Code

git.git
12 years agoMerge branch 'jk/grep-binary-attribute'
Junio C Hamano [Tue, 14 Feb 2012 20:57:18 +0000 (12:57 -0800)]
Merge branch 'jk/grep-binary-attribute'

* jk/grep-binary-attribute:
  grep: pre-load userdiff drivers when threaded
  grep: load file data after checking binary-ness
  grep: respect diff attributes for binary-ness
  grep: cache userdiff_driver in grep_source
  grep: drop grep_buffer's "name" parameter
  convert git-grep to use grep_source interface
  grep: refactor the concept of "grep source" into an object
  grep: move sha1-reading mutex into low-level code
  grep: make locking flag global

12 years agoMerge branch 'nd/pack-objects-parseopt'
Junio C Hamano [Tue, 14 Feb 2012 20:57:18 +0000 (12:57 -0800)]
Merge branch 'nd/pack-objects-parseopt'

* nd/pack-objects-parseopt:
  pack-objects: convert to use parse_options()
  pack-objects: remove bogus comment
  pack-objects: do not accept "--index-version=version,"

12 years agoMerge branch 'dp/i18n-libcharset'
Junio C Hamano [Tue, 14 Feb 2012 20:57:18 +0000 (12:57 -0800)]
Merge branch 'dp/i18n-libcharset'

* dp/i18n-libcharset:
  Makefile: introduce CHARSET_LIB to link with -lcharset

12 years agoMerge branch 'mh/war-on-extra-refs'
Junio C Hamano [Tue, 14 Feb 2012 20:57:18 +0000 (12:57 -0800)]
Merge branch 'mh/war-on-extra-refs'

* mh/war-on-extra-refs:
  refs: remove the extra_refs API
  clone: do not add alternate references to extra_refs
  everything_local(): mark alternate refs as complete
  fetch-pack.c: inline insert_alternate_refs()
  fetch-pack.c: rename some parameters from "path" to "refname"
  clone.c: move more code into the "if (refs)" conditional
  t5700: document a failure of alternates to affect fetch

12 years agoMerge branch 'lt/pull-no-edit'
Junio C Hamano [Tue, 14 Feb 2012 20:57:17 +0000 (12:57 -0800)]
Merge branch 'lt/pull-no-edit'

* lt/pull-no-edit:
  "git pull" doesn't know "--edit"

12 years agoMerge branch 'bl/gitweb-project-filter'
Junio C Hamano [Tue, 14 Feb 2012 20:57:17 +0000 (12:57 -0800)]
Merge branch 'bl/gitweb-project-filter'

* bl/gitweb-project-filter:
  gitweb: Harden and improve $project_filter page title

12 years agoMerge branch 'jn/ancient-meld-support'
Junio C Hamano [Tue, 14 Feb 2012 20:57:17 +0000 (12:57 -0800)]
Merge branch 'jn/ancient-meld-support'

* jn/ancient-meld-support:
  mergetools/meld: Use --help output to detect --output support

12 years agoMerge branch 'jk/userdiff-config-simplify'
Junio C Hamano [Tue, 14 Feb 2012 20:57:17 +0000 (12:57 -0800)]
Merge branch 'jk/userdiff-config-simplify'

* jk/userdiff-config-simplify:
  drop odd return value semantics from userdiff_config

12 years agoSync with 1.7.9.1
Junio C Hamano [Tue, 14 Feb 2012 17:56:23 +0000 (09:56 -0800)]
Sync with 1.7.9.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoGit 1.7.9.1 v1.7.9.1
Junio C Hamano [Tue, 14 Feb 2012 17:53:38 +0000 (09:53 -0800)]
Git 1.7.9.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'jk/maint-tag-show-fixes' into maint
Junio C Hamano [Tue, 14 Feb 2012 07:31:27 +0000 (23:31 -0800)]
Merge branch 'jk/maint-tag-show-fixes' into maint

* jk/maint-tag-show-fixes:
  tag: do not show non-tag contents with "-n"
  tag: die when listing missing or corrupt objects
  tag: fix output of "tag -n" when errors occur

Conflicts:
t/t7004-tag.sh

12 years agoMerge branch 'bw/inet-pton-ntop-compat' into maint
Junio C Hamano [Tue, 14 Feb 2012 07:26:31 +0000 (23:26 -0800)]
Merge branch 'bw/inet-pton-ntop-compat' into maint

* bw/inet-pton-ntop-compat:
  Drop system includes from inet_pton/inet_ntop compatibility wrappers

12 years agoMerge branch 'mp/make-cleanse-x-for-exe' into maint
Junio C Hamano [Tue, 14 Feb 2012 07:26:25 +0000 (23:26 -0800)]
Merge branch 'mp/make-cleanse-x-for-exe' into maint

* mp/make-cleanse-x-for-exe:
  Explicitly set X to avoid potential build breakage

12 years agoMerge branch 'jn/merge-no-edit-fix' into maint
Junio C Hamano [Tue, 14 Feb 2012 07:24:02 +0000 (23:24 -0800)]
Merge branch 'jn/merge-no-edit-fix' into maint

* jn/merge-no-edit-fix:
  merge: do not launch an editor on "--no-edit $tag"

12 years agoSync with maint
Junio C Hamano [Mon, 13 Feb 2012 19:47:01 +0000 (11:47 -0800)]
Sync with maint

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoUpdate draft release notes to 1.7.9.1
Junio C Hamano [Mon, 13 Feb 2012 19:46:47 +0000 (11:46 -0800)]
Update draft release notes to 1.7.9.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'js/add-e-submodule-fix' into maint
Junio C Hamano [Mon, 13 Feb 2012 19:42:18 +0000 (11:42 -0800)]
Merge branch 'js/add-e-submodule-fix' into maint

* js/add-e-submodule-fix:
  add -e: do not show difference in a submodule that is merely dirty

12 years agoMerge branch 'jc/parse-date-raw' into maint
Junio C Hamano [Mon, 13 Feb 2012 19:42:15 +0000 (11:42 -0800)]
Merge branch 'jc/parse-date-raw' into maint

* jc/parse-date-raw:
  parse_date(): '@' prefix forces git-timestamp
  parse_date(): allow ancient git-timestamp

12 years agoMerge branch 'jc/merge-ff-only-stronger-than-signed-merge' into maint
Junio C Hamano [Mon, 13 Feb 2012 19:42:11 +0000 (11:42 -0800)]
Merge branch 'jc/merge-ff-only-stronger-than-signed-merge' into maint

* jc/merge-ff-only-stronger-than-signed-merge:
  merge: do not create a signed tag merge under --ff-only option

12 years agoMerge branch 'jc/branch-desc-typoavoidance' into maint
Junio C Hamano [Mon, 13 Feb 2012 19:42:07 +0000 (11:42 -0800)]
Merge branch 'jc/branch-desc-typoavoidance' into maint

* jc/branch-desc-typoavoidance:
  branch --edit-description: protect against mistyped branch name
  tests: add write_script helper function

12 years agoMerge branch 'jn/rpm-spec' into maint
Junio C Hamano [Mon, 13 Feb 2012 19:42:04 +0000 (11:42 -0800)]
Merge branch 'jn/rpm-spec' into maint

* jn/rpm-spec:
  git.spec: Workaround localized messages not put in any RPM

12 years agot: use sane_unset instead of unset
Ævar Arnfjörð Bjarmason [Sun, 12 Feb 2012 01:05:12 +0000 (01:05 +0000)]
t: use sane_unset instead of unset

Change several tests to use the sane_unset function introduced in
v1.7.3.1-35-g00648ba instead of the built-in unset function.

This fixes a failure I was having on t9130-git-svn-authors-file.sh on
Solaris, and prevents several other issues from occurring.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoRemove Git's support for smoke testing
Ævar Arnfjörð Bjarmason [Fri, 23 Dec 2011 17:08:18 +0000 (17:08 +0000)]
Remove Git's support for smoke testing

I'm no longer running the Git smoke testing service at
smoke.git.nix.is due to Smolder being a fragile piece of software not
having time to follow through on making it easy for third parties to
run and submit their own smoke tests.

So remove the support in Git for sending smoke tests to
smoke.git.nix.is, it's still easy to modify the test suite to submit
smokes somewhere else.

This reverts the following commits:

    Revert "t/README: Add SMOKE_{COMMENT,TAGS}= to smoke_report target" -- e38efac87d
    Revert "t/README: Document the Smoke testing" -- d15e9ebc5c
    Revert "t/Makefile: Create test-results dir for smoke target" -- 617344d77b
    Revert "tests: Infrastructure for Git smoke testing" -- b6b84d1b74

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMakefile: Change the default compiler from "gcc" to "cc"
Ævar Arnfjörð Bjarmason [Tue, 20 Dec 2011 23:40:47 +0000 (23:40 +0000)]
Makefile: Change the default compiler from "gcc" to "cc"

Ever since the very first commit to git.git we've been setting CC to
"gcc". Presumably this is behavior that Linus copied from the Linux
Makefile.

However unlike Linux Git is written in ANSI C and supports a multitude
of compilers, including Clang, Sun Studio, xlc etc. On my Linux box
"cc" is a symlink to clang, and on a Solaris box I have access to "cc"
is Sun Studio's CC.

Both of these are perfectly capable of compiling Git, and it's
annoying to have to specify CC=cc on the command-line when compiling
Git when that's the default behavior of most other portable programs.

So change the default to "cc". Users who want to compile with GCC can
still add "CC=gcc" to the make(1) command-line, but those users who
don't have GCC as their "cc" will see expected behavior, and as a
bonus we'll be more likely to smoke out new compilation warnings from
our distributors since they'll me using a more varied set of compilers
by default.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMakefile: introduce CHARSET_LIB to link with -lcharset
Дилян Палаузов [Sun, 12 Feb 2012 16:23:36 +0000 (17:23 +0100)]
Makefile: introduce CHARSET_LIB to link with -lcharset

On some systems, the function locale_charset() may not be exported from
libiconv but is available from libcharset, and we need -lcharset when
linking.

Introduce a make variable CHARSET_LIB that can be set to -lcharsetlib
on such systems.  Also autodetect this in the configure script by first
looking for the symbol in libiconv, and then libcharset.

Signed-off-by: Дилян Палаузов <dilyan.palauzov@aegee.org>
12 years agoUpdate draft release notes to 1.7.10
Junio C Hamano [Mon, 13 Feb 2012 06:58:36 +0000 (22:58 -0800)]
Update draft release notes to 1.7.10

Again this round mostly consists of fixes for 1.7.9 in preparation for
merging these topics down to maint for 1.7.9.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'jc/checkout-out-of-unborn'
Junio C Hamano [Mon, 13 Feb 2012 06:43:44 +0000 (22:43 -0800)]
Merge branch 'jc/checkout-out-of-unborn'

* jc/checkout-out-of-unborn:
  git checkout -b: allow switching out of an unborn branch

12 years agoMerge branch 'jc/maint-mailmap-output'
Junio C Hamano [Mon, 13 Feb 2012 06:43:39 +0000 (22:43 -0800)]
Merge branch 'jc/maint-mailmap-output'

* jc/maint-mailmap-output:
  mailmap: always return a plain mail address from map_user()

12 years agoMerge branch 'jk/prompt-fallback-to-tty'
Junio C Hamano [Mon, 13 Feb 2012 06:43:34 +0000 (22:43 -0800)]
Merge branch 'jk/prompt-fallback-to-tty'

* jk/prompt-fallback-to-tty:
  prompt: fall back to terminal if askpass fails
  prompt: clean up strbuf usage

12 years agoMerge branch 'jn/gitweb-search-utf-8'
Junio C Hamano [Mon, 13 Feb 2012 06:43:24 +0000 (22:43 -0800)]
Merge branch 'jn/gitweb-search-utf-8'

* jn/gitweb-search-utf-8:
  gitweb: Allow UTF-8 encoded CGI query parameters and path_info

Conflicts:
gitweb/gitweb.perl

12 years agoMerge branch 'nd/diffstat-gramnum'
Junio C Hamano [Mon, 13 Feb 2012 06:43:19 +0000 (22:43 -0800)]
Merge branch 'nd/diffstat-gramnum'

* nd/diffstat-gramnum:
  Use correct grammar in diffstat summary line

12 years agoMerge branch 'nd/find-pack-entry-recent-cache-invalidation'
Junio C Hamano [Mon, 13 Feb 2012 06:43:03 +0000 (22:43 -0800)]
Merge branch 'nd/find-pack-entry-recent-cache-invalidation'

* nd/find-pack-entry-recent-cache-invalidation:
  find_pack_entry(): do not keep packed_git pointer locally
  sha1_file.c: move the core logic of find_pack_entry() into fill_pack_entry()

12 years agoMerge branch 'tt/profile-build-fix'
Junio C Hamano [Mon, 13 Feb 2012 06:42:46 +0000 (22:42 -0800)]
Merge branch 'tt/profile-build-fix'

* tt/profile-build-fix:
  Makefile: fix syntax for older make
  Fix build problems related to profile-directed optimization

12 years agoMerge branch 'nd/cache-tree-api-refactor'
Junio C Hamano [Mon, 13 Feb 2012 06:42:42 +0000 (22:42 -0800)]
Merge branch 'nd/cache-tree-api-refactor'

* nd/cache-tree-api-refactor:
  cache-tree: update API to take abitrary flags

12 years agoMerge branch 'fc/zsh-completion'
Junio C Hamano [Mon, 13 Feb 2012 06:42:35 +0000 (22:42 -0800)]
Merge branch 'fc/zsh-completion'

* fc/zsh-completion:
  completion: simplify __gitcomp and __gitcomp_nl implementations
  completion: use ls -1 instead of rolling a loop to do that ourselves
  completion: work around zsh option propagation bug

12 years agoMerge branch 'mp/make-cleanse-x-for-exe'
Junio C Hamano [Mon, 13 Feb 2012 06:42:17 +0000 (22:42 -0800)]
Merge branch 'mp/make-cleanse-x-for-exe'

* mp/make-cleanse-x-for-exe:
  Explicitly set X to avoid potential build breakage

12 years agoMerge branch 'bw/inet-pton-ntop-compat'
Junio C Hamano [Mon, 13 Feb 2012 06:42:14 +0000 (22:42 -0800)]
Merge branch 'bw/inet-pton-ntop-compat'

* bw/inet-pton-ntop-compat:
  Drop system includes from inet_pton/inet_ntop compatibility wrappers

12 years agoMerge branch 'jc/maint-commit-ignore-i-t-a'
Junio C Hamano [Mon, 13 Feb 2012 06:42:10 +0000 (22:42 -0800)]
Merge branch 'jc/maint-commit-ignore-i-t-a'

* jc/maint-commit-ignore-i-t-a:
  commit: ignore intent-to-add entries instead of refusing

Conflicts:
cache-tree.c

12 years agoMerge branch 'jk/maint-tag-show-fixes'
Junio C Hamano [Mon, 13 Feb 2012 06:42:06 +0000 (22:42 -0800)]
Merge branch 'jk/maint-tag-show-fixes'

* jk/maint-tag-show-fixes:
  tag: do not show non-tag contents with "-n"
  tag: die when listing missing or corrupt objects
  tag: fix output of "tag -n" when errors occur

Conflicts:
t/t7004-tag.sh

12 years agoMerge branch 'mm/empty-loose-error-message'
Junio C Hamano [Mon, 13 Feb 2012 06:42:02 +0000 (22:42 -0800)]
Merge branch 'mm/empty-loose-error-message'

* mm/empty-loose-error-message:
  fsck: give accurate error message on empty loose object files

12 years agoMerge branch 'nk/ctype-for-perf'
Junio C Hamano [Mon, 13 Feb 2012 06:41:56 +0000 (22:41 -0800)]
Merge branch 'nk/ctype-for-perf'

* nk/ctype-for-perf:
  ctype: implement islower/isupper macro
  ctype.c only wants git-compat-util.h

12 years agoMerge branch 'jn/merge-no-edit-fix'
Junio C Hamano [Mon, 13 Feb 2012 06:41:51 +0000 (22:41 -0800)]
Merge branch 'jn/merge-no-edit-fix'

* jn/merge-no-edit-fix:
  merge: do not launch an editor on "--no-edit $tag"

Conflicts:
builtin/merge.c

12 years agogitweb: Harden and improve $project_filter page title
Jakub Narebski [Sun, 12 Feb 2012 15:21:30 +0000 (16:21 +0100)]
gitweb: Harden and improve $project_filter page title

Commit 19d2d23 (gitweb: add project_filter to limit project list
to a subdirectory, 2012-01-30) added also support for displaying
$project_filter, if present, in page title.

Unfortunately it forgot to treat $project_filter as path, and escape
it using esc_path(), like it is done for $filename.

Also, it was not obvious that "$site_name - $project_filter" is about
project filtering: use "$site_name - projects in '$project_filter'".

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agorefs: remove the extra_refs API
Michael Haggerty [Sat, 11 Feb 2012 06:21:01 +0000 (07:21 +0100)]
refs: remove the extra_refs API

The extra_refs provided a kludgy way to create fake references at a
global level in the hope that they would only affect some particular
code path.  The last user of this API been rewritten, so strip this
stuff out before somebody else gets the bad idea of using it.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoclone: do not add alternate references to extra_refs
Michael Haggerty [Sat, 11 Feb 2012 06:21:00 +0000 (07:21 +0100)]
clone: do not add alternate references to extra_refs

Alternate references are directly (and now, correctly) handled by
fetch-pack, so there is no need to inform fetch-pack about them via
the extra_refs back channel.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoeverything_local(): mark alternate refs as complete
Michael Haggerty [Sat, 11 Feb 2012 06:20:59 +0000 (07:20 +0100)]
everything_local(): mark alternate refs as complete

Objects in an alternate object database are already available to the
local repository and therefore don't need to be fetched.  So mark them
as complete in everything_local().

This fixes a test in t5700.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agofetch-pack.c: inline insert_alternate_refs()
Michael Haggerty [Sat, 11 Feb 2012 06:20:58 +0000 (07:20 +0100)]
fetch-pack.c: inline insert_alternate_refs()

The logic of the (single) caller is clearer without encapsulating this
one line in a function.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agofetch-pack.c: rename some parameters from "path" to "refname"
Michael Haggerty [Sat, 11 Feb 2012 06:20:57 +0000 (07:20 +0100)]
fetch-pack.c: rename some parameters from "path" to "refname"

The parameters denote reference names, which are no longer 1:1 with
filesystem paths.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoclone.c: move more code into the "if (refs)" conditional
Michael Haggerty [Sat, 11 Feb 2012 06:20:56 +0000 (07:20 +0100)]
clone.c: move more code into the "if (refs)" conditional

The bahavior of a bunch of code before the "if (refs)" statement also
depends on whether refs is set, so make the logic clearer by shifting
this code into the if statement.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agot5700: document a failure of alternates to affect fetch
Michael Haggerty [Sat, 11 Feb 2012 06:20:55 +0000 (07:20 +0100)]
t5700: document a failure of alternates to affect fetch

If an alternate supplies some, but not all, of the objects needed for
a fetch, fetch-pack nevertheless generates "want" lines for the
alternate objects that are present.  Demonstrate this problem via a
failing test.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years ago"git pull" doesn't know "--edit"
Linus Torvalds [Sat, 11 Feb 2012 18:21:03 +0000 (10:21 -0800)]
"git pull" doesn't know "--edit"

Ok, so now "git merge" defaults to editing when interactive - lovely. But
when testing that, I noticed that while you can say

   git merge --[no-]edit ..branch..

that does not work with "git pull". You get a message like

  error: unknown option `no-edit'
  usage: git fetch [<options>] [<repository> [<refspec>...]]
     or: git fetch [<options>] <group>
     or: git fetch --multiple [<options>] [(<repository> | <group>)...]
     or: git fetch --all [<options>]

      -v, --verbose         be more verbose
      -q, --quiet           be more quiet
      --all                 fetch from all remotes
  ...

which is because that stupid shell script doesn't know about the new
flags, and just passes it to "git fetch" instead.

Now, I really wanted to just make "git pull" a built-in instead of that
nasty shell script, but I'm lazy. So here's the trivial updates to
git-pull.sh to at least teach it about -e/--edit/--no-edit.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agomergetools/meld: Use --help output to detect --output support
Jonathan Nieder [Fri, 10 Feb 2012 21:57:55 +0000 (15:57 -0600)]
mergetools/meld: Use --help output to detect --output support

In v1.7.7-rc0~3^2 (2011-08-19), git mergetool's "meld" support learned
to use the --output option when calling versions of meld that are
detected to support it (1.5.0 and newer, hopefully).

Alas, it misdetects old versions (before 1.1.5, 2006-06-11) of meld as
supporting the option, so on systems with such meld, instead of
getting a nice merge helper, the operator gets a dialog box with the
text "Wrong number of arguments (Got 5)".  (Version 1.1.5 is when meld
switched to using optparse.  One consequence of that change was that
errors in usage are detected and signalled through the exit status
even when --help was passed.)

Luckily there is a simpler check that is more reliable: the usage
string printed by "meld --help" reliably reflects whether --output is
supported in a given version.  Use it.

Reported-by: Jeff Epler <jepler@unpythonic.net>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoUpdate draft release notes to 1.7.10
Junio C Hamano [Fri, 10 Feb 2012 22:15:38 +0000 (14:15 -0800)]
Update draft release notes to 1.7.10

Document bunch of bugfix topics to be merged down to 'maint' soonish
for 1.7.9.1 maintenance release.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'js/add-e-submodule-fix'
Junio C Hamano [Fri, 10 Feb 2012 22:08:27 +0000 (14:08 -0800)]
Merge branch 'js/add-e-submodule-fix'

* js/add-e-submodule-fix:
  add -e: do not show difference in a submodule that is merely dirty

12 years agoMerge branch 'jc/parse-date-raw'
Junio C Hamano [Fri, 10 Feb 2012 22:08:12 +0000 (14:08 -0800)]
Merge branch 'jc/parse-date-raw'

* jc/parse-date-raw:
  parse_date(): '@' prefix forces git-timestamp
  parse_date(): allow ancient git-timestamp

12 years agoMerge branch 'jc/merge-ff-only-stronger-than-signed-merge'
Junio C Hamano [Fri, 10 Feb 2012 22:08:02 +0000 (14:08 -0800)]
Merge branch 'jc/merge-ff-only-stronger-than-signed-merge'

* jc/merge-ff-only-stronger-than-signed-merge:
  merge: do not create a signed tag merge under --ff-only option

12 years agoMerge branch 'jc/branch-desc-typoavoidance'
Junio C Hamano [Fri, 10 Feb 2012 22:07:58 +0000 (14:07 -0800)]
Merge branch 'jc/branch-desc-typoavoidance'

* jc/branch-desc-typoavoidance:
  branch --edit-description: protect against mistyped branch name

12 years agoMerge branch 'jn/rpm-spec'
Junio C Hamano [Fri, 10 Feb 2012 22:07:49 +0000 (14:07 -0800)]
Merge branch 'jn/rpm-spec'

* jn/rpm-spec:
  git.spec: Workaround localized messages not put in any RPM

12 years agoMerge branch 'jk/tests-write-script'
Junio C Hamano [Fri, 10 Feb 2012 22:07:42 +0000 (14:07 -0800)]
Merge branch 'jk/tests-write-script'

* jk/tests-write-script:
  t0300: use write_script helper
  tests: add write_script helper function

12 years agoSync with maint
Junio C Hamano [Fri, 10 Feb 2012 22:04:32 +0000 (14:04 -0800)]
Sync with maint

12 years agoUpdate draft release notes to 1.7.9.1
Junio C Hamano [Fri, 10 Feb 2012 22:04:20 +0000 (14:04 -0800)]
Update draft release notes to 1.7.9.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'jc/maint-request-pull-for-tag' into maint
Junio C Hamano [Fri, 10 Feb 2012 21:59:02 +0000 (13:59 -0800)]
Merge branch 'jc/maint-request-pull-for-tag' into maint

* jc/maint-request-pull-for-tag:
  request-pull: explicitly ask tags/$name to be pulled

12 years agoMerge branch 'tr/grep-l-with-decoration' into maint
Junio C Hamano [Fri, 10 Feb 2012 21:59:02 +0000 (13:59 -0800)]
Merge branch 'tr/grep-l-with-decoration' into maint

* tr/grep-l-with-decoration:
  grep: fix -l/-L interaction with decoration lines

12 years agoMerge branch 'jl/submodule-re-add' into maint
Junio C Hamano [Fri, 10 Feb 2012 21:59:01 +0000 (13:59 -0800)]
Merge branch 'jl/submodule-re-add' into maint

* jl/submodule-re-add:
  submodule add: fix breakage when re-adding a deep submodule

12 years agoMerge branch 'da/maint-mergetool-twoway' into maint
Junio C Hamano [Fri, 10 Feb 2012 21:59:01 +0000 (13:59 -0800)]
Merge branch 'da/maint-mergetool-twoway' into maint

* da/maint-mergetool-twoway:
  mergetool: Provide an empty file when needed

12 years agoctype: implement islower/isupper macro
Namhyung Kim [Fri, 10 Feb 2012 02:13:31 +0000 (11:13 +0900)]
ctype: implement islower/isupper macro

"perf" uses a the forked copy of this file, and wants to use these two
macros.

Signed-off-by: Namhyung Kim <namhyung.kim@lge.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoctype.c only wants git-compat-util.h
Namhyung Kim [Fri, 10 Feb 2012 02:13:30 +0000 (11:13 +0900)]
ctype.c only wants git-compat-util.h

The implementation of sane ctype macros only depends on symbols in
git-compat-util.h not cache.h

Signed-off-by: Namhyung Kim <namhyung.kim@lge.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoExplicitly set X to avoid potential build breakage
Michael Palimaka [Wed, 8 Feb 2012 10:59:04 +0000 (21:59 +1100)]
Explicitly set X to avoid potential build breakage

$X is appended to binary names for Windows builds (ie. git.exe).
Pollution from the environment can inadvertently trigger this behaviour,
resulting in 'git' turning into 'gitwhatever' without warning.

Signed-off-by: Michael Palimaka <kensington@astralcloak.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agomerge: do not launch an editor on "--no-edit $tag"
Junio C Hamano [Thu, 9 Feb 2012 21:30:52 +0000 (13:30 -0800)]
merge: do not launch an editor on "--no-edit $tag"

When the user explicitly asked us not to, don't launch an editor.

But do everything else the same way as the "edit" case, i.e. leave the
comment with verification result in the log template and record the
mergesig in the resulting merge commit for later inspection.

Based on initiail analysis by Jonathan Nieder.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMakefile: fix syntax for older make
Johannes Sixt [Thu, 9 Feb 2012 08:22:26 +0000 (09:22 +0100)]
Makefile: fix syntax for older make

It is necessary to write the else branch as a nested conditional. Also,
write the conditions with parentheses because we use them throughout the
Makefile.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agotag: do not show non-tag contents with "-n"
Junio C Hamano [Mon, 6 Feb 2012 18:13:27 +0000 (10:13 -0800)]
tag: do not show non-tag contents with "-n"

"git tag -n" did not check the type of the object it is reading the top n
lines from. At least, avoid showing the beginning of trees and blobs when
dealing with lightweight tags that point at them.

As the payload of a tag and a commit look similar in that they both start
with a header block, which is skipped for the purpose of "-n" output,
followed by human readable text, allow the message of commit objects to be
shown just like the contents of tag objects. This avoids regression for
people who have been using "tag -n" to show the log messages of commits
that are pointed at by lightweight tags.

Test script is from Jeff King.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocache-tree: update API to take abitrary flags
Nguyễn Thái Ngọc Duy [Mon, 16 Jan 2012 02:36:46 +0000 (09:36 +0700)]
cache-tree: update API to take abitrary flags

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoUpdate draft release notes to 1.7.10
Junio C Hamano [Tue, 7 Feb 2012 21:01:56 +0000 (13:01 -0800)]
Update draft release notes to 1.7.10

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'jc/maint-request-pull-for-tag'
Junio C Hamano [Tue, 7 Feb 2012 20:57:10 +0000 (12:57 -0800)]
Merge branch 'jc/maint-request-pull-for-tag'

* jc/maint-request-pull-for-tag:
  request-pull: explicitly ask tags/$name to be pulled

12 years agoMerge branch 'bl/gitweb-project-filter'
Junio C Hamano [Tue, 7 Feb 2012 20:57:05 +0000 (12:57 -0800)]
Merge branch 'bl/gitweb-project-filter'

* bl/gitweb-project-filter:
  gitweb: Make project search respect project_filter
  gitweb: improve usability of projects search form
  gitweb: place links to parent directories in page header
  gitweb: show active project_filter in project_list page header
  gitweb: limit links to alternate forms of project_list to active project_filter
  gitweb: add project_filter to limit project list to a subdirectory
  gitweb: prepare git_get_projects_list for use outside 'forks'.
  gitweb: move hard coded .git suffix out of git_get_projects_list

12 years agoMerge branch 'jx/i18n-more-marking'
Junio C Hamano [Tue, 7 Feb 2012 20:56:54 +0000 (12:56 -0800)]
Merge branch 'jx/i18n-more-marking'

* jx/i18n-more-marking:
  i18n: format_tracking_info "Your branch is behind" message
  i18n: git-commit whence_s "merge/cherry-pick" message

12 years agoMerge branch 'jn/svn-fe'
Junio C Hamano [Tue, 7 Feb 2012 20:56:38 +0000 (12:56 -0800)]
Merge branch 'jn/svn-fe'

* jn/svn-fe: (36 commits)
  vcs-svn: suppress a -Wtype-limits warning
  vcs-svn: allow import of > 4GiB files
  vcs-svn: rename check_overflow arguments for clarity
  vcs-svn/svndiff.c: squelch false "unused" warning from gcc
  vcs-svn: reset first_commit_done in fast_export_init
  vcs-svn: do not initialize report_buffer twice
  vcs-svn: avoid hangs from corrupt deltas
  vcs-svn: guard against overflow when computing preimage length
  vcs-svn: cap number of bytes read from sliding view
  test-svn-fe: split off "test-svn-fe -d" into a separate function
  vcs-svn: implement text-delta handling
  vcs-svn: let deltas use data from preimage
  vcs-svn: let deltas use data from postimage
  vcs-svn: verify that deltas consume all inline data
  vcs-svn: implement copyfrom_data delta instruction
  vcs-svn: read instructions from deltas
  vcs-svn: read inline data from deltas
  vcs-svn: read the preimage when applying deltas
  vcs-svn: parse svndiff0 window header
  vcs-svn: skeleton of an svn delta parser
  ...

12 years agoMerge branch 'rt/completion-branch-edit-desc'
Junio C Hamano [Tue, 7 Feb 2012 20:56:33 +0000 (12:56 -0800)]
Merge branch 'rt/completion-branch-edit-desc'

* rt/completion-branch-edit-desc:
  completion: --edit-description option for git-branch

12 years agocommit: ignore intent-to-add entries instead of refusing
Junio C Hamano [Tue, 7 Feb 2012 19:55:48 +0000 (11:55 -0800)]
commit: ignore intent-to-add entries instead of refusing

Originally, "git add -N" was introduced to help users from forgetting to
add new files to the index before they ran "git commit -a".  As an attempt
to help them further so that they do not forget to say "-a", "git commit"
to commit the index as-is was taught to error out, reminding the user that
they may have forgotten to add the final contents of the paths before
running the command.

This turned out to be a false "safety" that is useless.  If the user made
changes to already tracked paths and paths added with "git add -N", and
then ran "git add" to register the final contents of the paths added with
"git add -N", "git commit" will happily create a commit out of the index,
without including the local changes made to the already tracked paths. It
was not a useful "safety" measure to prevent "forgetful" mistakes from
happening.

It turns out that this behaviour is not just a useless false "safety", but
actively hurts use cases of "git add -N" that were discovered later and
have become popular, namely, to tell Git to be aware of these paths added
by "git add -N", so that commands like "git status" and "git diff" would
include them in their output, even though the user is not interested in
including them in the next commit they are going to make.

Fix this ancient UI mistake, and instead make a commit from the index
ignoring the paths added by "git add -N" without adding real contents.

Based on the work by Nguyễn Thái Ngọc Duy, and helped by injection of
sanity from Jonathan Nieder and others on the Git mailing list.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agodrop odd return value semantics from userdiff_config
Jeff King [Tue, 7 Feb 2012 18:23:02 +0000 (13:23 -0500)]
drop odd return value semantics from userdiff_config

When the userdiff_config function was introduced in be58e70
(diff: unify external diff and funcname parsing code,
2008-10-05), it used a return value convention unlike any
other config callback. Like other callbacks, it used "-1" to
signal error. But it returned "1" to indicate that it found
something, and "0" otherwise; other callbacks simply
returned "0" to indicate that no error occurred.

This distinction was necessary at the time, because the
userdiff namespace overlapped slightly with the color
configuration namespace. So "diff.color.foo" could mean "the
'foo' slot of diff coloring" or "the 'foo' component of the
"color" userdiff driver". Because the color-parsing code
would die on an unknown color slot, we needed the userdiff
code to indicate that it had matched the variable, letting
us bypass the color-parsing code entirely.

Later, in 8b8e862 (ignore unknown color configuration,
2009-12-12), the color-parsing code learned to silently
ignore unknown slots. This means we no longer need to
protect userdiff-matched variables from reaching the
color-parsing code.

We can therefore change the userdiff_config calling
convention to a more normal one. This drops some code from
each caller, which is nice. But more importantly, it reduces
the cognitive load for readers who may wonder why
userdiff_config is unlike every other config callback.

There's no need to add a new test confirming that this
works; t4020 already contains a test that sets
diff.color.external.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoadd -e: do not show difference in a submodule that is merely dirty
Johannes Schindelin [Tue, 7 Feb 2012 04:05:48 +0000 (05:05 +0100)]
add -e: do not show difference in a submodule that is merely dirty

When the HEAD of the submodule matches what is recorded in the index of
the superproject, and it has local changes or untracked files, the patch
offered by "git add -e" for editing shows a diff like this:

    diff --git a/submodule b/submodule
    <header>
    -deadbeef...
    +deadbeef...-dirty

Because applying such a patch has no effect to the index, this is a
useless noise.  Generate the patch with IGNORE_DIRTY_SUBMODULES flag to
prevent such a change from getting reported.

This patch also loses the "-dirty" suffix from the output when the HEAD of
the submodule is different from what is in the index of the superproject.
As such dirtiness expressed by the suffix does not affect the result of
the patch application at all, there is no information lost if we remove
it. The user could still run "git status" before "git add -e" if s/he
cares about the dirtiness.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit checkout -b: allow switching out of an unborn branch
Junio C Hamano [Mon, 30 Jan 2012 20:10:08 +0000 (12:10 -0800)]
git checkout -b: allow switching out of an unborn branch

Running "git checkout -b another" immediately after "git init" when you do
not even have a commit on 'master' fails with:

    $ git checkout -b another
    fatal: You are on a branch yet to be born

This is unnecessary, if we redefine "git checkout -b $name" that does not
take any $start_point (which has to be a commit) as "I want to check out a
new branch $name from the state I am in".

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: simplify __gitcomp and __gitcomp_nl implementations
Felipe Contreras [Thu, 2 Feb 2012 19:48:08 +0000 (11:48 -0800)]
completion: simplify __gitcomp and __gitcomp_nl implementations

These shell functions are written in an unnecessarily verbose way;
simplify their "conditionally use $<number> after checking $# against
<number>" logic by using shell's built-in conditional substitution
facilities.

Also remove the first of the two assignments to IFS in __gitcomp_nl
that does not have any effect.

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: use ls -1 instead of rolling a loop to do that ourselves
Felipe Contreras [Thu, 2 Feb 2012 09:05:29 +0000 (03:05 -0600)]
completion: use ls -1 instead of rolling a loop to do that ourselves

This simplifies the code a great deal.  In particular, it allows us to
get rid of __git_shopt, which is used only in this fuction to enable
'nullglob' in zsh.

[jn: squashed with a patch that actually gets rid of __git_shopt]

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: work around zsh option propagation bug
Felipe Contreras [Thu, 2 Feb 2012 19:26:15 +0000 (11:26 -0800)]
completion: work around zsh option propagation bug

When listing commands in zsh (git <TAB><TAB>), all of them will show up,
instead of only porcelain ones.

The root cause of this is because zsh versions from 4.3.0 to present
(4.3.15) do not correctly propagate the SH_WORD_SPLIT option into the
subshell in ${foo:=$(bar)} expressions. Because of this bug, the list of
all commands was treated as a single word in __git_list_porcelain_commands
and did not match any of the patterns that would usually cause plumbing to
be excluded.

With problematic versions of zsh, after running

emulate sh
fn () {
var='one two'
for v in $var; do echo $v; done
}
x=$(fn)
: ${y=$(fn)}

printing "$x" results in two lines as expected, but printing "$y" results
in a single line because $var is expanded as a single word when evaluating
fn to compute y.

So avoid the construct, and use an explicit 'test -n "$foo" || foo=$(bar)'
instead.

[jn: clarified commit message, indentation style fix]

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agomailmap: always return a plain mail address from map_user()
Junio C Hamano [Mon, 6 Feb 2012 01:31:51 +0000 (17:31 -0800)]
mailmap: always return a plain mail address from map_user()

The callers of map_user() give email and name to it, and expect to get the
up-to-date email and/or name to be used in their output. The function
rewrites the given buffers in place. To optimize the majority of cases,
the function returns 0 when it did not do anything, and it returns 1 when
the caller should use the updated contents.

The 'email' input to the function is terminated by '>' or a NUL (whichever
comes first) for historical reasons, but when a rewrite happens, the value
is replaced with the mailbox inside the <> pair.  However, it failed to
meet this expectation when it only rewrote the name part without rewriting
the email part, and the email in the input was terminated by '>'.

This causes an extra '>' to appear in the output of "blame -e", because the
caller does send in '>'-terminated email, and when the function returned 1
to tell it that rewriting happened, it appends '>' that is necessary when
the email part was rewritten.

The patch looks bigger than it actually is, because this change makes a
variable that points at the end of the email part in the input 'p' live
much longer than it used to, deserving a more descriptive name.

Noticed and diagnosed by Felipe Contreras and Jeff King.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agofsck: give accurate error message on empty loose object files
Matthieu Moy [Mon, 6 Feb 2012 16:24:52 +0000 (17:24 +0100)]
fsck: give accurate error message on empty loose object files

Since 3ba7a065527a (A loose object is not corrupt if it
cannot be read due to EMFILE), "git fsck" on a repository with an empty
loose object file complains with the error message

  fatal: failed to read object <sha1>: Invalid argument

This comes from a failure of mmap on this empty file, which sets errno to
EINVAL. Instead of calling xmmap on empty file, we display a clean error
message ourselves, and return a NULL pointer. The new message is

  error: object file .git/objects/09/<rest-of-sha1> is empty
  fatal: loose object <sha1> (stored in .git/objects/09/<rest-of-sha1>) is corrupt

The second line was already there before the regression in 3ba7a065527a,
and the first is an additional message, that should help diagnosing the
problem for the user.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agotag: die when listing missing or corrupt objects
Jeff King [Mon, 6 Feb 2012 08:13:42 +0000 (03:13 -0500)]
tag: die when listing missing or corrupt objects

We don't usually bother looking at tagged objects at all
when listing. However, if "-n" is specified, we open the
objects to read the annotations of the tags.  If we fail to
read an object, or if the object has zero length, we simply
silently return.

The first case is an indication of a broken or corrupt repo,
and we should notify the user of the error.

The second case is OK to silently ignore; however, the
existing code leaked the buffer returned by read_sha1_file.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agotag: fix output of "tag -n" when errors occur
Jeff King [Mon, 6 Feb 2012 08:13:12 +0000 (03:13 -0500)]
tag: fix output of "tag -n" when errors occur

When "git tag" is instructed to print lines from annotated
tags via "-n", it first prints the tag name, then attempts
to parse and print the lines of the tag object, and then
finally adds a trailing newline.

If an error occurs, we return early from the function and
never print the newline, screwing up the output for the next
tag. Let's factor the line-printing into its own function so
we can manage the early returns better, and make sure that
we always terminate the line.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoFix build problems related to profile-directed optimization
Theodore Ts'o [Mon, 6 Feb 2012 06:00:17 +0000 (01:00 -0500)]
Fix build problems related to profile-directed optimization

There was a number of problems I ran into when trying the
profile-directed optimizations added by Andi Kleen in git commit
7ddc2710b9.  (This was using gcc 4.4 found on many enterprise
distros.)

1) The -fprofile-generate and -fprofile-use commands are incompatible
with ccache; the code ends up looking in the wrong place for the gcda
files based on the ccache object names.

2) If the makefile notices that CFLAGS are different, it will rebuild
all of the binaries.  Hence the recipe originally specified by the
INSTALL file ("make profile-all" followed by "make install") doesn't
work.  It will appear to work, but the binaries will end up getting
built with no optimization.

This patch fixes this by using an explicit set of options passed via
the PROFILE variable then using this to directly manipulate CFLAGS and
EXTLIBS.

The developer can run "make PROFILE=BUILD all ; sudo make
PROFILE=BUILD install" automatically run a two-pass build with the
test suite run in between as the sample workload for the purpose of
recording profiling information to do the profile-directed
optimization.

Alternatively, the profiling version of binaries can be built using:

make PROFILE=GEN PROFILE_DIR=/var/cache/profile all
make PROFILE=GEN install

and then after git has been used for a while, the optimized version of
the binary can be built as follows:

make PROFILE=USE PROFILE_DIR=/var/cache/profile all
make PROFILE=USE install

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoSync with maint
Junio C Hamano [Mon, 6 Feb 2012 08:04:47 +0000 (00:04 -0800)]
Sync with maint

12 years agoPrepare for 1.7.9.1
Junio C Hamano [Mon, 6 Feb 2012 08:03:18 +0000 (00:03 -0800)]
Prepare for 1.7.9.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: --edit and --no-edit for git-merge
Adrian Weimann [Mon, 30 Jan 2012 19:29:33 +0000 (20:29 +0100)]
completion: --edit and --no-edit for git-merge

Signed-off-by: Adrian Weimann <adrian.weimann@googlemail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'sp/smart-http-failure-to-push' into maint
Junio C Hamano [Mon, 6 Feb 2012 07:58:43 +0000 (23:58 -0800)]
Merge branch 'sp/smart-http-failure-to-push' into maint

* sp/smart-http-failure-to-push:
  remote-curl: Fix push status report when all branches fail

12 years agoMerge branch 'jc/maint-log-first-parent-pathspec' into maint
Junio C Hamano [Mon, 6 Feb 2012 07:58:42 +0000 (23:58 -0800)]
Merge branch 'jc/maint-log-first-parent-pathspec' into maint

* jc/maint-log-first-parent-pathspec:
  Making pathspec limited log play nicer with --first-parent

12 years agoMerge branch 'cb/push-quiet' into maint
Junio C Hamano [Mon, 6 Feb 2012 07:58:42 +0000 (23:58 -0800)]
Merge branch 'cb/push-quiet' into maint

* cb/push-quiet:
  t5541: avoid TAP test miscounting
  fix push --quiet: add 'quiet' capability to receive-pack
  server_supports(): parse feature list more carefully

12 years agoMerge branch 'cb/maint-kill-subprocess-upon-signal' into maint
Junio C Hamano [Mon, 6 Feb 2012 07:58:42 +0000 (23:58 -0800)]
Merge branch 'cb/maint-kill-subprocess-upon-signal' into maint

* cb/maint-kill-subprocess-upon-signal:
  dashed externals: kill children on exit
  run-command: optionally kill children on exit

12 years agoSync with 1.7.6.6
Junio C Hamano [Mon, 6 Feb 2012 07:53:21 +0000 (23:53 -0800)]
Sync with 1.7.6.6

* maint-1.7.8:
  Git 1.7.6.6
  imap-send: remove dead code

12 years agoSync with 1.7.6.6
Junio C Hamano [Mon, 6 Feb 2012 07:52:53 +0000 (23:52 -0800)]
Sync with 1.7.6.6

* maint-1.7.7:
  Git 1.7.6.6
  imap-send: remove dead code

12 years agoSync with 1.7.6.6
Junio C Hamano [Mon, 6 Feb 2012 07:52:25 +0000 (23:52 -0800)]
Sync with 1.7.6.6

Signed-off-by: Junio C Hamano <gitster@pobox.com>