs to be replaced are closed beforehand

Especially on Windows where an opened file cannot be replaced, make
sure pack-objects always close packs it is about to replace. Even on
non Windows systems, this could save potential bad results if ever
objects were to be read from the new pack file using offset from the old
index.

This should fix t5303 on Windows.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Tested-by: Johannes Sixt <j6t@kdbg.org> (MinGW)
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agowork around Python warnings from AsciiDoc
Junio C Hamano [Mon, 8 Dec 2008 02:38:46 +0000 (18:38 -0800)]
work around Python warnings from AsciiDoc

It appears that a reference to an anchor defined as [[anchor-name]] from
another place using <<anchor-name>> syntax, when the anchor name contains
a string "-with-" in its name, triggers these warnings from Python
interpreter.

  asciidoc -b docbook -d book user-manual.txt
  <string>:1: Warning: 'with' will become a reserved keyword in Python 2.6
  <string>:1: Warning: 'with' will become a reserved keyword in Python 2.6
  <string>:1: Warning: 'with' will become a reserved keyword in Python 2.6
  <string>:1: Warning: 'with' will become a reserved keyword in Python 2.6

There currently is no reference to "Finding comments with given content",
but for consistency and for futureproofing, the anchor is also updated as
the other ones that are actually used and trigger these warnings.

Signed-off-by: Junio C Hamano <junio@pobox.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-svn: Make following parents atomic
Deskin Miller [Mon, 8 Dec 2008 13:31:31 +0000 (08:31 -0500)]
git-svn: Make following parents atomic

find_parent_branch generates branch@rev type branches when one has to
look back through SVN history to properly get the history for a branch
copied from somewhere not already being tracked by git-svn.  If in the
process of fetching this history, git-svn is interrupted, then when one
fetches again, it will use whatever was last fetched as the parent
commit and fail to fetch any more history which it didn't get to before
being terminated.  This is especially troubling in that different
git-svn copies of the same SVN repository can end up with different
commit sha1s, incorrectly showing the history as divergent and
precluding easy collaboration using git push and fetch.

To fix this, when we initialise the Git::SVN object $gs to search for
and perhaps fetch history, we check if there are any commits in SVN in
the range between the current revision $gs is at, and the top revision
for which we were asked to fill history.  If there are commits we're
missing in that range, we continue the fetch from the current revision
to the top, properly getting all history before using it as the parent
for the branch we're trying to create.

Signed-off-by: Deskin Miller <deskinm@umich.edu>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.6.0.5 v1.6.0.5
Junio C Hamano [Sun, 7 Dec 2008 11:03:16 +0000 (03:03 -0800)]
GIT 1.6.0.5

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years ago"git diff <tree>{3,}": do not reverse order of arguments
Matt McCutchen [Sat, 11 Oct 2008 01:56:15 +0000 (21:56 -0400)]
"git diff <tree>{3,}": do not reverse order of arguments

According to the message of commit 0fe7c1de16f71312e6adac4b85bddf0d62a47168,
"git diff" with three or more trees expects the merged tree first followed by
the parents, in order.  However, this command reversed the order of its
arguments, resulting in confusing diffs.  A comment /* Again, the revs are all
reverse */ suggested there was a reason for this, but I can't figure out the
reason, so I removed the reversal of the arguments.  Test case included.

Signed-off-by: Matt McCutchen <matt@mattmccutchen.net>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
15 years agotag: delete TAG_EDITMSG only on successful tag
Jeff King [Sat, 6 Dec 2008 19:40:34 +0000 (14:40 -0500)]
tag: delete TAG_EDITMSG only on successful tag

The user may put some effort into writing an annotated tag
message. When the tagging process later fails (which can
happen fairly easily, since it may be dependent on gpg being
correctly configured and used), there is no record left on
disk of the tag message.

Instead, let's keep the TAG_EDITMSG file around until we are
sure the tag has been created successfully. If we die
because of an error, the user can recover their text from
that file. Leaving the file in place causes no conflicts;
it will be silently overwritten by the next annotated tag
creation.

This matches the behavior of COMMIT_EDITMSG, which stays
around in case of error.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogitweb: Make project specific override for 'grep' feature work
Jakub Narebski [Sun, 7 Dec 2008 09:36:36 +0000 (10:36 +0100)]
gitweb: Make project specific override for 'grep' feature work

The 'grep' feature was marked in the comments as having project
specific config, but it lacked 'sub' key required for it to work.

Kind-of-Noticed-by: Matt Kraai <kraai@ftbfs.org>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agohttp.c: use 'git_config_string' to get 'curl_http_proxy'
Miklos Vajna [Sun, 7 Dec 2008 00:45:37 +0000 (01:45 +0100)]
http.c: use 'git_config_string' to get 'curl_http_proxy'

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agofetch-pack: Avoid memcpy() with src==dst
Thomas Rast [Sat, 6 Dec 2008 20:50:09 +0000 (21:50 +0100)]
fetch-pack: Avoid memcpy() with src==dst

memcpy() may only be used for disjoint memory areas, but when invoked
from cmd_fetch_pack(), we have my_args == &args.  (The argument cannot
be removed entirely because transport.c invokes with its own
variable.)

Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'jk/maint-commit-v-strip' into maint
Junio C Hamano [Wed, 3 Dec 2008 07:47:25 +0000 (23:47 -0800)]
Merge branch 'jk/maint-commit-v-strip' into maint

* jk/maint-commit-v-strip:
  commit: Fix stripping of patch in verbose mode.

15 years agoxdiff: give up scanning similar lines early
Davide Libenzi [Sat, 8 Nov 2008 05:24:33 +0000 (21:24 -0800)]
xdiff: give up scanning similar lines early

In a corner case of large files whose lines do not match uniquely, the
loop to eliminate a line that matches multiple locations adjacent to a run
of lines that do not uniquely match wasted too much cycles.  Fix this by
giving up early after scanning 100 lines in both direction.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'bc/maint-keep-pack' into maint
Junio C Hamano [Wed, 3 Dec 2008 07:00:04 +0000 (23:00 -0800)]
Merge branch 'bc/maint-keep-pack' into maint

* bc/maint-keep-pack:
  repack: only unpack-unreachable if we are deleting redundant packs
  t7700: test that 'repack -a' packs alternate packed objects
  pack-objects: extend --local to mean ignore non-local loose objects too
  sha1_file.c: split has_loose_object() into local and non-local counterparts
  t7700: demonstrate mishandling of loose objects in an alternate ODB
  builtin-gc.c: use new pack_keep bitfield to detect .keep file existence
  repack: do not fall back to incremental repacking with [-a|-A]
  repack: don't repack local objects in packs with .keep file
  pack-objects: new option --honor-pack-keep
  packed_git: convert pack_local flag into a bitfield and add pack_keep
  t7700: demonstrate mishandling of objects in packs with a .keep file

15 years agoMerge branch 'js/mingw-rename-fix' into maint
Junio C Hamano [Wed, 3 Dec 2008 06:38:07 +0000 (22:38 -0800)]
Merge branch 'js/mingw-rename-fix' into maint

* js/mingw-rename-fix:
  compat/mingw.c: Teach mingw_rename() to replace read-only files

15 years agoUser's Manual: remove duplicated url at the end of Appendix B
Miklos Vajna [Mon, 1 Dec 2008 18:50:02 +0000 (19:50 +0100)]
User's Manual: remove duplicated url at the end of Appendix B

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoUpdate draft release notes to 1.6.0.5
Junio C Hamano [Mon, 1 Dec 2008 02:33:20 +0000 (18:33 -0800)]
Update draft release notes to 1.6.0.5

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'st/maint-tag' into maint
Junio C Hamano [Mon, 1 Dec 2008 02:18:50 +0000 (18:18 -0800)]
Merge branch 'st/maint-tag' into maint

* st/maint-tag:
  tag: Add more tests about mixing incompatible modes and options
  tag: Check that options are only allowed in the appropriate mode

15 years agoMerge branch 'mk/maint-cg-push' into maint
Junio C Hamano [Mon, 1 Dec 2008 02:18:11 +0000 (18:18 -0800)]
Merge branch 'mk/maint-cg-push' into maint

* mk/maint-cg-push:
  git push: Interpret $GIT_DIR/branches in a Cogito compatible way

15 years agogenerate-cmdlist.sh: avoid selecting synopsis at wrong place
Nguyễn Thái Ngọc Duy [Sun, 30 Nov 2008 10:54:31 +0000 (17:54 +0700)]
generate-cmdlist.sh: avoid selecting synopsis at wrong place

In "common" man pages there is luckily no "NAME" anywhere except at
beginning of documents. If there is another "NAME", sed could
mis-select it and lead to common-cmds.h corruption. So better nail it
at beginning of line, which would reduce corruption chance.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'mv/fast-export' into maint
Junio C Hamano [Fri, 28 Nov 2008 03:23:27 +0000 (19:23 -0800)]
Merge branch 'mv/fast-export' into maint

* mv/fast-export:
  fast-export: use an unsorted string list for extra_refs
  Add new testcase to show fast-export does not always exports all tags

16 years agosha1_file.c: resolve confusion EACCES vs EPERM
Sam Vilain [Fri, 14 Nov 2008 07:19:34 +0000 (20:19 +1300)]
sha1_file.c: resolve confusion EACCES vs EPERM

An earlier commit 916d081 (Nicer error messages in case saving an object
to db goes wrong, 2006-11-09) confused EACCES with EPERM, the latter of
which is an unlikely error from mkstemp().

Signed-off-by: Sam Vilain <sam@vilain.net>
16 years agosha1_file: avoid bogus "file exists" error message
Joey Hess [Thu, 20 Nov 2008 18:56:28 +0000 (13:56 -0500)]
sha1_file: avoid bogus "file exists" error message

This avoids the following misleading error message:

error: unable to create temporary sha1 filename ./objects/15: File exists

mkstemp can fail for many reasons, one of which, ENOENT, can occur if
the directory for the temp file doesn't exist. create_tmpfile tried to
handle this case by always trying to mkdir the directory, even if it
already existed. This caused errno to be clobbered, so one cannot tell
why mkstemp really failed, and it truncated the buffer to just the
directory name, resulting in the strange error message shown above.

Note that in both occasions that I've seen this failure, it has not been
due to a missing directory, or bad permissions, but some other, unknown
mkstemp failure mode that did not occur when I ran git again. This code
could perhaps be made more robust by retrying mkstemp, in case it was a
transient failure.

Signed-off-by: Joey Hess <joey@kitenet.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit checkout: don't warn about unborn branch if -f is already passed
Matt McCutchen [Mon, 24 Nov 2008 06:55:22 +0000 (01:55 -0500)]
git checkout: don't warn about unborn branch if -f is already passed

I think it's unnecessary to warn that the checkout has been forced due to an
unborn current branch if -f has been explicitly passed.  For one project, I am
using git-new-workdir to create workdirs from a bare repository whose HEAD is
set to an unborn branch, and this warning started to irritate me.

Signed-off-by: Matt McCutchen <matt@mattmccutchen.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agobash: offer refs instead of filenames for 'git revert'
SZEDER Gábor [Thu, 27 Nov 2008 13:35:53 +0000 (14:35 +0100)]
bash: offer refs instead of filenames for 'git revert'

The completion script for 'git revert' currently offers options and
filenames.  However, 'git revert' doesn't take any filenames from the
command line, but a single commit.  Therefore, it's more sane to offer
refs instead.

Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agobash: remove dashed command leftovers
SZEDER Gábor [Thu, 27 Nov 2008 13:35:38 +0000 (14:35 +0100)]
bash: remove dashed command leftovers

Commit 5a625b07 (bash: remove fetch, push, pull dashed form leftovers,
2008-10-03) did that already, but there were still some git-cmd left
here and there.

Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'pw/maint-p4' into maint
Junio C Hamano [Thu, 27 Nov 2008 21:18:25 +0000 (13:18 -0800)]
Merge branch 'pw/maint-p4' into maint

* pw/maint-p4:
  git-p4: fix keyword-expansion regex

16 years agogit-p4: fix keyword-expansion regex
Pete Wyckoff [Wed, 26 Nov 2008 18:52:15 +0000 (13:52 -0500)]
git-p4: fix keyword-expansion regex

This text:

     my $dir = $File::Find::dir;
     return if ($dir !~ m,$options->{dirpat}$,);

was improperly converted to:

     my $dir = $File$dir !~ m,$options->{dirpat}$,);

by the keyword identifier expansion code.  Add a \n
to make sure the regex doesn't go across end-of-line
boundaries.

Signed-off-by: Pete Wyckoff <pw@padd.com>
Acked-by: Simon Hausmann <simon@lst.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix typos in the documentation.
Ralf Wildenhues [Thu, 27 Nov 2008 07:32:01 +0000 (08:32 +0100)]
Fix typos in the documentation.

Signed-off-by: Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'mm/maint-sort-config-doc' into maint
Junio C Hamano [Wed, 26 Nov 2008 19:57:15 +0000 (11:57 -0800)]
Merge branch 'mm/maint-sort-config-doc' into maint

* mm/maint-sort-config-doc:
  config.txt: alphabetize configuration sections

16 years agoconfig.txt: alphabetize configuration sections
Matt McCutchen [Wed, 26 Nov 2008 08:26:50 +0000 (03:26 -0500)]
config.txt: alphabetize configuration sections

I figured the sections might as well be in some order, so I chose alphabetical
but with "core" at the beginning.  This should help people add new variables
in the right places.

Signed-off-by: Matt McCutchen <matt@mattmccutchen.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoTeach "git diff" to honour --[no-]ext-diff
Junio C Hamano [Wed, 26 Nov 2008 17:58:41 +0000 (09:58 -0800)]
Teach "git diff" to honour --[no-]ext-diff

The original intention of 72909be (Add diff-option --ext-diff, 2007-06-30)
was to optionally allow the use of external diff viewer in "git log"
family (while keeping them disabled by default).  It exposed the "allow
external diff" bit to the UI, but forgot to adjust the "git diff" codepath
that was set up to always allow use of the external diff viewer.

Noticed by Nazri Ramliy; tests by René Scharfe squashed in.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agofast-export: use an unsorted string list for extra_refs
Johannes Schindelin [Sun, 23 Nov 2008 11:55:54 +0000 (12:55 +0100)]
fast-export: use an unsorted string list for extra_refs

The list extra_refs contains tags and the objects referenced by them,
so that they can be handled at the end.  When a tag references a
commit, that commit is added to the list using the same name.

Also, the function handle_tags_and_duplicates() relies on the order
the items were added to extra_refs, so clearly we do not want to
use a sorted list here.

Noticed by Miklos Vajna.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Tested-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoAdd new testcase to show fast-export does not always exports all tags
Miklos Vajna [Sat, 22 Nov 2008 18:22:48 +0000 (19:22 +0100)]
Add new testcase to show fast-export does not always exports all tags

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix misleading wording for git-cherry-pick
Bryan Drewery [Thu, 20 Nov 2008 05:11:42 +0000 (23:11 -0600)]
Fix misleading wording for git-cherry-pick

Documentation for -n implies that -x is normally
used, however this is no longer true.

Signed-off-by: Bryan Drewery <bryan@shatow.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agocompat/mingw.c: Teach mingw_rename() to replace read-only files
Johannes Sixt [Wed, 19 Nov 2008 16:25:27 +0000 (17:25 +0100)]
compat/mingw.c: Teach mingw_rename() to replace read-only files

On POSIX, rename() can replace files that are not writable. On Windows,
however, read-only files cannot be replaced without additional efforts:
We have to make the destination writable first.

Since the situations where the destination is read-only are rare, we do not
make the destination writable on every invocation, but only if the first
try to rename a file failed with an "access denied" error.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agorequest-pull: make usage string match manpage
Stefan Naewe [Mon, 17 Nov 2008 08:57:19 +0000 (09:57 +0100)]
request-pull: make usage string match manpage

The usage string of 'git request-pull' differs from he manpage
which gives the correct 'synopsis'.

Signed-off-by: Stefan Naewe <stefan.naewe@atlas-elektronik.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoTeach ls-files --with-tree=<tree> to work with options other than -c
Junio C Hamano [Sun, 16 Nov 2008 08:10:25 +0000 (00:10 -0800)]
Teach ls-files --with-tree=<tree> to work with options other than -c

Originally --with-tree=<tree> was designed for the sole purpose of
checking if a given pathspec makes sense as a parameter to git-commit
using it in conjunction with --error-unmatch.  It had logic to avoid
showing the same entry (one came from the original index, another from the
overlayed tree) twice so that it works with -c (i.e. "show-cached"), but
otherwise it was not designed to work with the flags such as -m, -d, etc.

This teaches the same logic to cover the codepath for -m and -d.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agobuiltin-ls-files.c: coding style fix.
Junio C Hamano [Sun, 16 Nov 2008 08:15:43 +0000 (00:15 -0800)]
builtin-ls-files.c: coding style fix.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: git-svn: fix example for centralized SVN clone
Jan Krüger [Fri, 14 Nov 2008 17:45:14 +0000 (18:45 +0100)]
Documentation: git-svn: fix example for centralized SVN clone

The example that tells users how to centralize the effort of the initial
git svn clone operation doesn't work properly. It uses rebase but that
only works if HEAD exists. This adds one extra command to create a
somewhat sensible HEAD that should work in all cases.

Signed-off-by: Jan Krüger <jk@jk.gs>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: fix links to "everyday.html"
Christian Couder [Fri, 14 Nov 2008 12:26:47 +0000 (13:26 +0100)]
Documentation: fix links to "everyday.html"

In some places the links are wrong. They should be:
"link:everyday.html", instead of: "linkgit:everyday[7]".
This patch fixes that.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agorevision.c: use proper data type in call to sizeof() within xrealloc
Brandon Casey [Thu, 13 Nov 2008 20:20:37 +0000 (14:20 -0600)]
revision.c: use proper data type in call to sizeof() within xrealloc

A type char** was being used instead of char*.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agorepack: only unpack-unreachable if we are deleting redundant packs
Brandon Casey [Thu, 13 Nov 2008 20:11:46 +0000 (14:11 -0600)]
repack: only unpack-unreachable if we are deleting redundant packs

The -A option calls pack-objects with the --unpack-unreachable option so
that the unreachable objects in local packs are left in the local object
store loose. But if the -d option to repack was _not_ used, then these
unpacked loose objects are redundant and unnecessary.

Update tests in t7701.

Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agodate/time: do not get confused by fractional seconds
Linus Torvalds [Sun, 17 Aug 2008 04:25:40 +0000 (21:25 -0700)]
date/time: do not get confused by fractional seconds

The date/time parsing code was confused if the input time HH:MM:SS is
followed by fractional seconds.  Since we do not record anything finer
grained than seconds, we could just drop fractional part, but there is a
twist.

We have taught people that not just spaces but dot can be used as word
separators when spelling things like:

    $ git log --since 2.days
    $ git show @{12:34:56.7.days.ago}

and we shouldn't mistake "7" in the latter example as a fraction and
discard it.

The rules are:

 - valid days of month/mday are always single or double digits.

 - valid years are either two or four digits

   No, we don't support the year 600 _anyway_, since our encoding is based
   on the UNIX epoch, and the day we worry about the year 10,000 is far
   away and we can raise the limit to five digits when we get closer.

 - Other numbers (eg "600 days ago") can have any number of digits, but
   they cannot start with a zero. Again, the only exception is for
   two-digit numbers, since that is fairly common for dates ("Dec 01" is
   not unheard of)

So that means that any milli- or micro-second would be thrown out just
because the number of digits shows that it cannot be an interesting date.

A milli- or micro-second can obviously be a perfectly fine number
according to the rules above, as long as it doesn't start with a '0'. So
if we have

12:34:56.123

then that '123' gets parsed as a number, and we remember it. But because
it's bigger than 31, we'll never use it as such _unless_ there is
something after it to trigger that use.

So you can say "12:34:56.123.days.ago", and because of the "days", that
123 will actually be meaninful now.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agot7700: test that 'repack -a' packs alternate packed objects
Brandon Casey [Thu, 13 Nov 2008 00:50:26 +0000 (18:50 -0600)]
t7700: test that 'repack -a' packs alternate packed objects

Previously, when 'repack -a' was called and there were no packs in the local
repository without a .keep file, the repack would fall back to calling
pack-objects with '--unpacked --incremental'. This resulted in the created
pack file, if any, to be missing the packed objects in the alternate object
store. Test that this specific case has been fixed.

Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoStart 1.6.0.5 cycle
Junio C Hamano [Wed, 12 Nov 2008 23:03:03 +0000 (15:03 -0800)]
Start 1.6.0.5 cycle

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix pack.packSizeLimit and --max-pack-size handling
Nicolas Pitre [Wed, 12 Nov 2008 18:23:58 +0000 (13:23 -0500)]
Fix pack.packSizeLimit and --max-pack-size handling

If the limit was sufficiently low, having a single object written
could bust the limit (by design), but caused the remaining allowed
size to go negative for subsequent objects, which for an unsigned
variable is a rather huge limit.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agocheckout: Fix "initial checkout" detection
Junio C Hamano [Wed, 12 Nov 2008 19:52:35 +0000 (11:52 -0800)]
checkout: Fix "initial checkout" detection

Earlier commit 5521883 (checkout: do not lose staged removal, 2008-09-07)
tightened the rule to prevent switching branches from losing local
changes, so that staged removal of paths can be protected, while
attempting to keep a loophole to still allow a special case of switching
out of an un-checked-out state.

However, the loophole was made a bit too tight, and did not allow
switching from one branch (in an un-checked-out state) to check out
another branch.

The change to builtin-checkout.c in this commit loosens it to allow this,
by not insisting the original commit and the new commit to be the same.

It also introduces a new function, is_index_unborn (and an associated
macro, is_cache_unborn), to check if the repository is truly in an
un-checked-out state more reliably, by making sure that $GIT_INDEX_FILE
did not exist when populating the in-core index structure.  A few places
the earlier commit 5521883 added the check for the initial checkout
condition are updated to use this function.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agocommit: Fix stripping of patch in verbose mode.
Jeff King [Wed, 12 Nov 2008 08:25:52 +0000 (03:25 -0500)]
commit: Fix stripping of patch in verbose mode.

When the "-v" option is given, we put diff of what is to be committed into
the commit template, and then strip it back out again after the user has
edited it.

We used to look for the diff by searching for the "diff --git a/"
header. With diff.mnemonicprefix set in the configuration, however, this
pattern does not match.  The pattern is loosened to cover this case.

Also, if the user puts their own diff in the message (e.g., as a sample
output), then we will accidentally trigger the pattern, removing part of
their output.

We can avoid doing this stripping altogether if the user didn't use "-v"
in the first place, so we know that any match we find will be a false
positive.

[jc: this fix was split out of a series originally meant for master.]

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>