Code

git.git
14 years agoie_match_stat(): do not ignore skip-worktree bit with CE_MATCH_IGNORE_VALID
Nguyễn Thái Ngọc Duy [Mon, 14 Dec 2009 11:43:58 +0000 (18:43 +0700)]
ie_match_stat(): do not ignore skip-worktree bit with CE_MATCH_IGNORE_VALID

Previously CE_MATCH_IGNORE_VALID flag is used by both valid and
skip-worktree bits. While the two bits have similar behaviour, sharing
this flag means "git update-index --really-refresh" will ignore
skip-worktree while it should not. Instead another flag is
introduced to ignore skip-worktree bit, CE_MATCH_IGNORE_VALID only
applies to valid bit.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agotests: rename duplicate t1009
Jeff King [Wed, 25 Nov 2009 22:46:16 +0000 (17:46 -0500)]
tests: rename duplicate t1009

We should avoid duplicate test numbers, since things like
GIT_SKIP_TESTS consider something like t1009.5 to be
unambiguous.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agosparse checkout: inhibit empty worktree
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:13 +0000 (20:47 +0700)]
sparse checkout: inhibit empty worktree

The way sparse checkout works, users may empty their worktree
completely, because of non-matching sparse-checkout spec, or empty
spec. I believe this is not desired. This patch makes Git refuse to
produce such worktree.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAdd tests for sparse checkout
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:12 +0000 (20:47 +0700)]
Add tests for sparse checkout

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoread-tree: add --no-sparse-checkout to disable sparse checkout support
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:11 +0000 (20:47 +0700)]
read-tree: add --no-sparse-checkout to disable sparse checkout support

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agounpack-trees(): ignore worktree check outside checkout area
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:10 +0000 (20:47 +0700)]
unpack-trees(): ignore worktree check outside checkout area

verify_absent() and verify_uptodate() are used to ensure worktree
is safe to be updated, then CE_REMOVE or CE_UPDATE will be set.
Finally check_updates() bases on CE_REMOVE, CE_UPDATE and the
recently added CE_WT_REMOVE to update working directory accordingly.

The entries that are checked may eventually be left out of checkout
area (done later in apply_sparse_checkout()). We don't want to update
outside checkout area. This patch teaches Git to assume "good",
skip these checks when it's sure those entries will be outside checkout
area, and clear CE_REMOVE|CE_UPDATE that could be set due to this
assumption.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agounpack_trees(): apply $GIT_DIR/info/sparse-checkout to the final index
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:09 +0000 (20:47 +0700)]
unpack_trees(): apply $GIT_DIR/info/sparse-checkout to the final index

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agounpack-trees(): "enable" sparse checkout and load $GIT_DIR/info/sparse-checkout
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:08 +0000 (20:47 +0700)]
unpack-trees(): "enable" sparse checkout and load $GIT_DIR/info/sparse-checkout

This patch introduces core.sparseCheckout, which will control whether
sparse checkout support is enabled in unpack_trees()

It also loads sparse-checkout file that will be used in the next patch.
I split it out so the next patch will be shorter, easier to read.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agounpack-trees.c: generalize verify_* functions
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:07 +0000 (20:47 +0700)]
unpack-trees.c: generalize verify_* functions

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agounpack-trees(): add CE_WT_REMOVE to remove on worktree alone
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:06 +0000 (20:47 +0700)]
unpack-trees(): add CE_WT_REMOVE to remove on worktree alone

CE_REMOVE now removes both worktree and index versions. Sparse
checkout must be able to remove worktree version while keep the
index intact when checkout area is narrowed.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoIntroduce "sparse checkout"
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:05 +0000 (20:47 +0700)]
Introduce "sparse checkout"

With skip-worktree bit, you can manually set it to unwanted files,
then remove them: you would have the so-called sparse checkout. The
disadvantages are:

 - Porcelain tools are not aware of this. Everytime you do an
   operation that may update working directory, skip-worktree may be
   cleared out. You have to set them again.

 - You still have to remove skip-worktree'd files manually, which is
   boring and ineffective.

These will be addressed in the following patches. This patch gives an
idea what is "sparse checkout" in Documentation/git-read-tree.txt.
This file is chosen instead of git-checkout.txt because it is quite
technical and user-unfriendly. I'd expect git-checkout.txt to have
something when Porcelain support is done.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agodir.c: export excluded_1() and add_excludes_from_file_1()
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:04 +0000 (20:47 +0700)]
dir.c: export excluded_1() and add_excludes_from_file_1()

These functions are used to handle .gitignore. They are now exported
so that sparse checkout can reuse.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoexcluded_1(): support exclude files in index
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:03 +0000 (20:47 +0700)]
excluded_1(): support exclude files in index

Index does not really have "directories", attempts to match "foo/"
against index will fail unless someone tries to reconstruct directories
from a list of file.

Observing that dtype in this function can never be NULL (otherwise
it would segfault), dtype NULL will be used to say "hey.. you are
matching against index" and behave properly.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agounpack-trees(): carry skip-worktree bit over in merged_entry()
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:02 +0000 (20:47 +0700)]
unpack-trees(): carry skip-worktree bit over in merged_entry()

In this code path, we would remove "old" and replace it with "merge".
"old" may have skip-worktree bit, so re-add it to "merge".

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoRead .gitignore from index if it is skip-worktree
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:01 +0000 (20:47 +0700)]
Read .gitignore from index if it is skip-worktree

This adds index as a prerequisite for directory listing (with
exclude).  At the moment directory listing is used by "git clean",
"git add", "git ls-files" and "git status"/"git commit" and
unpack_trees()-related commands.  These commands have been
checked/modified to populate index before doing directory listing.

add_excludes_from_file() does not enable this feature, because it
is used to read .git/info/exclude and some explicit files specified
by "git ls-files".

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAvoid writing to buffer in add_excludes_from_file_1()
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:47:00 +0000 (20:47 +0700)]
Avoid writing to buffer in add_excludes_from_file_1()

In the next patch, the buffer that is being used within
add_excludes_from_file_1() comes from another function and does not
have extra space to put \n at the end.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoTeach Git to respect skip-worktree bit (writing part)
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:46:59 +0000 (20:46 +0700)]
Teach Git to respect skip-worktree bit (writing part)

This part is mainly to remove CE_VALID shortcuts (and as a
consequence, ce_uptodate() shortcuts as it may be turned on by
CE_VALID) in writing code path if skip-worktree is used. Various tests
are added to avoid future breakages.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoTeach Git to respect skip-worktree bit (reading part)
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:46:58 +0000 (20:46 +0700)]
Teach Git to respect skip-worktree bit (reading part)

grep: turn on --cached for files that is marked skip-worktree
ls-files: do not check for deleted file that is marked skip-worktree
update-index: ignore update request if it's skip-worktree, while still allows removing
diff*: skip worktree version

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoIntroduce "skip-worktree" bit in index, teach Git to get/set this bit
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:46:57 +0000 (20:46 +0700)]
Introduce "skip-worktree" bit in index, teach Git to get/set this bit

Detail about this bit is in Documentation/git-update-index.txt.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAdd test-index-version
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:46:56 +0000 (20:46 +0700)]
Add test-index-version

Commit 06aaaa0bf70fe37d198893f4e25fa73b6516f8a9 may step index format
version up and down, depends on whether extended flags present in the
index. This adds a test to check for index format version.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoupdate-index: refactor mark_valid() in preparation for new options
Nguyễn Thái Ngọc Duy [Thu, 20 Aug 2009 13:46:55 +0000 (20:46 +0700)]
update-index: refactor mark_valid() in preparation for new options

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoPrevent diff machinery from examining assume-unchanged entries on worktree
Nguyễn Thái Ngọc Duy [Tue, 11 Aug 2009 15:43:59 +0000 (22:43 +0700)]
Prevent diff machinery from examining assume-unchanged entries on worktree

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint'
Junio C Hamano [Tue, 11 Aug 2009 05:23:59 +0000 (22:23 -0700)]
Merge branch 'maint'

* maint:
  Fix typos in git-remote.txt and git-symbolic-ref.txt
  git-instaweb: fix mod_perl detection for apache2

15 years agoMerge branch 'maint-1.6.3' into maint
Junio C Hamano [Tue, 11 Aug 2009 05:23:48 +0000 (22:23 -0700)]
Merge branch 'maint-1.6.3' into maint

* maint-1.6.3:
  Fix typos in git-remote.txt and git-symbolic-ref.txt

15 years agoMerge branch 'jk/push-quiet'
Junio C Hamano [Tue, 11 Aug 2009 05:18:21 +0000 (22:18 -0700)]
Merge branch 'jk/push-quiet'

* jk/push-quiet:
  transport: don't show push status if --quiet is given
  transport: pass "quiet" flag to pack-objects
  push: add --quiet flag

15 years agoMerge branch 'jc/maint-merge-recursive-fix'
Junio C Hamano [Tue, 11 Aug 2009 05:17:01 +0000 (22:17 -0700)]
Merge branch 'jc/maint-merge-recursive-fix'

* jc/maint-merge-recursive-fix:
  merge-recursive: don't segfault while handling rename clashes

15 years agoMerge branch 'zf/maint-gitweb-acname'
Junio C Hamano [Tue, 11 Aug 2009 05:16:47 +0000 (22:16 -0700)]
Merge branch 'zf/maint-gitweb-acname'

* zf/maint-gitweb-acname:
  gitweb: parse_commit_text encoding fix

15 years agoMerge branch 'ns/am-raw-email'
Junio C Hamano [Tue, 11 Aug 2009 05:16:42 +0000 (22:16 -0700)]
Merge branch 'ns/am-raw-email'

* ns/am-raw-email:
  git-am: print fair error message when format detection fails
  am: allow individual e-mail files as input

15 years agoMerge branch 'np/maint-limit-delta-cache'
Junio C Hamano [Tue, 11 Aug 2009 05:16:09 +0000 (22:16 -0700)]
Merge branch 'np/maint-limit-delta-cache'

* np/maint-limit-delta-cache:
  don't let the delta cache grow unbounded in 'git repack'

15 years agoMerge branch 'jp/symlink-dirs'
Junio C Hamano [Tue, 11 Aug 2009 05:15:41 +0000 (22:15 -0700)]
Merge branch 'jp/symlink-dirs'

* jp/symlink-dirs:
  t6035-merge-dir-to-symlink depends on SYMLINKS prerequisite
  git-checkout: be careful about untracked symlinks
  lstat_cache: guard against full match of length of 'name' parameter
  Demonstrate bugs when a directory is replaced with a symlink

15 years agoMerge branch 'mk/grep-max-depth'
Junio C Hamano [Tue, 11 Aug 2009 05:15:12 +0000 (22:15 -0700)]
Merge branch 'mk/grep-max-depth'

* mk/grep-max-depth:
  grep: Add --max-depth option.

15 years agoMerge branch 'js/run-command-updates'
Junio C Hamano [Tue, 11 Aug 2009 05:14:57 +0000 (22:14 -0700)]
Merge branch 'js/run-command-updates'

* js/run-command-updates:
  api-run-command.txt: describe error behavior of run_command functions
  run-command.c: squelch a "use before assignment" warning
  receive-pack: remove unnecessary run_status report
  run_command: report failure to execute the program, but optionally don't
  run_command: encode deadly signal number in the return value
  run_command: report system call errors instead of returning error codes
  run_command: return exit code as positive value
  MinGW: simplify waitpid() emulation macros

15 years agoFix typos in git-remote.txt and git-symbolic-ref.txt
Štěpán Němec [Tue, 11 Aug 2009 00:52:07 +0000 (02:52 +0200)]
Fix typos in git-remote.txt and git-symbolic-ref.txt

Signed-off-by: Štěpán Němec <stepnem@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoCheck return value of ftruncate call in http.c
Jeff Lasslett [Mon, 10 Aug 2009 16:05:06 +0000 (00:05 +0800)]
Check return value of ftruncate call in http.c

In new_http_object_request(), check ftruncate() call return value and
handle possible errors.

Signed-off-by: Jeff Lasslett <jeff.lasslett@gmail.com>
Signed-off-by: Tay Ray Chuan <rctay89@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agohttp.c: replace usage of temporary variable for urls
Tay Ray Chuan [Mon, 10 Aug 2009 15:59:55 +0000 (23:59 +0800)]
http.c: replace usage of temporary variable for urls

Use preq->url in new_http_pack_request and freq->url in
new_http_object_request when calling curl_setopt(CURLOPT_URL), instead
of using an intermediate variable, 'url'.

Signed-off-by: Tay Ray Chuan <rctay89@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agohttp.c: free preq when aborting
Tay Ray Chuan [Mon, 10 Aug 2009 15:55:48 +0000 (23:55 +0800)]
http.c: free preq when aborting

Free preq in new_http_pack_request when aborting. preq was allocated
before jumping to the 'abort' label so this is safe.

Signed-off-by: Tay Ray Chuan <rctay89@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'master' of git://git.bogomips.org/git-svn
Junio C Hamano [Mon, 10 Aug 2009 20:22:33 +0000 (13:22 -0700)]
Merge branch 'master' of git://git.bogomips.org/git-svn

* 'master' of git://git.bogomips.org/git-svn:
  git-svn: ignore leading blank lines in svn:ignore
  svn: Honor --prefix option in init without --stdlayout
  svn: Add && to t9107-git-svn-migrate.sh

15 years agogit-svn: ignore leading blank lines in svn:ignore
Michael Haggerty [Fri, 7 Aug 2009 19:21:21 +0000 (21:21 +0200)]
git-svn: ignore leading blank lines in svn:ignore

Subversion ignores all blank lines in svn:ignore properties.  The old
git-svn code ignored blank lines everywhere except for the first line
of the svn:ignore property.  This patch makes the "git svn
show-ignore" and "git svn create-ignore" commands ignore leading blank
lines, too.

Also include leading blank lines in the test suite.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Acked-by: Eric Wong <normalperson@yhbt.net>
15 years agosvn: Honor --prefix option in init without --stdlayout
Adam Brewster [Tue, 4 Aug 2009 01:40:38 +0000 (21:40 -0400)]
svn: Honor --prefix option in init without --stdlayout

Most users who type

  git svn init file:///tmp/repo --prefix=my-svn/

would expect the root of the svn repository to be tracked by
refs/remotes/my-svn/git-svn.

Acked-by: Eric Wong <normalperson@yhbt.net>
15 years agosvn: Add && to t9107-git-svn-migrate.sh
Adam Brewster [Tue, 4 Aug 2009 01:40:37 +0000 (21:40 -0400)]
svn: Add && to t9107-git-svn-migrate.sh

It was probably intended for the test to fail unless all of the
commands succeed.

[ew: fixed tests to actually work]

Signed-off-by: Eric Wong <normalperson@yhbt.net>
15 years agogit-instaweb: fix mod_perl detection for apache2
Mark A Rada [Sat, 8 Aug 2009 22:24:02 +0000 (18:24 -0400)]
git-instaweb: fix mod_perl detection for apache2

The script was looking for something that matched the '^our $gitbin'
regex, which no longer exists in gitweb.cgi.

Now it looks for 'MOD_PERL', which should be on the line that checks
to see if the script is running in a mod_perl environment.

Signed-off-by: Mark Rada <marada@uwaterloo.ca>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocument 'stash clear' recovery via unreachable commits
Thomas Rast [Sun, 9 Aug 2009 00:47:36 +0000 (02:47 +0200)]
Document 'stash clear' recovery via unreachable commits

Add an example to the stash documentation that shows how to quickly
find candidate commits among the 'git fsck --unreachable' output.
Unless you have merges of branch names containing WIP, or edit your
merge messages to say WIP, there will be no false positives.

Snippet written by Björn "doener" Steinbrink and me after zepolen_
asked on IRC.

Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot/Makefile: include config.mak
Jeff King [Sun, 9 Aug 2009 08:50:37 +0000 (04:50 -0400)]
t/Makefile: include config.mak

This is useful if you want to specify GIT_TEST_OPTS that you
always use.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agotests: allow user to specify trash directory location
Jeff King [Sun, 9 Aug 2009 08:39:45 +0000 (04:39 -0400)]
tests: allow user to specify trash directory location

The tests generate a large amount of I/O activity creating
and destroying repositories and files. We can improve the
time it takes to run the test suite by creating trash
directories on filesystems with better performance
characteristic, even though we may not want the rest of the
git repository on those filesystems (e.g., because they are
not network connected, or because they are temporary
ramdisks).

For example, on a dual processor system:

  $ cd t && time make -j32
  real    1m51.562s
  user    0m59.260s
  sys     1m20.933s

  # /dev/shm is tmpfs
  $ cd t && time make -j32 GIT_TEST_OPTS="--root=/dev/shm"
  real    1m1.484s
  user    0m53.555s
  sys     1m5.264s

We almost halve the wall clock time, and we utilize the
dual processors much better.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agotests: provide $TRASH_DIRECTORY variable
Jeff King [Sun, 9 Aug 2009 08:38:11 +0000 (04:38 -0400)]
tests: provide $TRASH_DIRECTORY variable

Most scripts don't care about the absolute path to the trash
directory. The one exception was t4014 script, which pieced
together $TEST_DIRECTORY and $test itself to get an absolute
directory.

Instead, let's provide a $TRASH_DIRECTORY which specifies
the same thing. This keeps the $test variable internal to
test-lib.sh and paves the way for trash directories in other
locations.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agotests: use "$TEST_DIRECTORY" instead of ".."
Jeff King [Sun, 9 Aug 2009 08:37:52 +0000 (04:37 -0400)]
tests: use "$TEST_DIRECTORY" instead of ".."

The $TEST_DIRECTORY variable allows tests to find the
top-level test directory regardless of the current working
directory.

In the past, this has been used to accomodate tests which
change directories, but it is also the first step to being
able to move trash directories outside of the
$TEST_DIRECTORY hierarchy.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot0001-init: split the existence test from the permission test
Johannes Sixt [Sun, 9 Aug 2009 16:02:55 +0000 (18:02 +0200)]
t0001-init: split the existence test from the permission test

The test for correct permissions after init created a deep directory
must be guarded by POSIXPERM. But testing that the deep dirctory exists
is good even on platforms that do not provide the POSIXPERM prerequiste.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot0001-init: fix a file name
Johannes Sixt [Sun, 9 Aug 2009 15:38:04 +0000 (17:38 +0200)]
t0001-init: fix a file name

Without this change, grep fails because it does not find the file
instead of because it does not find the text in the file.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot6035-merge-dir-to-symlink depends on SYMLINKS prerequisite
Johannes Sixt [Sun, 9 Aug 2009 15:35:02 +0000 (17:35 +0200)]
t6035-merge-dir-to-symlink depends on SYMLINKS prerequisite

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoapi-run-command.txt: describe error behavior of run_command functions
Johannes Sixt [Sat, 8 Aug 2009 20:44:20 +0000 (22:44 +0200)]
api-run-command.txt: describe error behavior of run_command functions

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint'
Junio C Hamano [Sat, 8 Aug 2009 05:35:17 +0000 (22:35 -0700)]
Merge branch 'maint'

* maint:
  verify-pack -v: do not report "chain length 0"
  t5510: harden the way verify-pack is used

15 years agoMerge branch 'maint-1.6.3' into maint
Junio C Hamano [Sat, 8 Aug 2009 03:44:26 +0000 (20:44 -0700)]
Merge branch 'maint-1.6.3' into maint

* maint-1.6.3:
  verify-pack -v: do not report "chain length 0"
  t5510: harden the way verify-pack is used

15 years agoMerge branch 'maint-1.6.2' into maint-1.6.3
Junio C Hamano [Sat, 8 Aug 2009 03:44:15 +0000 (20:44 -0700)]
Merge branch 'maint-1.6.2' into maint-1.6.3

* maint-1.6.2:
  verify-pack -v: do not report "chain length 0"
  t5510: harden the way verify-pack is used

15 years agoMerge branch 'maint-1.6.1' into maint-1.6.2
Junio C Hamano [Sat, 8 Aug 2009 03:44:09 +0000 (20:44 -0700)]
Merge branch 'maint-1.6.1' into maint-1.6.2

* maint-1.6.1:
  verify-pack -v: do not report "chain length 0"
  t5510: harden the way verify-pack is used

15 years agoMerge branch 'maint-1.6.0' into maint-1.6.1
Junio C Hamano [Sat, 8 Aug 2009 03:44:02 +0000 (20:44 -0700)]
Merge branch 'maint-1.6.0' into maint-1.6.1

* maint-1.6.0:
  verify-pack -v: do not report "chain length 0"
  t5510: harden the way verify-pack is used

15 years agoverify-pack -v: do not report "chain length 0"
Junio C Hamano [Fri, 7 Aug 2009 22:36:31 +0000 (15:36 -0700)]
verify-pack -v: do not report "chain length 0"

When making a histogram of delta chain length in the pack, the program
collects number of objects whose delta depth exceeds the MAX_CHAIN limit
in histogram[0], and showed it as the number of items that exceeds the
limit correctly.  HOWEVER, it also showed the same number labeled as
"chain length = 0".

In fact, we are not showing the number of objects whose chain length is
zero, i.e. the base objects.  Correct this.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot5510: harden the way verify-pack is used
Junio C Hamano [Sat, 8 Aug 2009 03:12:13 +0000 (20:12 -0700)]
t5510: harden the way verify-pack is used

The test ignored the exit status from verify pack command, and also relied
on not seeing any delta chain statistics.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoShow usage string for 'git log -h', 'git show -h' and 'git diff -h'
Matthieu Moy [Thu, 6 Aug 2009 10:47:21 +0000 (12:47 +0200)]
Show usage string for 'git log -h', 'git show -h' and 'git diff -h'

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-am: print fair error message when format detection fails
Nicolas Sebrecht [Fri, 7 Aug 2009 01:08:13 +0000 (20:08 -0500)]
git-am: print fair error message when format detection fails

Avoid git ending with this message:
"Patch format  is not supported."

With improved error message in the format detection failure case by
Giuseppe Bilotta.

Signed-off-by: Nicolas Sebrecht <ni.s@laposte.net>
Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoam: allow individual e-mail files as input
Junio C Hamano [Fri, 7 Aug 2009 01:08:12 +0000 (20:08 -0500)]
am: allow individual e-mail files as input

We traditionally allowed a mbox file or a directory name of a maildir (but
never an individual file inside a maildir) to be given to "git am".  Even
though an individual file in a maildir (or more generally, a piece of
RFC2822 e-mail) is not a mbox file, it contains enough information to
create a commit out of it, so there is no reason to reject one.  Running
mailsplit on such a file feels stupid, but it does not hurt.

This builds on top of a5a6755 (git-am foreign patch support: introduce
patch_format, 2009-05-27) that introduced mailbox format detection.  The
codepath to deal with a mbox requires it to begin with "From " line and
also allows it to begin with "From: ", but a random piece of e-mail can
and often do begin with any valid RFC2822 header lines.

Instead of checking the first line, we extract all the lines up to the
first empty line, and make sure they look like e-mail headers.

A test is added to t4150 to demonstrate this feature.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-ls-files.txt: clarify what "other files" mean for --other
Matthieu Moy [Thu, 6 Aug 2009 15:11:48 +0000 (17:11 +0200)]
git-ls-files.txt: clarify what "other files" mean for --other

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-describe: Die early if there are no possible descriptions
Björn Steinbrink [Thu, 6 Aug 2009 12:15:14 +0000 (14:15 +0200)]
git-describe: Die early if there are no possible descriptions

If we find no refs that may be used for git-describe with the current
options, then die early instead of pointlessly walking the whole
history.

In git.git with all the tags dropped, this makes "git describe" go down
from 0.244 to 0.003 seconds for me. This is especially noticeable with
"git submodule status" which calls describe with increasing levels of
allowed refs to be matched. For a submodule without tags, this means
that it walks the whole history in the submodule twice (first annotated,
then plain tags), just to find out that it can't describe the commit
anyway.

Signed-off-by: Björn Steinbrink <B.Steinbrink@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agorebase: consistent error messages for staged and unstaged changes.
Matthieu Moy [Wed, 5 Aug 2009 14:56:25 +0000 (16:56 +0200)]
rebase: consistent error messages for staged and unstaged changes.

Previous version expose the output of the plumbing update-index to the
user, which novice users have difficulty to understand.

We still need to run update-index to refresh the cache (if
diff.autorefreshindex is false, git diff won't do it).

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-tag(1): Refer to git-check-ref-format(1) for <name>
Jari Aalto [Wed, 5 Aug 2009 21:20:32 +0000 (00:20 +0300)]
git-tag(1): Refer to git-check-ref-format(1) for <name>

Explain briefly what characters are prohibited in tag <name>
and point to git-check-ref-format(1) manual page for
further information.

Signed-off-by: Jari Aalto <jari.aalto@cante.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-rev-list.txt: Clarify the use of multiple revision arguments
Junio C Hamano [Wed, 5 Aug 2009 16:42:33 +0000 (09:42 -0700)]
git-rev-list.txt: Clarify the use of multiple revision arguments

If one thinks of a revision as the set of commits which can be reached
from the rev, and of ^rev as the complement, then multiple arguments to
git rev-list can be neither understood as the intersection nor the union
of the individual sets.

But set language is the natural as well as logical language in which to
phrase this. So, add a paragraph which explains multiple arguments using
set language.

Suggested-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit.el: Clarify documentation of git-commit-tree
David Kågedal [Fri, 31 Jul 2009 07:23:09 +0000 (09:23 +0200)]
git.el: Clarify documentation of git-commit-tree

Signed-off-by: David Kågedal <davidk@lysator.liu.se>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agodon't let the delta cache grow unbounded in 'git repack'
Nicolas Pitre [Wed, 5 Aug 2009 20:55:07 +0000 (16:55 -0400)]
don't let the delta cache grow unbounded in 'git repack'

I have 4GB of RAM on my system which should, in theory, be quite enough
to repack a 600 MB repository.  However the unbounded delta cache size
always pushes it into swap, at which point everything virtually comes to
a halt.  So unbounded caches are never a good idea.

A default of 256MB should be a good compromize between memory usage and
speed where medium sized repositories are still likely to fit in the
cache with a reasonable memory usage, and larger repositories are going
to take quite some time to repack already anyway.

While at it, clarify the associated config variable documentation
entries a bit.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agotransport: don't show push status if --quiet is given
Jeff King [Wed, 5 Aug 2009 20:23:26 +0000 (16:23 -0400)]
transport: don't show push status if --quiet is given

When --quiet is given, the user generally only wants to see
errors. So let's suppress printing the ref status table
unless there is an error, in which case we print out the
whole table.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agotransport: pass "quiet" flag to pack-objects
Jeff King [Wed, 5 Aug 2009 20:22:36 +0000 (16:22 -0400)]
transport: pass "quiet" flag to pack-objects

When pushing over the git protocol, pack-objects gives
progress reports about the pack being sent. If "push" is
given the --quiet flag, it now passes "-q" to pack-objects,
suppressing this output.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopush: add --quiet flag
Jeff King [Wed, 5 Aug 2009 20:22:19 +0000 (16:22 -0400)]
push: add --quiet flag

Some transports produce output even without "--verbose"
turned on. This provides a way to tell them to be more
quiet (whereas simply redirecting might lose error
messages).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'sb/read-tree'
Junio C Hamano [Wed, 5 Aug 2009 19:40:07 +0000 (12:40 -0700)]
Merge branch 'sb/read-tree'

* sb/read-tree:
  read-tree: migrate to parse-options
  read-tree: convert unhelpful usage()'s to helpful die()'s

15 years agoMerge branch 'jc/apply-epoch-patch'
Junio C Hamano [Wed, 5 Aug 2009 19:40:00 +0000 (12:40 -0700)]
Merge branch 'jc/apply-epoch-patch'

* jc/apply-epoch-patch:
  apply: notice creation/removal patches produced by GNU diff

15 years agoMerge branch 'sb/parse-options'
Junio C Hamano [Wed, 5 Aug 2009 19:39:37 +0000 (12:39 -0700)]
Merge branch 'sb/parse-options'

* sb/parse-options:
  prune-packed: migrate to parse-options
  verify-pack: migrate to parse-options
  verify-tag: migrate to parse-options
  write-tree: migrate to parse-options

15 years agoMerge branch 'ns/init-mkdir'
Junio C Hamano [Wed, 5 Aug 2009 19:39:33 +0000 (12:39 -0700)]
Merge branch 'ns/init-mkdir'

* ns/init-mkdir:
  git init: optionally allow a directory argument

Conflicts:
builtin-init-db.c

15 years agoMerge branch 'mk/init-db-parse-options'
Junio C Hamano [Wed, 5 Aug 2009 19:39:06 +0000 (12:39 -0700)]
Merge branch 'mk/init-db-parse-options'

* mk/init-db-parse-options:
  init-db: migrate to parse-options

15 years agoMerge branch 'jk/maint-show-tag'
Junio C Hamano [Wed, 5 Aug 2009 19:38:54 +0000 (12:38 -0700)]
Merge branch 'jk/maint-show-tag'

* jk/maint-show-tag:
  show: add space between multiple items
  show: suppress extra newline when showing annotated tag

15 years agoMerge branch 'sb/maint-pull-rebase'
Junio C Hamano [Wed, 5 Aug 2009 19:38:39 +0000 (12:38 -0700)]
Merge branch 'sb/maint-pull-rebase'

* sb/maint-pull-rebase:
  pull: support rebased upstream + fetch + pull --rebase
  t5520-pull: Test for rebased upstream + fetch + pull --rebase

15 years agoMerge branch 'ne/futz-upload-pack'
Junio C Hamano [Wed, 5 Aug 2009 19:38:29 +0000 (12:38 -0700)]
Merge branch 'ne/futz-upload-pack'

* ne/futz-upload-pack:
  Shift object enumeration out of upload-pack

Conflicts:
upload-pack.c

15 years agoMerge branch 'maint'
Junio C Hamano [Wed, 5 Aug 2009 19:37:40 +0000 (12:37 -0700)]
Merge branch 'maint'

* maint:
  gitweb/README: Document $base_url
  Documentation: git submodule: add missing options to synopsis
  Better usage string for reflog.
  hg-to-git: don't import the unused popen2 module
  send-email: remove debug trace
  config: Keep inner whitespace verbatim

15 years agoMerge branch 'maint-1.6.3' into maint
Junio C Hamano [Wed, 5 Aug 2009 19:37:24 +0000 (12:37 -0700)]
Merge branch 'maint-1.6.3' into maint

* maint-1.6.3:
  Better usage string for reflog.
  hg-to-git: don't import the unused popen2 module
  send-email: remove debug trace
  config: Keep inner whitespace verbatim

15 years agogitweb: parse_commit_text encoding fix
Zoltán Füzesi [Sun, 2 Aug 2009 07:42:24 +0000 (09:42 +0200)]
gitweb: parse_commit_text encoding fix

Call to_utf8 when parsing author and committer names, otherwise they will appear
with bad encoding if they written by using chop_and_escape_str.

Signed-off-by: Zoltán Füzesi <zfuzesi@eaglet.hu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogitweb/README: Document $base_url
Jakub Narebski [Tue, 4 Aug 2009 15:54:32 +0000 (17:54 +0200)]
gitweb/README: Document $base_url

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: git submodule: add missing options to synopsis
Jens Lehmann [Sat, 1 Aug 2009 18:49:47 +0000 (20:49 +0200)]
Documentation: git submodule: add missing options to synopsis

The option --merge was missing for submodule update and --cached for
submodule summary.

Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-merge-base/git-show-branch --merge-base: Documentation and test
Michael J Gruber [Wed, 5 Aug 2009 07:59:20 +0000 (09:59 +0200)]
git-merge-base/git-show-branch --merge-base: Documentation and test

Currently, the documentation suggests that 'git merge-base -a' and 'git
show-branch --merge-base' are equivalent (in fact it claims that the
former cannot handle more than two revs).

Alas, the handling of more than two revs is very different. Document
this by tests and correct the documentation to reflect this.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-merge-base/git-show-branch: Cleanup documentation and usage
Michael J Gruber [Wed, 5 Aug 2009 07:59:19 +0000 (09:59 +0200)]
git-merge-base/git-show-branch: Cleanup documentation and usage

Make sure that usage strings and documentation coincide with each other
and with the actual code.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agot6010-merge-base.sh: Depict the octopus test graph
Michael J Gruber [Wed, 5 Aug 2009 07:59:18 +0000 (09:59 +0200)]
t6010-merge-base.sh: Depict the octopus test graph

...so that it is easier to reuse it for other tests.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoBetter usage string for reflog.
Matthieu Moy [Wed, 5 Aug 2009 15:36:28 +0000 (17:36 +0200)]
Better usage string for reflog.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agohg-to-git: don't import the unused popen2 module
Miklos Vajna [Mon, 3 Aug 2009 16:41:34 +0000 (18:41 +0200)]
hg-to-git: don't import the unused popen2 module

Importing the popen2 module in Python-2.6 results in the
"DeprecationWarning: The popen2 module is deprecated.  Use the
subprocess module." message. The module itself isn't used in fact, so
just removing it solves the problem.

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agosend-email: remove debug trace
Erik Faye-Lund [Tue, 4 Aug 2009 21:57:34 +0000 (21:57 +0000)]
send-email: remove debug trace

Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
15 years agorun-command.c: squelch a "use before assignment" warning
David Soria Parra [Tue, 4 Aug 2009 09:28:40 +0000 (11:28 +0200)]
run-command.c: squelch a "use before assignment" warning

i686-apple-darwin9-gcc-4.0.1 (GCC) 4.0.1 (Apple Inc. build 5490) compiler
(and probably others) mistakenly thinks variable failed_errno is used
before assigned.  Work it around by giving it a fake initialization.

Signed-off-by: David Soria Parra <dsp@php.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogitweb: fix 'Use of uninitialized value' error in href()
Giuseppe Bilotta [Fri, 31 Jul 2009 06:48:49 +0000 (08:48 +0200)]
gitweb: fix 'Use of uninitialized value' error in href()

Equality between file_parent and file_name was being checked without a
preliminary check for existence of the parameters.

Fix by wrapping the equality check in appropriate if (defined ...),
rearranging the lines to prevent excessive length.

Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
Acked-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agotechnical-docs: document tree-walking API
Stephen Boyd [Tue, 4 Aug 2009 04:13:21 +0000 (21:13 -0700)]
technical-docs: document tree-walking API

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoFix typos on pt_BR/gittutorial.txt translation
André Goddard Rosa [Fri, 31 Jul 2009 16:50:50 +0000 (13:50 -0300)]
Fix typos on pt_BR/gittutorial.txt translation

With extra fixes from Thadeu and Carlos as well.

Signed-off-by: André Goddard Rosa <andre.goddard@gmail.com>
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@holoscopio.com>
Signed-off-by: Carlos R. Mafra <crmafra2@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: git-send-email: correct statement about standard ports
Wesley J. Landaker [Fri, 31 Jul 2009 17:45:00 +0000 (11:45 -0600)]
Documentation: git-send-email: correct statement about standard ports

The current documentation states that servers typically listen on port
465 and calls this "ssmtp". While it's true that many mail servers use
port 465 for SSL smtp, this is non-standard, and hails from the days
before smtp and submission TLS support, that arrived in RFC2487 and
RFC3207. Port 465 is actually assigned by IANA for unrelated purposes,
and is mostly still used by mail servers today only to support Outlook
Express.

In any case, this patch helps the documentation better reflect both
standards and reality, while still helpfully mentioning ports numbers
that a user may wish to specify.

Signed-off-by: Wesley J. Landaker <wjl@icecavern.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDocumentation: git-send-email: fix submission port number
Wesley J. Landaker [Thu, 30 Jul 2009 23:08:53 +0000 (17:08 -0600)]
Documentation: git-send-email: fix submission port number

The current documentation confuses non-standard SSL smtp port 465 with
submission port 587 (RFC 4406). This patch just changes the referenced
number.

Signed-off-by: Wesley J. Landaker <wjl@icecavern.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoconfig: Keep inner whitespace verbatim
Björn Steinbrink [Thu, 30 Jul 2009 11:41:57 +0000 (13:41 +0200)]
config: Keep inner whitespace verbatim

Configuration values are expected to be quoted when they have leading or
trailing whitespace, but inner whitespace should be kept verbatim even if
the value is not quoted. This is already documented in git-config(1), but
the code caused inner whitespace to be collapsed to a single space,
breaking, for example, clones from a path that has two consecutive spaces
in it, as future fetches would only see a single space.

Reported-by: John te Bokkel <tanj.tanj@gmail.com>
Signed-off-by: Björn Steinbrink <B.Steinbrink@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit fast-export: add --no-data option
Geoffrey Irving [Tue, 28 Jul 2009 02:20:22 +0000 (22:20 -0400)]
git fast-export: add --no-data option

When using git fast-export and git fast-import to rewrite the history
of a repository with large binary files, almost all of the time is
spent dealing with blobs.  This is extremely inefficient if all we want
to do is rewrite the commits and tree structure.  --no-data skips the
output of blobs and writes SHA-1s instead of marks, which provides a
massive speedup.

Signed-off-by: Geoffrey Irving <irving@naml.us>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoparse-opt: optionally show "--no-" option string
Johannes Schindelin [Mon, 27 Jul 2009 18:49:56 +0000 (20:49 +0200)]
parse-opt: optionally show "--no-" option string

It is usually better to have positive options, to avoid confusing double
negations.  However, sometimes it is desirable to show the negative option
in the help.

Introduce the flag PARSE_OPT_NEGHELP to do that.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agomerge-recursive: don't segfault while handling rename clashes
Junio C Hamano [Fri, 31 Jul 2009 00:38:15 +0000 (17:38 -0700)]
merge-recursive: don't segfault while handling rename clashes

When a branch moves A to B while the other branch created B (or moved C to
B), the code tried to rename one of them to B~something to preserve both
versions, and failed to register temporary resolution for the original
path B at stage#0 during virtual ancestor computation.  This left the
index in unmerged state and caused a segfault.

A better solution is to merge these two versions of B's in place and use
the (potentially conflicting) result as the intermediate merge result in
the virtual ancestor.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-checkout: be careful about untracked symlinks
Linus Torvalds [Thu, 30 Jul 2009 03:22:25 +0000 (20:22 -0700)]
git-checkout: be careful about untracked symlinks

This fixes the case where an untracked symlink that points at a directory
with tracked paths confuses the checkout logic, demostrated in t6035.

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