CVS username is generated from local part email address.
We take the whole local part but restrict the character set to the
Portable Filename Character Set, which is used for Unix login names
according to Single Unix Specification v3.
This will obviously report different usernames from existing repositories
for commits with the local part of the author e-mail address that contains
characters outside the PFCS. Hopefully this won't break an old CVS
checkout from an earlier version of git-cvsserver, because the names are
always shown afresh to the CVS clients and not kept on the client side.
Signed-off-by: Fabian Emmes <fabian.emmes@rwth-aachen.de> Signed-off-by: Lars Noschinski <lars@public.noschinski.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
show <tag>: reuse pp_user_info() instead of duplicating code
We used to extract the tagger information "by hand" in "git show <tag>",
but the function pp_user_info() already does that. Even better:
it respects the commit_format and date_format specified by the user.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
It comes quite as a surprise to an unsuspecting Git user that calling
"git add submodule/file" (which is a mistake, alright) _removes_
the submodule in the index, and adds the file. Instead, complain loudly.
While at it, be nice when the user said "git add submodule/" which is
most likely the consequence of tab-completion, and stage the submodule,
instead of trying to add the contents of that directory.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
bundle: allow rev-list options to exclude annotated tags
With options such as "--all --since=2.weeks.ago", annotated tags used to
be included, when they should have been excluded. The reason is that we
heavily abuse the revision walker to determine what needs to be included
or excluded. And the revision walker does not show tags at all (and
therefore never marks tags as uninteresting).
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since the OPML project list view was hand-coding the RSS and HTML URLs,
it didn't respect global options such as use_pathinfo. Make it use
href() to ensure consistency with the rest of the gitweb setup.
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>
cvsserver annotates each commit message by "via git-CVS emulator". This is
made configurable via gitcvs.commitmsgannotation.
Signed-off-by: Fabian Emmes <fabian.emmes@rwth-aachen.de> Signed-off-by: Lars Noschinski <lars@public.noschinski.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git checkout: do not allow switching to a tree-ish that is not a commit
"git checkout -b newbranch $commit^{tree}" mistakenly created a new branch
rooted at the current HEAD, because in that case, the two structure fields
used to see if the command was invoked without any argument (hence it
needs to default to checking out the HEAD) were populated incorrectly.
Upon seeing a command line argument that we took as a rev, we should store
that string in new.name, even if that does not name a commit. This will
correctly trigger the existing safety logic.
Signed-off-by: Junio C Hamano <gitster@pobox.com> Acked-by: Daniel Barkalow <barkalow@iabervon.org>
A git patch that does not change the executable bit records the mode bits
on its "index" line. "git apply" used to interpret this mode exactly the
same way as it interprets the mode recorded on "new mode" line, as the
wish by the patch submitter to set the mode to the one recorded on the
line.
The reason the mode does not agree between the submitter and the receiver
in the first place is because there is _another_ commit that only appears
on one side but not the other since their histories diverged, and that
commit changes the mode. The patch has "index" line but not "new mode"
line because its change is about updating the contents without affecting
the mode. The application of such a patch is an explicit wish by the
submitter to only cherry-pick the commit that updates the contents without
cherry-picking the commit that modifies the mode. Viewed this way, the
current behaviour is problematic, even though the command does warn when
the mode of the path being patched does not match this mode, and a careful
user could detect this inconsistencies between the patch submitter and the
patch receiver.
This changes the semantics of the mode recorded on the "index" line;
instead of interpreting it as the submitter's wish to set the mode to the
recorded value, it merely informs what the mode submitter happened to
have, and the presense of the "index" line is taken as submitter's wish to
keep whatever the mode is on the receiving end.
This is based on the patch originally done by Alexander Potashev with a
minor fix; the tests are mine.
* cb/mergetool:
mergetool: Don't keep temporary merge files unless told to
mergetool: Add prompt to continue after failing to merge a file
Add -y/--no-prompt option to mergetool
Fix some tab/space inconsistencies in git-mergetool.sh
objects to be pruned immediately don't have to be loosened
When there is no grace period before pruning unreferenced objects, it is
pointless to push those objects in their loose form just to delete them
right away.
Also be more explicit about the possibility of using "now" in the
gc.pruneexpire config variable (needed for the above behavior to
happen).
Signed-off-by: Nicolas Pitre <nico@cam.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The upstream branch <upstream> now defaults to the first tracked
remote branch, which is set by the configuration variables
branch.<name>.remote and branch.<name>.merge of the current branch.
Without such a remote branch, the command "git cherry [-v]" fails with
usage output as before and an additional message.
Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
builtin-shortlog.c: use string_list_append(), and don't strdup unnecessarily
Make insert_one_record() use string_list_append(), instead of duplicating
its code. Because of this, do not free the "util" member when clearing the
"onelines" string lists: with the new code path it is not initialized to
any value (was being initialized to NULL previously).
Also, avoid unnecessary strdup() calls when inserting names in log->list.
This list always has "strdup_strings" activated, hence strdup'ing namebuf is
unnecessary. This change also removes a latent memory leak in the old code.
NB: The duplicated code mentioned above predated the appearance of
string_list_append().
Signed-off-by: Adeodato Simó <dato@net.com.org.es> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* np/auto-thread:
Force t5302 to use a single thread
pack-objects: don't use too many threads with few objects
autodetect number of CPUs by default when using threads
Instead of listing short option (e.g. "-U<n>") as a shorthand for its
longer counterpart (e.g. "--unified=<n>"), list the synonyms together. It
saves one indirection to find what the reader wants.
Signed-off-by: jidanni <jidanni@jidanni.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This variable was added in 5f8b9fc (git-send-email: add a new
sendemail.cc configuration variable, 2008-04-27), but is not yet refered
to by the documentation.
Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff: add option to show context between close hunks
Merge two hunks if there is only the specified number of otherwise unshown
context between them. For --inter-hunk-context=1, the resulting patch has
the same number of lines but shows uninterrupted context instead of a
context header line in between.
Patches generated with this option are easier to read but are also more
likely to conflict if the file to be patched contains other changes.
This patch keeps the default for this option at 0. It is intended to just
make the feature available in order to see its advantages and downsides.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
"makeinfo" failed to generate gitman.info from gitman.texi input file
because the combined manual page file contains several nodes with the
same name (DESCRIPTION, OPTIONS, SEE ALSO etc.). An Info document should
contain unique node names.
This patch creates a simple (read: ugly) work-around by suppressing the
validation of the final Info file. Jumping to nodes in the Info document
still works but they are not very useful. Common man-page headings like
DESCRIPTION and OPTIONS appear in the Info node list and they point to
the man page where they appear first (that is git-add currently).
Also, this patch adds directory-entry information for Info document to
make the document appear in the top-level Info directory.
Signed-off-by: Teemu Likonen <tlikonen@iki.fi> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix the building of user-manual.texi and gitman.texi documents
Previously "docbook2x-texi" failed to generate user-manual.texi and
gitman.texi files from .xml input files because "iconv" stopped at
"illegal input sequence" error. This was due to some UTF-8 octets in the
input .xml files. This patch adds option --encoding=UTF-8 for
"docbook2x-texi" to allow the building of .texi files complete.
Signed-off-by: Teemu Likonen <tlikonen@iki.fi> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When $filter was empty, the path passed to check_export_ok would
contain an extra '/', which some implementations of export_auth_hook
are sensitive to.
It makes more sense to fix this here than to handle the special case
in each implementation of export_auth_hook.
Signed-off-by: Devin Doucette <devin@doucette.cc> Acked-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
grep: grep cache entries if they are "assume unchanged"
"Assume unchanged" bit means "please pretend that I have never touched
this file", so if user removes the file, we should not care.
This patch teaches "git grep" to use cache version in such
situations. External grep case has not been fixed yet. But given that
on the platform that CE_VALID bit may be used like Windows, external
grep is not available anyway, I would wait for people to raise their
hands before touching it.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
git-send-email.txt: move --format-patch paragraph to a proper location
git-shortlog.txt: improve documentation about .mailmap files
pretty: support multiline subjects with format:
pretty: factor out format_subject()
pretty: factor out skip_empty_lines()
merge-file: handle freopen() failure
daemon: cleanup: factor out xstrdup_tolower()
daemon: cleanup: replace loop with if
daemon: handle freopen() failure
describe: Avoid unnecessary warning when using --all
git-shortlog.txt: improve documentation about .mailmap files
The description on .mailmap made it seem like they are only useful for
commits with a wrong address for an author, but they are about fixing the
real name. Explain this better in the text, and replace the existing
example with a new one that hopefully makes things clearer.
Signed-off-by: Adeodato Simó <dato@net.com.org.es> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git log --pretty=format:%s (and tformat:) used to display the first
line of the subject, unlike the other --pretty options, which would
construct a subject line from all lines of the first paragraph of
the commit message.
For consistency and increased code reuse, change format: to do the
same as the other options.
In the version that was factored out, we can't rely on the len of the
struct strbuf to find out if a line separator needs to be added, as
it might already contain something. Add a guard variable ("first")
instead.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Replace a loop around an enter_repo() call, which was used to retry
a single time with a different parameter in case the first call fails,
with two calls and an if. This is shorter and cleaner.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
describe: Avoid unnecessary warning when using --all
In 212945d4 ("Teach git-describe to verify annotated tag names
before output") git-describe learned how to output a warning if
an annotated tag object was matched but its internal name doesn't
match the local ref name.
However, "git describe --all" causes the local ref name to be
prefixed with "tags/", so we need to skip over this prefix before
comparing the local ref name with the name recorded inside of the
tag object.
Patch-by: René Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The established behaviour of "git ls-tree $tree_ish" run from a subdirectory
"sub/dir" in a work tree is to limit the output to the paths in the
subdirectory, and strip off the leading "sub/dir" from the output, since 3c5e846 (ls-tree: major rewrite to do pathspec, 2005-11-26).
This was a "usability" feature made back in the days when the line between
Porcelain and plumbing was blurry, and in retrospect, it probably was
misguided. The behaviour may be what the end user would expect when the
command is run interactively from a subdirectory, but it also means that a
scripted Porcelain that wants to use the command to list the full contents
of a tree object has to do cd_to_toplevel (and save the output from
"rev-parse --show-prefix" before doing so, so that it can be used as a
pathspec if it wants to limit its operation to the original subdirectory
in other commands).
This new option makes the command operate on the full tree object,
regardless of where in the work tree it is run from. It also implies the
behaviour that is triggered by the existing --full-name option.
merge-recursive: mark rename/delete conflict as unmerged
When a file was renamed in one branch, but deleted in the other, one
should expect the index to contain an unmerged entry, namely the
target of the rename. Make it so.
Noticed by Constantine Plotnikov.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
http-push: support full URI in handle_remote_ls_ctx()
The program calls remote_ls() to get list of files from the server over
HTTP; handle_remote_ls_ctx() is used to parse its response to populate
"struct remote_ls_ctx" that is returned from remote_ls().
The handle_remote_ls_ctx() function assumed that the server returns a
local path in href field, but RFC 4918 (14.7) demand of support full URI
(e.g. "http://localhost:8080/repo.git").
This resulted in push failure (e.g. git-http-push issues a PROPFIND
request to "/repo.git/alhost:8080/repo.git/refs/" to the server).
Signed-off-by: Kirill A. Korinskiy <catap@catap.ru> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* js/rebase-i-p:
rebase -i -p: leave a --cc patch when a merge could not be redone
rebase -i -p: Fix --continue after a merge could not be redone
Show a failure of rebase -p if the merge had a conflict
git-revert: record the parent against which a revert was made
As described in Documentation/howto/revert-a-faulty-merge.txt, re-merging
from a previously reverted a merge of a side branch may need a revert of
the revert beforehand. Record against which parent the revert was made in
the commit, so that later the user can figure out what went on.
Signed-off-by: Robin Rosenberg <robin.rosenberg@dewire.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* git://git.kernel.org/pub/scm/gitk/gitk:
gitk: Force the focus to the main window on Windows
gitk: Allow unbalanced quotes/braces in commit headers
gitk: Update German translation
gitk: Mark forgotten strings (header sentence parts in color chooser) for translation
gitk: Ensure that "Reset branch" menu entry is enabled
gitk: Use check-buttons' -text property instead of separate labels
gitk: Map / to focus the search box
gitk: Fix bugs in blaming code
gitk: Force the focus to the main window on Windows
On msysGit, the focus is first on the (Tk) console. This console is then
hidden, but keeps the focus. Work around that by forcing the focus onto
the gitk window.
This fixes msysGit issue 14. Diagnosed and originally fixed by
Johannes Schindelin.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
gitk: Allow unbalanced quotes/braces in commit headers
When parsing commits, gitk treats the headers of the commit as tcl
lists. This causes errors if the header contains an unbalanced quote
or open brace. Splitting the line on spaces allows us to treat it as
a set of words instead of as a tcl list, which prevents errors.
Signed-off-by: Kevin Ballard <kevin@sb.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
gitk: Ensure that "Reset branch" menu entry is enabled
Consider this sequence of events:
1. Detach HEAD and fire up gitk
2. Call the context menu on some commit. Notice that the last menu entry
says "Detached HEAD: can't reset" and it is disabled.
3. Now checkout some regular branch (e.g. 'master') using the context menu.
4. Call the context menu again on some commit.
Previously, at this point the last menu entry said "Reset master branch
to here", but it was still disabled. With this fix it is now enabled again.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
gitk: Use check-buttons' -text property instead of separate labels
Previously the check-buttons' labels in the Preferences were separate
widgets. This had the disadvantage that in order to toggle the
check-button with the mouse the check-box had to be clicked. With
this change the check-box can also be toggled by clicking the label.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
The / key is often used to initiate searches (less, vim, some web
browsers). This changes the binding for the / (slash) key from 'find
next' to 'focus the search box' to follow this convention.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
doc/git-fsck: change the way for getting heads' SHA1s
The straightforward way with using 'cat .git/refs/heads/*' doesn't work
with packed refs as well as branches of the form topic/topic1. So let's
use git-for-each-ref for getting the heads' SHA1s in this example.
Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/git-show-branch: work around "single quote" typesetting glitch
The displayed example is typeset with acute accents around the string that
should be surrounded by a pair of single quotes in manpage. Replace them
with double quotes (the semantics of the example does not change).
Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Make sure lockfiles are unlocked when dying on SIGPIPE
We cleaned up lockfiles upon receiving the usual suspects HUP, TERM, QUIT
but a wicked user could kill us of asphyxiation by piping our output to a
pipe that does not read. Protect ourselves by catching SIGPIPE and clean
up the lockfiles as well in such a case.
connect.c: stricter port validation, silence compiler warning
In addition to checking if the provided port is numeric, also check
that the string isn't empty and that the port number is within the
valid range. Incidentally, this silences a compiler warning about
ignoring strtol's return value.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Linus and Junio explained issues that are involved in reverting a merge
and how to continue working with a branch that was updated since such a
revert on the mailing list. This is to help new people who did not see
these messages.
Signed-off-by: Nanako Shiraishi <nanako3@lavabit.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
- call lock_remote() to issue a DAV_LOCK request to the server to lock
info/refs and branch refs being pushed into; handle_new_lock_ctx() is
used to parse its response to populate "struct remote_lock" that is
returned from lock_remote();
- send objects;
- call unlock_remote() to drop the lock.
The handle_new_lock_ctx() function assumed that the server will use a
lock token in opaquelocktoken URI scheme, which may have been an Ok
assumption under RFC 2518, but under RFC 4918 which obsoletes the older
standard it is not necessarily true.
This resulted in push failure (often resulted in "cannot lock existing
info/refs" error message) when talking to a server that does not use
opaquelocktoken URI scheme.
Signed-off-by: Kirill A. Korinskiy <catap@catap.ru> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-sh-setup: Fix scripts whose PWD is a symlink into a git work-dir
I want directories of my working tree to be linked to from various
paths on my filesystem where third-party components expect them, both
in development and production environments. A build system's install
step could solve this, but I develop scripts and web pages that don't
need to be built. Git's submodule system could solve this, but we
tend to develop, branch, and test those directories all in unison, so
one big repository feels more natural. We prefer to edit and commit
on the symlinked paths, not the canonical ones, and in that setting,
"git pull" fails to find the top-level directory of the repository
while other commands work fine.
"git pull" fails because POSIX shells have a notion of current working
directory that is different from getcwd(). The shell stores this path
in PWD. As a result, "cd ../" can be interpreted differently in a
shell script than chdir("../") in a C program. The shell interprets
"../" by essentially stripping the last textual path component from
PWD, whereas C chdir() follows the ".." link in the current directory
on the filesystem. When PWD is a symlink, these are different
destinations. As a result, Git's C commands find the correct
top-level working tree, and shell scripts do not.
Changes:
* When interpreting a relative upward (../) path in cd_to_toplevel,
prepend the cwd without symlinks, given by /bin/pwd
* Add tests for cd_to_toplevel and "git pull" in a symlinked
directory that failed before this fix, plus contrasting scenarios
that already worked
Signed-off-by: Marcel M. Cary <marcel@oak.homeunix.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
rebase -i -p: Fix --continue after a merge could not be redone
When a merge that has a conflict was rebased, then rebase stopped to let
the user resolve the conflicts. However, thereafter --continue failed
because the author-script was not saved. (This is rebase -i's way to
preserve a commit's authorship.) This fixes it by doing taking the same
failure route after a merge that is also taken after a normal cherry-pick.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Show a failure of rebase -p if the merge had a conflict
This extends t3409-rebase-preserve-merges by a case where the merge that
is rebased has a conflict. Therefore, the rebase stops and expects that
the user resolves the conflict. However, currently rebase --continue
fails because .git/rebase-merge/author-script is missing.
The test script had allocated two identical clones, but only one of them
(clone2) was used. Now we use both as indicated in the comment. Also,
two instances of && was missing in the setup part.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Even though newer Porcelain tools always record the tagger information
when creating new tags, export/import pair should be able to faithfully
reproduce ancient tag objects that lack tagger information.
Signed-off-by: Junio C Hamano <gitster@pobox.com> Acked-by: Shawn O. Pearce <spearce@spearce.org>
SubmittingPatches: mention the usage of real name in Signed-off-by: lines
Especially with something that is supposed to hopefully have some legal
value down the line if somebody starts making noises, it really would be
nice to have a real person to associate things with. Suggest this in the
SubmittingPatches document.
Signed-off-by: Miklos Vajna <vmiklos@frugalware.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Like many git commands, git-mergetool allows "--" to signal
the end of option processing. This adds a missing "shift"
statement so that this is correctly handled.
Signed-off-by: David Aguilar <davvid@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a test which checks that negated patterns such as "!foo.html" can
override previous patterns such as "*.html". This is documented
behaviour but had not been tested so far.
Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* lt/readlink:
combine-diff.c: use strbuf_readlink()
builtin-blame.c: use strbuf_readlink()
make_absolute_path(): check bounds when seeing an overlong symlink
Make 'prepare_temp_file()' ignore st_size for symlinks
Make 'diff_populate_filespec()' use the new 'strbuf_readlink()'
Make 'index_path()' use 'strbuf_readlink()'
Make 'ce_compare_link()' use the new 'strbuf_readlink()'
Add generic 'strbuf_readlink()' helper function
but the DESCRIPTION says that this form checks the paths out "from the
index, or from a named commit." A later sentence refers to the same
argument as "<tree-ish> argument", but it is not clear that these two
sentences are talking about the same command line argument for first-time
readers.
Signed-off-by: Nanako Shiraishi <nanako3@lavabit.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* git://repo.or.cz/git-gui:
git-gui 0.12
git-gui: Get rid of the last remnants of GIT_CONFIG_LOCAL
git-gui: Update Hungarian translation for 0.12
git-gui: Fixed typos in Swedish translation.
git-gui: Updated Swedish translation (515t0f0u).
git gui: update Italian translation
git-gui: Update Japanese translation for 0.12
git-gui: Starting translation for Norwegian
git-gui: Update German (completed) translation.
git-gui: Update po template to include 'Mirroring %s' message
git-gui: Fix commit encoding handling.
git-gui: Fix handling of relative paths in blame.
githooks documentation: add a note about the +x mode
In a freshly initialized repo it is only necessary to rename the .sample
hooks, but when using older repos (initialized with older git init)
enabled the +x mode is still necessary - docuement this.
Signed-off-by: Miklos Vajna <vmiklos@frugalware.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>