Merge branch 'cb/maint-stash-orphaned-file' into maint
* cb/maint-stash-orphaned-file:
stash tests: stash can lose data in a file removed from the index
stash: Don't overwrite files that have gone from the index
* jn/shortlog:
pretty: Respect --abbrev option
shortlog: Document and test --format option
t4201 (shortlog): Test output format with multiple authors
t4201 (shortlog): guard setup with test_expect_success
Documentation/shortlog: scripted users should not rely on implicit HEAD
Merge branch 'sp/maint-dumb-http-pack-reidx' into maint
* sp/maint-dumb-http-pack-reidx:
http.c::new_http_pack_request: do away with the temp variable filename
http-fetch: Use temporary files for pack-*.idx until verified
http-fetch: Use index-pack rather than verify-pack to check packs
Allow parse_pack_index on temporary files
Extract verify_pack_index for reuse from verify_pack
Introduce close_pack_index to permit replacement
http.c: Remove unnecessary strdup of sha1_to_hex result
http.c: Don't store destination name in request structures
http.c: Drop useless != NULL test in finish_http_pack_request
http.c: Tiny refactoring of finish_http_pack_request
t5550-http-fetch: Use subshell for repository operations
http.c: Remove bad free of static block
Merge branch 'sp/maint-describe-tiebreak-with-tagger-date' into maint
* sp/maint-describe-tiebreak-with-tagger-date:
describe: Break annotated tag ties by tagger date
tag.c: Parse tagger date (if present)
tag.c: Refactor parse_tag_buffer to be saner to program
tag.h: Remove unused signature field
tag.c: Correct indentation
* rc/maint-curl-helper:
remote-curl: ensure that URLs have a trailing slash
http: make end_url_with_slash() public
t5541-http-push: add test for URLs with trailing slash
* hg/maint-attr-fix:
attr: Expand macros immediately when encountered.
attr: Allow multiple changes to an attribute on the same line.
attr: Fixed debug output for macro expansion.
Merge branch 'mh/status-optionally-refresh' into maint
* mh/status-optionally-refresh:
t7508: add a test for "git status" in a read-only repository
git status: refresh the index if possible
t7508: add test for "git status" refreshing the index
add-interactive: Clarify “remaining hunks in the file”
The "a" and "d" commands to ‘add --patch’ (accept/reject rest of file)
interact with "j", "g", and "/" (skip some hunks) in a perhaps
confusing way: after accepting or rejecting all _later_ hunks in the
file, they return to the earlier, skipped hunks and prompt the user
about them again.
This behavior can be very useful in practice. One can still accept or
reject _all_ undecided hunks in a file by using the "g" command to
move to hunk #1 first.
Reported-by: Frédéric Brière <fbriere@fbriere.net> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
AIX doesn't ship with "less" by default, and their "more" is
more featureful than average, so the latter is a more
sensible choice. People who really want less can set the
compile-time option themselves, or users can set $PAGER.
Signed-off-by: Jeff King <peff@peff.net> Tested-by: Tor Arntsen <tor@spacetec.no> Signed-off-by: Junio C Hamano <gitster@pobox.com>
commit --author was added by 146ea06 (git commit --author=$name: look $name up
in existing commits), but its documentation was sorely lacking compared to its
excellent commit message. This commit tries to improve the documentation.
Signed-off-by: Jay Soffian <jaysoffian@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-compat-util.h: use apparently more common __sgi macro to detect SGI IRIX
IRIX 6.5.26m does not define the 'sgi' macro, but it does define an '__sgi'
macro. Since later IRIX versions (6.5.29m) define both macros, and since
an underscore prefixed macro is preferred anyway, use '__sgi' to detect
compilation on SGI IRIX.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Signed-off-by: Gary V. Vaughan <gary@thewrittenword.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation: A...B shortcut for checkout and rebase
Describe the A...B shortcuts for checkout and rebase [-i] which were
introduced in these commits:
619a64e ("checkout A...B" switches to the merge base between A and B, 2009-10-18) 61dfa1b ("rebase --onto A...B" replays history on the merge base between A and B, 2009-11-20) 230a456 (rebase -i: teach --onto A...B syntax, 2010-01-07)
Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/SubmittingPatches: Fix typo in GMail section
Commit e498257d introduced a typo while improving the GMail section
of SubmittingPatches.
Signed-off-by: Tim Henigan <tim.henigan@gmail.com> Acked-by: Michael J Gruber <git@drmicha.warpmail.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Setting NO_CURL leaves some variables like REMOTE_CURL_ALIASES
empty, which creates no fun when for-looping over
$(REMOTE_CURL_ALIASES) unconditionally. Make it conditional.
Reported-by: Paul Walker <PWalker752@aol.com> Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
get_cwd_relative(): do not misinterpret suffix as subdirectory
If the current working directory is the same as the work tree path
plus a suffix, e.g. 'work' and 'work-xyz', then the suffix '-xyz'
would be interpreted as a subdirectory of 'work'.
Signed-off-by: Clemens Buchacher <drizzd@aon.at> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If we are in a git directory, get_git_work_tree() can return NULL.
While trying to determine whether or not the given paths are outside
the work tree, the following command would read from it anyways and
trigger a segmentation fault.
git diff / /
Signed-off-by: Clemens Buchacher <drizzd@aon.at> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix checkout of large files to network shares on Windows XP
Bigger writes to network drives on Windows XP fail. Cap them at 31MB to
allow them to succeed. Callers need to be prepared for write() calls
that do less work than requested anyway.
On local drives, write() calls are translated to WriteFile() calls with
a cap of 64KB on Windows XP and 256KB on Vista. Thus a cap of 31MB won't
affect the number of WriteFile() calls which do the actual work. There's
still room for some other version of Windows to use a chunk size of 1MB
without increasing the number of system calls.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
start_command: close cmd->err descriptor when fork/spawn fails
Fix the problem where the cmd->err passed into start_command wasn't
being properly closed when certain types of errors occurr. (Compare
the affected code with the clean shutdown code later in the function.)
On Windows, this problem would be triggered if mingw_spawnvpe()
failed, which would happen if the command to be executed was malformed
(e.g. a text file that didn't start with a #! line). If cmd->err was
a pipe, the failure to close it could result in a hang while the other
side was waiting (forever) for either input or pipe close, e.g. while
trying to shove the output into the side band. On msysGit, this
problem was causing a hang in t5516-fetch-push.
[J6t: With a slight adjustment of the test case, the hang is also
observed on Linux.]
Signed-off-by: bert Dvornik <dvornik+git@gmail.com> Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix "Out of memory? mmap failed" for files larger than 4GB on Windows
The git_mmap implementation was broken for file sizes that wouldn't fit
into a size_t (32 bits). This was caused by intermediate variables that
were only 32 bits wide when they should be 64 bits.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
According to the default hooks/post-receive file, the hook is called
with three arguments on stdin:
<oldrev> <newrev> <refname>
In command-line mode, the arguments come in a different order, because
the email hook instead calls:
generate_email $2 $3 $1
Add a comment to explain why, based on comments from the mailing list
and the commit message to v1.5.1~9. Thanks to Andy for the
explanation.
Requested-by: martin f. krafft <madduck@debian.org> Cc: Andy Parkins <andyparkins@gmail.com> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The check whether size is zero was done after if size <= SMALL_FILE_SIZE,
as result, zero size case was never triggered. Instead zero length file
was treated as any other small file. This did not caused any problem, but
if we have a special case for size equal to zero, it is better to make it
work and avoid redundant malloc().
Signed-off-by: Dmitry Potapov <dpotapov@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Restrict the tags used to generate the version string to those that
begin with "v", since git's tags for git-core (ie. excluding git-gui)
are all of the form "vX.Y...".
This is to avoid using private tags by the user in a clone of the git
code repository, which may break certain machinery (eg. Makefile, gitk).
Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If we know we are creating a bare repository, we use setenv
to set the GIT_DIR directory to the current directory
(either where we already were, or one we created and chdir'd
into with "git init --bare <dir>").
However, with "git --bare init <dir>" (note the --bare as a
git wrapper option), the setup code actually sets GIT_DIR
for us, but it uses the wrong, original cwd when a directory
is given. Because our setenv does not use the overwrite
flag, it is ignored.
We need to set the overwrite flag, but only when we are
given a directory on the command line. That still allows:
GIT_DIR=foo.git git init --bare
to work. The behavior is changed for:
GIT_DIR=foo.git git init --bare bar.git
which used to create the repository in foo.git, but now will
use bar.git. This is more sane, as command line options
should generally override the environment.
Noticed by Oliver Hoffmann.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When cherry-picking, usually the new and old commit encodings are both
UTF-8. Most old iconv implementations do not support this trivial
conversion, so on old platforms, out->message remains NULL, and later
attempts to read it segfault.
Fix this by noticing the input and output encodings match and skipping
the iconv step, like the other reencode_string() call sites already do.
Also stop segfaulting on other iconv failures: if iconv fails for some
other reason, the best we can do is to pass the old message through.
This fixes a regression introduced in v1.7.1-rc0~15^2~2 (revert:
clarify label on conflict hunks, 2010-03-20).
Reported-by: Andreas Krey <a.krey@gmx.de> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Makefile: Fix 'clean' target to remove all gitweb build files
In particular the gitweb/GITWEB-BUILD-OPTIONS file was not being
removed by the main Makefile. However, the gitweb/Makefile has a
'clean' target that correctly removes all the build products.
In order to fix the problem, rather than duplicate the clean-up
instructions, we change the main Makefile so that it delegates
the clean-up actions to the gitweb Makefile.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Acked-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The documentation erroneously mentions the GIT_NOTES_REWRITE_REF
override in the description of notes.rewrite.<command>. Move it
under notes.rewriteRef where it belongs.
Signed-off-by: Leif Arne Storset <lstorset@opera.com> Acked-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
edf563f (status: make "how to stage" messages optional, 2009-09-09)
introduced advice.statusHints without tests. Add a few tests to describe
and test the status quo.
Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
test-lib: some shells do not let $? propagate into an eval
In 3bf7886 (test-lib: Let tests specify commands to be run at end of
test, 2010-05-02), the git test harness learned to run cleanup
commands unconditionally at the end of a test. During each test,
the intended cleanup actions are collected in the test_cleanup variable
and evaluated. That variable looks something like this:
All cleanup actions are run unconditionally but if one of them fails
it is properly reported through $eval_ret.
On FreeBSD, unfortunately, $? is set at the beginning of an ‘eval’
to 0 instead of the exit status of the previous command. This results
in tests using test_expect_code appearing to fail and all others
appearing to pass, unless their cleanup fails. Avoid the problem by
setting eval_ret before the ‘eval’ begins.
Thanks to Jeff King for the explanation.
Cc: Jeff King <peff@peff.net> Cc: Johannes Sixt <j6t@kdbg.org> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/shortlog: scripted users should not rely on implicit HEAD
When passed no revision arguments, ‘git shortlog’ reads a log from
stdin if and only if stdin is not a tty. So scripts that need to
function identically when standard input is a terminal (as when run
interactively) and not (as when run through a cron job) should either
supply a log themselves or specify the desired revisions explicitly.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
test-lib: Let tests specify commands to be run at end of test
Certain actions can imply that if the test fails early, recovery from
within other tests is too much to expect:
- creating unwritable directories, like the EACCESS test in t0001-init
- setting unusual configuration, like user.signingkey in t7004-tag
- crashing and leaving the index lock held, like t3600-rm once did
Some test scripts work around this by running cleanup actions outside
the supervision of the test harness, with the unfortunate consequence
that those commands are not appropriately echoed and their output not
suppressed. Others explicitly save exit status, clean up, and then
reset the exit status within the tests, which has excellent behavior
but makes the tests hard to read. Still others ignore the problem.
Allow tests a fourth option: by calling this function, tests can
stack up commands they would like to be run to clean up.
Commands passed to test_when_finished during a test are
unconditionally run in the test environment immediately before the
test is completed, in last-in-first-out order. If some cleanup
command fails, then the other cleanup commands are still run before
the failure is reported and the test script allowed to continue.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
xdi_diff_outf() overrides the structure members of its last parameter,
ignoring any value that callers pass in. It's no surprise then that all
callers pass a pointer to an uninitialized structure. They also don't
read it after the call, so the parameter is neither used for input nor
for output. Turn it into a local variable of xdi_diff_outf().
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
test-lib.sh: Add explicit license detail, with change from GPLv2 to GPLv2+.
Dear Junio,
this is a resend of relicensing patch for test suite library, which
was initially sent by Carl Worth. Since the time you sent me acks for
this patch collected by you, I collected 8 additional acks as is
documented at
https://git.wiki.kernel.org/index.php/Test-lib_reclicensing. There are
still three contributors missing: Bert Wesarg, Stephan Beyer and Bryan
Donlan. The contributions of first two are clearly not copyrightable.
I'm not sure about the copyrightability of Bryan Donlan's
contributions (git log -p --author='Bryan Donlan' t/test-lib.sh).
Carl told me that in your ack collection process you missed only three
acks. So I wonder whether you already did some analysis of which
contributions are copyrightable. If so, are the missing acks in the
list bellow?
Thanks
Michal
8<--------8<--------8<--------
This file has had no explicit license information noted in it, but
has clearly been created and modified according to the terms of GPLv2
as with the rest of the git code base.
The purpose of relicensing is to allow other GPLv3+ projects (in
particular, the notmuch project: http://notmuchmail.org) to use this
same test-suite structure and to contribute changes back as well.
Signed-off-by: Carl Worth <cworth@cworth.org> Signed-off-by: Michal Sojka <sojkam1@fel.cvut.cz> Acked-by: Alex Riesen <raa.lkml@gmail.com> Acked-by: Brandon Casey <drafnel@gmail.com> Acked-by: Clemens Buchacher <drizzd@aon.at> Acked-by: David Reiss <dreiss@facebook.com> Acked-by: Emil Sit <sit@emilsit.net> Acked-by: Eric Wong <normalperson@yhbt.net> Acked-by: Fredrik Kuivinen <frekui@gmail.com> Acked-by: Gerrit Pape <pape@smarden.org> Acked-by: Christian Couder <chriscool@tuxfamily.org> Acked-by: Jakub Narebski <jnareb@gmail.com> Acked-by: Jeff King <peff@peff.net> Acked-by: Johan Herland <johan@herland.net> Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Acked-by: Johannes Sixt <j6t@kdbg.org> Acked-by: Jonathan Nieder <jrnieder@gmail.com> Acked-by: Josh Triplett <josh@joshtriplett.org> Acked-by: Junio C Hamano <gitster@pobox.com> Acked-by: Lea Wiemann <lewiemann@gmail.com> Acked-by: Markus Heidelberg <markus.heidelberg@web.de> Acked-by: Martin Waitz <tali@admingilde.org> Acked-by: Matthew Ogilvie <mmogilvi_git@miniinfo.net> Acked-by: Matthias Lederhofer <matled@gmx.net> Acked-by: Michael J Gruber <git@drmicha.warpmail.net> Acked-by: Michele Ballabio <barra_cuda@katamail.com> Acked-by: Miklos Vajna <vmiklos@frugalware.org> Acked-by: Nicolas Pitre <nico@fluxnic.net> Acked-by: Pavel Roskin <proski@gnu.org> Acked-by: Petr Baudis <pasky@ucw.cz> Acked-by: Pierre Habouzit <madcoder@debian.org> Acked-by: Robin Rosenberg <robin.rosenberg@dewire.com> Acked-by: Shawn O. Pearce <spearce@spearce.org> Acked-by: Stephen Boyd <bebarino@gmail.com> Acked-by: Sverre Rabbelier <srabbelier@gmail.com> Acked-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Ever since the xdiff library had been introduced to git, all its callers
have used the flag XDF_NEED_MINIMAL. It makes sure that the smallest
possible diff is produced, but that takes quite some time if there are
lots of differences that can be expressed in multiple ways.
This flag makes a difference for only 0.1% of the non-merge commits in
the git repo of Linux, both in terms of diff size and execution time.
The patches there are mostly nice and small.
SungHyun Nam however reported a case in a different repo where a diff
took more than 20 times longer to generate with XDF_NEED_MINIMAL than
without. Rebasing became really slow.
This patch removes this flag from all callers. The default of xdiff is
saner because it has minimal to no impact in the normal case of small
diffs and doesn't incur that much of a speed penalty for large ones.
A follow-up patch may introduce a command line option to set the flag if
the user needs it, similar to GNU diff's -d/--minimal.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-submodule inherits variables from the environment it is started in,
expects the internal variables init= and recursive= to have an empty
value, but doesn't initialize them appropriately. Thanks to the
selftests, this can be reproduced through
init=1 make test
recursive=1 make test
With this commit the variables are initialized, and the selftests
succeed even if these variables have some values in the environment.
The bug was discovered through the Debian autobuilders
http://bugs.debian.org/569594
Signed-off-by: Gerrit Pape <pape@smarden.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation improvements for the description of short format.
Incorporates the detailed explanation from Jeff King in
<20100410040959.GA11977@coredump.intra.peff.net> and fixes
the bug noted by Junio C Hamano in
<7vmxxc1i8g.fsf@alter.siamese.dyndns.org>.
Signed-off-by: Eric S. Raymond <esr@thyrsus.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/Makefile: fix interrupted builds of user-manual.xml
Unlike gcc, asciidoc does not atomically write its output file or
delete it when interrupted. If it is interrupted in the middle of
writing an XML file, the result will be truncated input for xsltproc.
XSLTPROC user-manual.html
user-manual.xml:998: parser error : Premature end of data in t
Take care of this case by writing to a temporary and renaming it when
finished.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
stash tests: stash can lose data in a file removed from the index
If a file is removed from the index and then modified in the working
tree then stash will discard the working tree file with no way to
recover the changes.
This can might be done in one of a number of ways.
git rm file
vi file # edit a new version
git stash
or with git mv
git mv file newfile
vi file # make a new file with the old name
git stash
Signed-off-by: Charles Bailey <charles@hashpling.org>
When passed an empty list, objects_array_remove_duplicates() corrupts it
by changing the number of entries from 0 to 1.
The problem lies in the condition of its main loop:
for (ref = 0; ref < array->nr - 1; ref++) {
The loop body manipulates the supplied object array. In the case of an
empty array, it should not be doing anything at all. But array->nr is an
unsigned quantity, so the code enters the loop, in particular increasing
array->nr. Fix this by comparing (ref + 1 < array->nr) instead.
The list of commits to bundle appears to be empty because of another bug:
by the time the revision-walking machinery gets to look at it, standard
input has already been consumed by rev-list, so this function gets an
empty list of revisions.
After this patch, git bundle --stdin still does not work; it just doesn’t
segfault any more.
Reported-by: Joey Hess <joey@kitenet.net> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
As long as no rev-list arguments are supplied on the command line,
git bundle create --stdin currently segfaults. With added rev-list
arguments, it does not segfault, but the revisions from stdin are
ignored.
Thanks to Joey Hess <joey@kitenet.net> for the report.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, there are 6 tests which are not even written but are
'test_expect_failure message false'.
Do not abuse test_expect_failure as a to do marker, but mark them as
'#TODO' instead.
Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net> Acked-by: Nguyen Thai Ngoc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
http-fetch: Use temporary files for pack-*.idx until verified
Verify that a downloaded pack-*.idx file is consistent and valid
as an index file before we rename it into its final destination.
This prevents a corrupt index file from later being treated as a
usable file, confusing readers.
Check that we do not have the pack index file before invoking
fetch_pack_index(); that way, we can do without the has_pack_index()
check in fetch_pack_index().
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
http-fetch: Use index-pack rather than verify-pack to check packs
To ensure we don't leave a corrupt pack file positioned as though
it were a valid pack file, run index-pack on the temporary pack
before we rename it to its final name. If index-pack crashes out
when it discovers file corruption (e.g. GitHub's error HTML at the
end of the file), simply delete the temporary files to cleanup.
By waiting until the pack has been validated before we move it
to its final name, we eliminate a race condition where another
concurrent reader might try to access the pack at the same time
that we are still trying to verify its not corrupt.
Switching from verify-pack to index-pack is a change in behavior,
but it should turn out better for users. The index-pack algorithm
tries to minimize disk seeks, as well as the number of times any
given object is inflated, by organizing its work along delta chains.
The verify-pack logic does not attempt to do this, thrashing the
delta base cache and the filesystem cache.
By recreating the index file locally, we also can automatically
upgrade from a v1 pack table of contents to v2. This makes the
CRC32 data available for use during later repacks, even if the
server didn't have them on hand.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Acked-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>