Code

git.git
15 years agoSpeed up reflog pruning of unreachable commits
Junio Hamano [Tue, 31 Mar 2009 04:34:14 +0000 (21:34 -0700)]
Speed up reflog pruning of unreachable commits

Instead of doing the (potentially very expensive) "in_merge_base()"
check for each commit that might be pruned if it is unreachable, do a
preparatory reachability graph of the commit space, so that the common
case of being reachable can be tested directly.

[ Cleaned up a bit and tweaked to actually work.  - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoClean up reflog unreachability pruning decision
Linus Torvalds [Tue, 31 Mar 2009 16:45:22 +0000 (09:45 -0700)]
Clean up reflog unreachability pruning decision

This clarifies the pruning rules for unreachable commits by having a
separate helpder function for the unreachability decision.

It's preparation for actual bigger changes to come to speed up the
decision when the reachability calculations become a bottleneck.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.6.2.3 v1.6.2.3
Junio C Hamano [Sun, 12 Apr 2009 22:43:24 +0000 (15:43 -0700)]
GIT 1.6.2.3

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Sun, 12 Apr 2009 22:34:53 +0000 (15:34 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  State the effect of filter-branch on graft explicitly
  process_{tree,blob}: Remove useless xstrdup calls

15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Sun, 12 Apr 2009 22:20:29 +0000 (15:20 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  State the effect of filter-branch on graft explicitly
  process_{tree,blob}: Remove useless xstrdup calls

15 years agoState the effect of filter-branch on graft explicitly
Daniel Cheng (aka SDiZ) [Fri, 10 Apr 2009 06:26:49 +0000 (14:26 +0800)]
State the effect of filter-branch on graft explicitly

Signed-off-by: Daniel Cheng (aka SDiZ) <j16sdiz+freenet@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoprocess_{tree,blob}: Remove useless xstrdup calls
Linus Torvalds [Fri, 10 Apr 2009 22:20:18 +0000 (15:20 -0700)]
process_{tree,blob}: Remove useless xstrdup calls

On Wed, 8 Apr 2009, Björn Steinbrink wrote:
>
> The name of the processed object was duplicated for passing it to
> add_object(), but that already calls path_name, which allocates a new
> string anyway. So the memory allocated by the xstrdup calls just went
> nowhere, leaking memory.

Ack, ack.

There's another easy 5% or so for the built-in object walker: once we've
created the hash from the name, the name isn't interesting any more, and
so something trivial like this can help a bit.

Does it matter? Probably not on its own. But a few more memory saving
tricks and it might all make a difference.

Linus

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoStart 1.6.2.3 preparation
Junio C Hamano [Thu, 9 Apr 2009 06:40:33 +0000 (23:40 -0700)]
Start 1.6.2.3 preparation

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'jc/shared-literally' into maint
Junio C Hamano [Thu, 9 Apr 2009 06:23:41 +0000 (23:23 -0700)]
Merge branch 'jc/shared-literally' into maint

* jc/shared-literally:
  t1301: loosen test for forced modes
  set_shared_perm(): sometimes we know what the final mode bits should look like
  move_temp_to_file(): do not forget to chmod() in "Coda hack" codepath
  Move chmod(foo, 0444) into move_temp_to_file()
  "core.sharedrepository = 0mode" should set, not loosen

15 years agoMerge branch 'jc/maint-1.6.0-diff-borrow-carefully' into maint
Junio C Hamano [Thu, 9 Apr 2009 06:23:17 +0000 (23:23 -0700)]
Merge branch 'jc/maint-1.6.0-diff-borrow-carefully' into maint

* jc/maint-1.6.0-diff-borrow-carefully:
  diff --cached: do not borrow from a work tree when a path is marked as assume-unchanged

15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Thu, 9 Apr 2009 06:22:09 +0000 (23:22 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  process_{tree,blob}: Remove useless xstrdup calls

15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Thu, 9 Apr 2009 06:22:05 +0000 (23:22 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  process_{tree,blob}: Remove useless xstrdup calls

15 years agoMerge branch 'jc/maint-1.6.0-keep-pack' into maint
Junio C Hamano [Thu, 9 Apr 2009 06:21:10 +0000 (23:21 -0700)]
Merge branch 'jc/maint-1.6.0-keep-pack' into maint

* jc/maint-1.6.0-keep-pack:
  pack-objects: don't loosen objects available in alternate or kept packs
  t7700: demonstrate repack flaw which may loosen objects unnecessarily
  Remove --kept-pack-only option and associated infrastructure
  pack-objects: only repack or loosen objects residing in "local" packs
  git-repack.sh: don't use --kept-pack-only option to pack-objects
  t7700-repack: add two new tests demonstrating repacking flaws
  is_kept_pack(): final clean-up
  Simplify is_kept_pack()
  Consolidate ignore_packed logic more
  has_sha1_kept_pack(): take "struct rev_info"
  has_sha1_pack(): refactor "pretend these packs do not exist" interface
  git-repack: resist stray environment variable

Conflicts:
t/t7700-repack.sh

15 years agoMerge branch 'jc/maint-1.6.0-blame-s' into maint-1.6.1
Junio C Hamano [Thu, 9 Apr 2009 06:02:17 +0000 (23:02 -0700)]
Merge branch 'jc/maint-1.6.0-blame-s' into maint-1.6.1

* jc/maint-1.6.0-blame-s:
  blame: read custom grafts given by -S before calling setup_revisions()

Conflicts:
builtin-blame.c

15 years agoMerge branch 'bc/maint-1.6.1-branch-deleted-was' into maint-1.6.1
Junio C Hamano [Thu, 9 Apr 2009 06:01:15 +0000 (23:01 -0700)]
Merge branch 'bc/maint-1.6.1-branch-deleted-was' into maint-1.6.1

* bc/maint-1.6.1-branch-deleted-was:
  git-branch: display "was sha1" on branch deletion rather than just "sha1"

15 years agoMerge branch 'js/maint-1.6.0-exec-path-env' into maint-1.6.1
Junio C Hamano [Thu, 9 Apr 2009 06:01:10 +0000 (23:01 -0700)]
Merge branch 'js/maint-1.6.0-exec-path-env' into maint-1.6.1

* js/maint-1.6.0-exec-path-env:
  Propagate --exec-path setting to external commands via GIT_EXEC_PATH

15 years agoMerge branch 'tr/maint-1.6.1-doc-format-patch--root' into maint-1.6.1
Junio C Hamano [Thu, 9 Apr 2009 06:00:21 +0000 (23:00 -0700)]
Merge branch 'tr/maint-1.6.1-doc-format-patch--root' into maint-1.6.1

* tr/maint-1.6.1-doc-format-patch--root:
  Documentation: format-patch --root clarifications

15 years agoprocess_{tree,blob}: Remove useless xstrdup calls
Björn Steinbrink [Wed, 8 Apr 2009 11:28:54 +0000 (13:28 +0200)]
process_{tree,blob}: Remove useless xstrdup calls

The name of the processed object was duplicated for passing it to
add_object(), but that already calls path_name, which allocates a new
string anyway. So the memory allocated by the xstrdup calls just went
nowhere, leaking memory.

This reduces the RSS usage for a "rev-list --all --objects" by about 10% on
the gentoo repo (fully packed) as well as linux-2.6.git:

    gentoo:
                    | old           | new
    ----------------|-------------------------------
    RSS             |       1537284 |       1388408
    VSZ             |       1816852 |       1667952
    time elapsed    |       1:49.62 |       1:48.99
    min. page faults|        417178 |        379919

    linux-2.6.git:
                    | old           | new
    ----------------|-------------------------------
    RSS             |        324452 |        292996
    VSZ             |        491792 |        460376
    time elapsed    |       0:14.53 |       0:14.28
    min. page faults|         89360 |         81613

Signed-off-by: Björn Steinbrink <B.Steinbrink@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-pull.sh: better warning message for "git pull" on detached head.
Matthieu Moy [Wed, 8 Apr 2009 07:24:03 +0000 (09:24 +0200)]
git-pull.sh: better warning message for "git pull" on detached head.

Otherwise, git complains about not finding a branch to pull from in
'branch..merge', which is hardly understandable. While we're there,
reword the sentences slightly.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Wed, 8 Apr 2009 06:05:14 +0000 (23:05 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  Documentation: clarify .gitattributes search
  git-checkout.txt: clarify that <branch> applies when no path is given.
  git-checkout.txt: fix incorrect statement about HEAD and index

15 years agoChange double quotes to single quotes in message
Jari Aalto [Sat, 21 Mar 2009 09:00:54 +0000 (11:00 +0200)]
Change double quotes to single quotes in message

Most of the time when we give branch name in the message, we quote it
inside a pair of single-quotes.  git-checkout uses double-quotes; this
patch corrects the inconsistency.

Signed-off-by: Jari Aalto <jari.aalto@cante.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Wed, 8 Apr 2009 05:51:14 +0000 (22:51 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  Documentation: clarify .gitattributes search
  git-checkout.txt: clarify that <branch> applies when no path is given.
  git-checkout.txt: fix incorrect statement about HEAD and index

15 years agoDocumentation: clarify .gitattributes search
Jason Merrill [Mon, 6 Apr 2009 15:03:36 +0000 (11:03 -0400)]
Documentation: clarify .gitattributes search

Use the term "toplevel of the work tree" in gitattributes.txt and
gitignore.txt to define the limits of the search for those files.

Signed-off-by: Jason Merrill <jason@redhat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-checkout.txt: clarify that <branch> applies when no path is given.
Matthieu Moy [Tue, 7 Apr 2009 12:43:53 +0000 (14:43 +0200)]
git-checkout.txt: clarify that <branch> applies when no path is given.

Otherwise, the sentence "Defaults to HEAD." can be mis-read to mean
that "git checkout -- hello.c" checks-out from HEAD.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-checkout.txt: fix incorrect statement about HEAD and index
Matthieu Moy [Mon, 6 Apr 2009 20:45:21 +0000 (22:45 +0200)]
git-checkout.txt: fix incorrect statement about HEAD and index

The command "git checkout" checks out from the index by default, not
HEAD (the introducing comment were correct, but the detailled
explanation added below were not).

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit submodule: fix usage line
Julien Danjou [Tue, 31 Mar 2009 15:50:12 +0000 (17:50 +0200)]
git submodule: fix usage line

Actually, you have to set the -b option after the add command.

Signed-off-by: Julien Danjou <julien@danjou.info>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Sun, 5 Apr 2009 08:04:54 +0000 (01:04 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  commit: abort commit if interactive add failed
  git-repack: use non-dashed update-server-info

15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Sun, 5 Apr 2009 08:04:38 +0000 (01:04 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  commit: abort commit if interactive add failed
  git-repack: use non-dashed update-server-info

Conflicts:
builtin-commit.c

15 years agodoc/git-pack-refs: fix two grammar issues
Markus Heidelberg [Sat, 4 Apr 2009 10:35:22 +0000 (12:35 +0200)]
doc/git-pack-refs: fix two grammar issues

Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agocommit: abort commit if interactive add failed
Jeff King [Fri, 3 Apr 2009 19:28:56 +0000 (15:28 -0400)]
commit: abort commit if interactive add failed

Previously we ignored the result of calling add_interactive,
which meant that if an error occurred we simply committed
whatever happened to be in the index.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-repack: use non-dashed update-server-info
Dan McGee [Sat, 4 Apr 2009 16:59:55 +0000 (11:59 -0500)]
git-repack: use non-dashed update-server-info

Signed-off-by: Dan McGee <dpmcgee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.6.2.2 v1.6.2.2
Junio C Hamano [Thu, 2 Apr 2009 19:21:57 +0000 (12:21 -0700)]
GIT 1.6.2.2

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'jc/maint-1.6.0-blame-s' into maint
Junio C Hamano [Thu, 2 Apr 2009 19:14:37 +0000 (12:14 -0700)]
Merge branch 'jc/maint-1.6.0-blame-s' into maint

* jc/maint-1.6.0-blame-s:
  blame: read custom grafts given by -S before calling setup_revisions()

Conflicts:
builtin-blame.c

15 years agoMerge branch 'js/maint-diff-temp-smudge' into maint
Junio C Hamano [Thu, 2 Apr 2009 19:02:30 +0000 (12:02 -0700)]
Merge branch 'js/maint-diff-temp-smudge' into maint

* js/maint-diff-temp-smudge:
  Smudge the files fed to external diff and textconv

15 years agoMerge branch 'js/maint-1.6.0-exec-path-env' into maint
Junio C Hamano [Thu, 2 Apr 2009 19:02:25 +0000 (12:02 -0700)]
Merge branch 'js/maint-1.6.0-exec-path-env' into maint

* js/maint-1.6.0-exec-path-env:
  Propagate --exec-path setting to external commands via GIT_EXEC_PATH

15 years agoMerge branch 'bc/maint-1.6.1-branch-deleted-was' into maint
Junio C Hamano [Thu, 2 Apr 2009 19:02:18 +0000 (12:02 -0700)]
Merge branch 'bc/maint-1.6.1-branch-deleted-was' into maint

* bc/maint-1.6.1-branch-deleted-was:
  git-branch: display "was sha1" on branch deletion rather than just "sha1"

Conflicts:
builtin-branch.c

15 years agoMerge branch 'tr/maint-1.6.1-doc-format-patch--root' into maint
Junio C Hamano [Thu, 2 Apr 2009 19:00:59 +0000 (12:00 -0700)]
Merge branch 'tr/maint-1.6.1-doc-format-patch--root' into maint

* tr/maint-1.6.1-doc-format-patch--root:
  Documentation: format-patch --root clarifications

15 years agoMerge branch 'dm/maint-docco' into maint
Junio C Hamano [Thu, 2 Apr 2009 18:58:39 +0000 (11:58 -0700)]
Merge branch 'dm/maint-docco' into maint

* dm/maint-docco:
  Documentation: Remove spurious uses of "you" in git-bisect.txt.
  Documentation: minor grammatical fix in git-check-ref-format.txt
  Documentation: minor grammatical fixes in git-check-attr.txt
  Documentation: minor grammatical fixes in git-cat-file.txt
  Documentation: minor grammatical fixes and rewording in git-bundle.txt
  Documentation: remove some uses of the passive voice in git-bisect.txt
  Documentation: reword example text in git-bisect.txt.
  Documentation: reworded the "Description" section of git-bisect.txt.
  Documentation: minor grammatical fixes in git-branch.txt.
  Documentation: minor grammatical fixes in git-blame.txt.
  Documentation: reword the "Description" section of git-bisect.txt.
  Documentation: minor grammatical fixes in git-archive.txt.

15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Thu, 2 Apr 2009 05:36:05 +0000 (22:36 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  Documentation: Remove an odd "instead"

15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Thu, 2 Apr 2009 05:35:57 +0000 (22:35 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  Documentation: Remove an odd "instead"

15 years agoDocumentation: Remove an odd "instead"
Holger Weiß [Tue, 31 Mar 2009 16:57:01 +0000 (18:57 +0200)]
Documentation: Remove an odd "instead"

Signed-off-by: Holger Weiß <holger@zedat.fu-berlin.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot1301: loosen test for forced modes
Jeff King [Tue, 31 Mar 2009 20:36:00 +0000 (16:36 -0400)]
t1301: loosen test for forced modes

One of the aspects of the test checked explicitly for the
g+s bit to be set on created directories. However, this is
only the means to an end (the "end" being having the correct
group set). And in fact, on systems where
DIR_HAS_BSD_GROUP_SEMANTICS is set, we do not even need to
use this "means" at all, causing the test to fail.

This patch removes that part of the test. In an ideal world
it would be replaced by a test to check that the group was
properly assigned, but that is difficult to automate because
it requires the user running the test suite be a member of
multiple groups.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agofix portability problem with IS_RUN_COMMAND_ERR
Jeff King [Tue, 31 Mar 2009 12:29:23 +0000 (08:29 -0400)]
fix portability problem with IS_RUN_COMMAND_ERR

Some old versions of gcc don't seem to like us negating an
enum constant. Let's work around it by negating the other
half of the comparison instead.

Reported by Pierre Poissinger on gcc 2.9.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agomailmap: resurrect lower-casing of email addresses
Johannes Schindelin [Tue, 31 Mar 2009 00:18:36 +0000 (02:18 +0200)]
mailmap: resurrect lower-casing of email addresses

Commit 0925ce4(Add map_user() and clear_mailmap() to mailmap) broke the
lower-casing of email addresses.  This mostly did not matter if your
.mailmap has only lower-case email addresses;  However, we did not
require .mailmap to contain lowercase-only email addresses.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoUpdate draft release notes to 1.6.2.2
Junio C Hamano [Mon, 30 Mar 2009 20:29:31 +0000 (13:29 -0700)]
Update draft release notes to 1.6.2.2

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Mon, 30 Mar 2009 20:25:27 +0000 (13:25 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  Fix bash completion in path with spaces
  bash completion: only show 'log --merge' if merging
  git-tag(1): add hint about commit messages
  Documentation: update graph api example.

Conflicts:
contrib/completion/git-completion.bash

15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Mon, 30 Mar 2009 20:23:53 +0000 (13:23 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  Fix bash completion in path with spaces
  bash completion: only show 'log --merge' if merging
  git-tag(1): add hint about commit messages
  Documentation: update graph api example.

15 years agoFix bash completion in path with spaces
Daniel Cheng (aka SDiZ) [Mon, 30 Mar 2009 11:27:37 +0000 (19:27 +0800)]
Fix bash completion in path with spaces

Signed-off-by: Daniel Cheng (aka SDiZ) <j16sdiz+freenet@gmail.com>
Trivially-acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agobash completion: only show 'log --merge' if merging
Thomas Rast [Mon, 16 Feb 2009 16:34:57 +0000 (17:34 +0100)]
bash completion: only show 'log --merge' if merging

The gitk completion only shows --merge if MERGE_HEAD is present.
Do it the same way for git-log completion.

Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-tag(1): add hint about commit messages
Nico -telmich- Schottelius [Mon, 30 Mar 2009 11:58:21 +0000 (13:58 +0200)]
git-tag(1): add hint about commit messages

If a tag is not annotated, git tag displays the commit message
instead. Add this hint to the manpage to unhide this secret.

Signed-off-by: Nico Schottelius <nico@ikn.schottelius.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: update graph api example.
Allan Caffee [Mon, 30 Mar 2009 14:13:45 +0000 (10:13 -0400)]
Documentation: update graph api example.

As of commit 03300c0 the graph API uses '*' for all nodes including merges.
This updates the example in the documentation to match.

Signed-off-by: Allan Caffee <allan.caffee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Mon, 30 Mar 2009 06:11:13 +0000 (23:11 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  import-zips: fix thinko

15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Mon, 30 Mar 2009 06:11:03 +0000 (23:11 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  import-zips: fix thinko

15 years agogit-svn: fix ls-tree usage with dash-prefixed paths
Eric Wong [Sun, 29 Mar 2009 06:10:45 +0000 (23:10 -0700)]
git-svn: fix ls-tree usage with dash-prefixed paths

To find the blob object name given a tree and pathname, we were
incorrectly calling "git ls-tree" with a "--" argument followed
by the pathname of the file we wanted to get.

  git ls-tree <TREE> -- --dashed/path/name.c

Unlike many command-line interfaces, the "--" alone does not
symbolize the end of non-option arguments on the command-line.

ls-tree interprets the "--" as a prefix to match against, thus
the entire contents of the --dashed/* hierarchy would be
returned because the "--" matches "--dashed" and every path
under it.

Thanks to Anton Gyllenberg for pointing me toward the
Twisted repository as a real-world example of this case.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoimport-zips: fix thinko
Johannes Schindelin [Sun, 29 Mar 2009 20:42:27 +0000 (22:42 +0200)]
import-zips: fix thinko

Embarrassingly, the common prefix calculation did not work properly, due
to a mistake in the assignment: instead of assigning the dirname of the
current file name, the dirname of the current common prefix needs to
be assigned to common prefix, when the current prefix does not match the
current file name.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoset_shared_perm(): sometimes we know what the final mode bits should look like
Junio C Hamano [Sat, 28 Mar 2009 06:21:00 +0000 (23:21 -0700)]
set_shared_perm(): sometimes we know what the final mode bits should look like

adjust_shared_perm() first obtains the mode bits from lstat(2), expecting
to find what the result of applying user's umask is, and then tweaks it
as necessary.  When the file to be adjusted is created with mkstemp(3),
however, the mode thusly obtained does not have anything to do with user's
umask, and we would need to start from 0444 in such a case and there is no
point running lstat(2) for such a path.

This introduces a new API set_shared_perm() to bypass the lstat(2) and
instead force setting the mode bits to the desired value directly.
adjust_shared_perm() becomes a thin wrapper to the function.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agomove_temp_to_file(): do not forget to chmod() in "Coda hack" codepath
Junio C Hamano [Sat, 28 Mar 2009 06:14:39 +0000 (23:14 -0700)]
move_temp_to_file(): do not forget to chmod() in "Coda hack" codepath

Now move_temp_to_file() is responsible for doing everything that is
necessary to turn a tempfile in $GIT_DIR into its final form, it must make
sure "Coda hack" codepath correctly makes the file read-only.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Sat, 28 Mar 2009 07:41:50 +0000 (00:41 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  test-lib: Clean up comments and Makefile.
  diff --no-index: Do not generate patch output if other output is requested

15 years agotest-lib: Clean up comments and Makefile.
Emil Sit [Sat, 28 Mar 2009 01:57:18 +0000 (21:57 -0400)]
test-lib: Clean up comments and Makefile.

Bring documentation in test-lib and clean target
in Makefile in-line with abc5d372.

Signed-off-by: Emil Sit <sit@emilsit.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMove chmod(foo, 0444) into move_temp_to_file()
Johan Herland [Thu, 26 Mar 2009 15:16:47 +0000 (16:16 +0100)]
Move chmod(foo, 0444) into move_temp_to_file()

When writing out a loose object or a pack (index), move_temp_to_file() is
called to finalize the resulting file. These files (loose files and packs)
should all have permission mode 0444 (modulo adjust_shared_perm()).
Therefore, instead of doing chmod(foo, 0444) explicitly from each callsite
(or even forgetting to chmod() at all), do the chmod() call from within
move_temp_to_file().

Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years ago"core.sharedrepository = 0mode" should set, not loosen
Junio C Hamano [Wed, 25 Mar 2009 23:19:36 +0000 (16:19 -0700)]
"core.sharedrepository = 0mode" should set, not loosen

This fixes the behaviour of octal notation to how it is defined in the
documentation, while keeping the traditional "loosen only" semantics
intact for "group" and "everybody".

Three main points of this patch are:

 - For an explicit octal notation, the internal shared_repository variable
   is set to a negative value, so that we can tell "group" (which is to
   "OR" in 0660) and 0660 (which is to "SET" to 0660);

 - git-init did not set shared_repository variable early enough to affect
   the initial creation of many files, notably copied templates and the
   configuration.  We set it very early when a command-line option
   specifies a custom value.

 - Many codepaths create files inside $GIT_DIR by various ways that all
   involve mkstemp(), and then call move_temp_to_file() to rename it to
   its final destination.  We can add adjust_shared_perm() call here; for
   the traditional "loosen-only", this would be a no-op for many codepaths
   because the mode is already loose enough, but with the new behaviour it
   makes a difference.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: format-patch --root clarifications
Thomas Rast [Thu, 26 Mar 2009 17:29:25 +0000 (18:29 +0100)]
Documentation: format-patch --root clarifications

Users were confused about the meaning and use of the --root option.
Notably, since 68c2ec7 (format-patch: show patch text for the root
commit, 2009-01-10), --root has nothing to do with showing the patch
text for the root commit any more.

Shorten and clarify the corresponding paragraph in the DESCRIPTION
section, document --root under OPTIONS, and add an explicit note that
root commits are formatted regardless.

Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agodocumentation: update cvsimport description of "-r" for recent clone
Carlo Marcelo Arenas Belon [Thu, 26 Mar 2009 07:08:09 +0000 (00:08 -0700)]
documentation: update cvsimport description of "-r" for recent clone

the "--use-separate-remote" option no longer exists, having since
become the default for a clone.

Signed-off-by: Carlo Marcelo Arenas Belon <carenas@sajinet.com.pe>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: Remove spurious uses of "you" in git-bisect.txt.
David J. Mellor [Thu, 26 Mar 2009 03:44:44 +0000 (20:44 -0700)]
Documentation: Remove spurious uses of "you" in git-bisect.txt.

These were added by accident in a42dea3.

This patch also rewords the description of how ranges of commits can be
skipped.

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agodiff --no-index: Do not generate patch output if other output is requested
Johannes Sixt [Wed, 25 Mar 2009 17:19:46 +0000 (18:19 +0100)]
diff --no-index: Do not generate patch output if other output is requested

Previously, 'git diff --no-index --stat a b' generated patch output in
addition to the --stat output (or whatever other output format was
requested). Now only the requested output is generated, and patch
output remains the default.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoIncrease the size of the die/warning buffer to avoid truncation
Shawn O. Pearce [Tue, 24 Mar 2009 23:10:35 +0000 (16:10 -0700)]
Increase the size of the die/warning buffer to avoid truncation

Long messages like those from lockfile.c when a lock can't be
obtained truncate with only 256 bytes in the message buffer.
Bump it to 1024 to give more space for these longer cases.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Tue, 24 Mar 2009 22:31:21 +0000 (15:31 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  close_sha1_file(): make it easier to diagnose errors
  avoid possible overflow in delta size filtering computation

15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Tue, 24 Mar 2009 22:31:15 +0000 (15:31 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  close_sha1_file(): make it easier to diagnose errors
  avoid possible overflow in delta size filtering computation

15 years agoclose_sha1_file(): make it easier to diagnose errors
Linus Torvalds [Tue, 24 Mar 2009 19:31:36 +0000 (12:31 -0700)]
close_sha1_file(): make it easier to diagnose errors

A bug report with "unable to write sha1 file" made us realize that we do
not have enough information to guess why close() is failing.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoavoid possible overflow in delta size filtering computation
Nicolas Pitre [Tue, 24 Mar 2009 19:56:12 +0000 (15:56 -0400)]
avoid possible overflow in delta size filtering computation

On a 32-bit system, the maximum possible size for an object is less than
4GB, while 64-bit systems may cope with larger objects.  Due to this
limitation, variables holding object sizes are using an unsigned long
type (32 bits on 32-bit systems, or 64 bits on 64-bit systems).

When large objects are encountered, and/or people play with large delta
depth values, it is possible for the maximum allowed delta size
computation to overflow, especially on a 32-bit system.  When this
occurs, surviving result bits may represent a value much smaller than
what it is supposed to be, or even zero.  This prevents some objects
from being deltified although they do get deltified when a smaller depth
limit is used.  Fix this by always performing a 64-bit multiplication.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: minor grammatical fix in git-check-ref-format.txt
David J. Mellor [Mon, 23 Mar 2009 01:00:17 +0000 (18:00 -0700)]
Documentation: minor grammatical fix in git-check-ref-format.txt

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: minor grammatical fixes in git-check-attr.txt
David J. Mellor [Mon, 23 Mar 2009 01:00:16 +0000 (18:00 -0700)]
Documentation: minor grammatical fixes in git-check-attr.txt

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: minor grammatical fixes in git-cat-file.txt
David J. Mellor [Mon, 23 Mar 2009 01:00:15 +0000 (18:00 -0700)]
Documentation: minor grammatical fixes in git-cat-file.txt

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: minor grammatical fixes and rewording in git-bundle.txt
David J. Mellor [Mon, 23 Mar 2009 01:00:14 +0000 (18:00 -0700)]
Documentation: minor grammatical fixes and rewording in git-bundle.txt

This commit also converts all reference specifications to a monospaced font,
as the embedded ~ character used in some of the references sometimes causes
the text up to the next ~ to be displayed incorrectly as a subscript when the
HTML pages are generated. This was tested with asciidoc 8.2.5.

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: remove some uses of the passive voice in git-bisect.txt
David J. Mellor [Mon, 23 Mar 2009 03:11:10 +0000 (20:11 -0700)]
Documentation: remove some uses of the passive voice in git-bisect.txt

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Sun, 22 Mar 2009 22:41:00 +0000 (15:41 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  everyday: use the dashless form of git-init

15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Sun, 22 Mar 2009 22:40:55 +0000 (15:40 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  everyday: use the dashless form of git-init

15 years agodiff --cached: do not borrow from a work tree when a path is marked as assume-unchanged
Junio C Hamano [Sun, 22 Mar 2009 22:26:07 +0000 (15:26 -0700)]
diff --cached: do not borrow from a work tree when a path is marked as assume-unchanged

When the index says that the file in the work tree that corresponds to the
blob object that is used for comparison is known to be unchanged, "diff"
reads from the file and applies convert_to_git(), instead of inflating the
object, to feed the internal diff engine with, because an earlier
benchnark found that it tends to be faster to use this optimization.

However, the index can lie when the path is marked as assume-unchanged.
Disable the optimization for such paths.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoSmudge the files fed to external diff and textconv
Johannes Schindelin [Sat, 21 Mar 2009 11:42:52 +0000 (12:42 +0100)]
Smudge the files fed to external diff and textconv

When preparing temporary files for an external diff or textconv, it is
easier on the external tools, especially when they are implemented using
platform tools, if they are fed the input after convert_to_working_tree().

This fixes msysGit issue 177.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoeveryday: use the dashless form of git-init
David Aguilar [Sun, 22 Mar 2009 09:15:13 +0000 (02:15 -0700)]
everyday: use the dashless form of git-init

The 'Everyday GIT' guide was using the old dashed form
of git-init.

Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoUpdate draft release notes to 1.6.2.2
Junio C Hamano [Sun, 22 Mar 2009 06:21:15 +0000 (23:21 -0700)]
Update draft release notes to 1.6.2.2

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agols-files: require worktree when --deleted is given
Jeff King [Tue, 17 Mar 2009 09:03:19 +0000 (05:03 -0400)]
ls-files: require worktree when --deleted is given

The code will end up calling lstat() to check whether the
file still exists; obviously this doesn't work if we're not
in the worktree.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopickaxe: count regex matches only once
René Scharfe [Mon, 16 Mar 2009 18:38:42 +0000 (19:38 +0100)]
pickaxe: count regex matches only once

When --pickaxe-regex is used, forward past the end of matches instead of
advancing to the byte after their start.  This way matches count only
once, even if the regular expression matches their tail -- like in the
fixed-string fork of the code.

E.g.: /.*/ used to count the number of bytes instead of the number of
lines.  /aa/ resulted in a count of two in "aaa" instead of one.

Also document the fact that regexec() needs a NUL-terminated string as
its second argument by adding an assert().

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Sun, 22 Mar 2009 06:10:42 +0000 (23:10 -0700)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:

15 years agoMerge branch 'tr/maint-1.6.0-send-email-irt' into maint
Junio C Hamano [Sun, 22 Mar 2009 06:09:21 +0000 (23:09 -0700)]
Merge branch 'tr/maint-1.6.0-send-email-irt' into maint

* tr/maint-1.6.0-send-email-irt:
  send-email: test --no-thread --in-reply-to combination
  send-email: respect in-reply-to regardless of threading

Conflicts:
t/t9001-send-email.sh

15 years agoMerge branch 'mg/maint-submodule-normalize-path' into maint
Junio C Hamano [Sun, 22 Mar 2009 06:08:27 +0000 (23:08 -0700)]
Merge branch 'mg/maint-submodule-normalize-path' into maint

* mg/maint-submodule-normalize-path:
  git submodule: Fix adding of submodules at paths with ./, .. and //
  git submodule: Add test cases for git submodule add

15 years agoMerge branch 'rs/memmem' into maint
Junio C Hamano [Sun, 22 Mar 2009 06:08:21 +0000 (23:08 -0700)]
Merge branch 'rs/memmem' into maint

* rs/memmem:
  optimize compat/ memmem()
  diffcore-pickaxe: use memmem()

15 years agoMerge branch 'js/rsync-local' into maint
Junio C Hamano [Sun, 22 Mar 2009 06:03:17 +0000 (23:03 -0700)]
Merge branch 'js/rsync-local' into maint

* js/rsync-local:
  rsync transport: allow local paths, and fix tests

15 years agoMerge branch 'db/maint-missing-origin' into maint
Junio C Hamano [Sun, 22 Mar 2009 06:02:55 +0000 (23:02 -0700)]
Merge branch 'db/maint-missing-origin' into maint

* db/maint-missing-origin:
  Remove total confusion from git-fetch and git-push
  Give error when no remote is configured

15 years agoMerge branch 'jc/maint-1.6.0-read-tree-overlay' into maint
Junio C Hamano [Sun, 22 Mar 2009 06:02:47 +0000 (23:02 -0700)]
Merge branch 'jc/maint-1.6.0-read-tree-overlay' into maint

* jc/maint-1.6.0-read-tree-overlay:
  read-tree A B C: do not create a bogus index and do not segfault

15 years agoMerge branch 'js/maint-1.6.0-path-normalize' into maint-1.6.1
Junio C Hamano [Sun, 22 Mar 2009 05:59:19 +0000 (22:59 -0700)]
Merge branch 'js/maint-1.6.0-path-normalize' into maint-1.6.1

* js/maint-1.6.0-path-normalize:
  Remove unused normalize_absolute_path()
  Test and fix normalize_path_copy()
  Fix GIT_CEILING_DIRECTORIES on Windows
  Move sanitary_path_copy() to path.c and rename it to normalize_path_copy()
  Make test-path-utils more robust against incorrect use

15 years agoMerge branch 'jc/maint-1.6.0-pack-directory' into maint-1.6.1
Junio C Hamano [Sun, 22 Mar 2009 05:53:36 +0000 (22:53 -0700)]
Merge branch 'jc/maint-1.6.0-pack-directory' into maint-1.6.1

* jc/maint-1.6.0-pack-directory:
  Fix odb_mkstemp() on AIX
  Make sure objects/pack exists before creating a new pack

Conflicts:
wrapper.c

15 years agoformat-patch: --numbered-files and --stdout aren't mutually exclusive
Stephen Boyd [Sun, 22 Mar 2009 04:32:43 +0000 (21:32 -0700)]
format-patch: --numbered-files and --stdout aren't mutually exclusive

For example:

    git format-patch --numbered-files --stdout --attach HEAD~~

will create two messages with files 1 and 2 attached respectively.
Without --attach/--inline but with --stdout, --numbered-files option
can be simply ignored, because we are not creating any file ourselves.

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopack-objects: don't loosen objects available in alternate or kept packs
Brandon Casey [Sat, 21 Mar 2009 22:26:11 +0000 (17:26 -0500)]
pack-objects: don't loosen objects available in alternate or kept packs

If pack-objects is called with the --unpack-unreachable option then it
will unpack (i.e. loosen) all unreferenced objects from local not-kept
packs, including those that also exist in packs residing in an alternate
object database or a locally kept pack.  The only user of this option is
git-repack.

In this case, repack will follow the call to pack-objects with a call to
prune-packed, which will delete these newly loosened objects, making the
act of loosening a waste of time.  The unnecessary loosening can be
avoided by checking whether an object exists in a non-local pack or a
locally kept pack before loosening it.

This fixes the 'local packed unreachable obs that exist in alternate ODB
are not loosened' test in t7700.

Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoPropagate --exec-path setting to external commands via GIT_EXEC_PATH
Johannes Sixt [Sat, 21 Mar 2009 22:21:18 +0000 (23:21 +0100)]
Propagate --exec-path setting to external commands via GIT_EXEC_PATH

Let PATH0=$PATH that was set before the invocation.
Let /foo be a build directory.
Let /pfx be the installation prefix.
Let pfxexecpath=/pfx/libexec/git-core.

The following is going on when 'git --exec-path=/foo gc' is invoked:

1. git sets PATH=/foo:$PATH0 using the path from --exec-path

2. gc execs 'git repack' (note: no dash).

3. Since there is a git in /foo (it's a build directory), /foo/git is
   taken.

4. No explicit exec-path is set this time, hence, this secondary git sets
   PATH=$pfxexecpath:/foo:$PATH

5. Since 'repack' is not a built-in, execv_dashed_external execs
   'git-repack' (note: dash).

6. There is a $pfxexecpath/git-repack, and it is taken.

7. This git-repack runs 'git pack-objects' (note: no dash).

8. There is no git in $pfxexecpath, but there is one in /foo. Hence,
   /foo/git is run.

9. pack-objects is a builtin, hence, in effect /foo/git-pack-objects
   is run.

As you can see, the way in which we previously set the PATH allowed to
mix gits of different vintage.  By setting GIT_EXEC_PATH when --exec-path
was given on the command line, we reduce the confusion.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot7700: demonstrate repack flaw which may loosen objects unnecessarily
Brandon Casey [Sat, 21 Mar 2009 22:25:30 +0000 (17:25 -0500)]
t7700: demonstrate repack flaw which may loosen objects unnecessarily

If an unreferenced object exists in both a local pack and in either a pack
residing in an alternate object database or a local kept pack, then the
pack-objects call made by repack will loosen that object only to have it
immediately pruned by repack's call to prune-packed.

Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-branch: display "was sha1" on branch deletion rather than just "sha1"
Brandon Casey [Sun, 22 Mar 2009 00:09:17 +0000 (19:09 -0500)]
git-branch: display "was sha1" on branch deletion rather than just "sha1"

Make it more pleasant to read about a branch deletion by adding "was".
Jeff King suggested this, and I ignored it.  He was right.

Update t3200 test again to match the change in output.

Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoRemove --kept-pack-only option and associated infrastructure
Brandon Casey [Fri, 20 Mar 2009 03:47:54 +0000 (22:47 -0500)]
Remove --kept-pack-only option and associated infrastructure

This option to pack-objects/rev-list was created to improve the -A and -a
options of repack.  It was found to be lacking in that it did not provide
the ability to differentiate between local and non-local kept packs, and
found to be unnecessary since objects residing in local kept packs can be
filtered out by the --honor-pack-keep option.

Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopack-objects: only repack or loosen objects residing in "local" packs
Brandon Casey [Fri, 20 Mar 2009 03:47:52 +0000 (22:47 -0500)]
pack-objects: only repack or loosen objects residing in "local" packs

These two features were invented for use by repack when repack will delete
the local packs that have been made redundant.  The packs accessible
through alternates are not deleted by repack, so the objects contained in
them are still accessible after the local packs are deleted.  They do not
need to be repacked into the new pack or loosened.  For the case of
loosening they would immediately be deleted by the subsequent prune-packed
that is called by repack anyway.

This fixes the test
'packed unreachable obs in alternate ODB are not loosened' in t7700.

Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-repack.sh: don't use --kept-pack-only option to pack-objects
Brandon Casey [Fri, 20 Mar 2009 03:47:51 +0000 (22:47 -0500)]
git-repack.sh: don't use --kept-pack-only option to pack-objects

The --kept-pack-only option to pack-objects treats all kept packs as equal.
This results in objects that reside in an alternate pack that has a .keep
file, not being packed into a newly created pack when the user specifies the
-a option to repack.  Since the user may not have any control over the
alternate database, git should not refrain from repacking those objects
even though they are in a pack with a .keep file.

This fixes the 'packed obs in alternate ODB kept pack are repacked' test in
t7700.

Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>