git
16 years agoMerge branch 'js/pick-root'
Junio C Hamano [Sun, 13 Jul 2008 22:15:13 +0000 (15:15 -0700)] 
Merge branch 'js/pick-root'

* js/pick-root:
  Allow cherry-picking root commits

16 years agoMerge branch 'ab/bundle'
Junio C Hamano [Sun, 13 Jul 2008 22:15:08 +0000 (15:15 -0700)] 
Merge branch 'ab/bundle'

* ab/bundle:
  Teach git-bundle to read revision arguments from stdin like git-rev-list.

16 years agoMerge branch 'tr/add-i-e'
Junio C Hamano [Sun, 13 Jul 2008 22:14:59 +0000 (15:14 -0700)] 
Merge branch 'tr/add-i-e'

* tr/add-i-e:
  git-add--interactive: manual hunk editing mode
  git-add--interactive: remove hunk coalescing
  git-add--interactive: replace hunk recounting with apply --recount

16 years agoapply: fix copy/rename breakage
Junio C Hamano [Thu, 10 Jul 2008 02:58:23 +0000 (19:58 -0700)] 
apply: fix copy/rename breakage

7ebd52a (Merge branch 'dz/apply-again', 2008-07-01) taught "git-apply" to
grok a (non-git) patch that is a concatenation of separate patches that
touch the same file number of times, by recording the postimage of patch
application of previous round and using it as the preimage for later
rounds.

This "incremental" mode of patch application fundamentally contradicts
with the way git rename/copy patches are designed.  When a git patch talks
about a file A getting modified, and a new file B created out of A, like
this:

diff --git a/A b/A
--- a/A
+++ b/A
... change text here ...
diff --git a/A b/B
copy from A
copy to B
--- a/A
+++ b/B
... change text here ...

the second change to produce B does not depend on what is done to A with
the first change in any way.  This is explicitly done so for reviewability
of individual patches.

With this commit, we do not look at 'fn_table' that records the postimage
of previous round when applying a patch to produce a new file out of an
existing file.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'jk/pager-config'
Junio C Hamano [Wed, 9 Jul 2008 23:58:46 +0000 (16:58 -0700)] 
Merge branch 'jk/pager-config'

* jk/pager-config:
  Allow per-command pager config

16 years agoMerge branch 'js/apply-root'
Junio C Hamano [Wed, 9 Jul 2008 23:58:21 +0000 (16:58 -0700)] 
Merge branch 'js/apply-root'

* js/apply-root:
  git-apply --directory: make --root more similar to GNU diff
  apply --root: thinkofix.
  Teach "git apply" to prepend a prefix with "--root=<root>"

16 years agoMerge branch 'jc/reflog-expire'
Junio C Hamano [Wed, 9 Jul 2008 23:57:35 +0000 (16:57 -0700)] 
Merge branch 'jc/reflog-expire'

* jc/reflog-expire:
  Make default expiration period of reflog used for stash infinite
  Per-ref reflog expiry configuration

16 years agogit-send-email: Fix authenticating on some servers when using TLS.
Robert Shearman [Wed, 9 Jul 2008 21:39:40 +0000 (22:39 +0100)] 
git-send-email: Fix authenticating on some servers when using TLS.

Send HELO again after a successful STARTTLS command to refresh the list of
extensions. These may be different to what is returned over a clear
connection (for example the AUTH command may be accepted over a secure
connection, but not over a clear connection).

Furthermore, this behaviour is recommended by RFC 2487
(http://www.ietf.org/rfc/rfc2487.txt).

Signed-off-by: Robert Shearman <robertshearman@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoProvide fallback definitions of PRIu32 and PRIx32
Johannes Sixt [Wed, 9 Jul 2008 20:38:14 +0000 (22:38 +0200)] 
Provide fallback definitions of PRIu32 and PRIx32

Since 6e1c23442 we make use of these C99 constructs, but this commit did
not provide fallbacks for non-C99 systems.

Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agot9100-git-svn-basic.sh: Fix determination of utf-8 locale
Ramsay Jones [Tue, 8 Jul 2008 22:59:25 +0000 (23:59 +0100)] 
t9100-git-svn-basic.sh: Fix determination of utf-8 locale

When setting the GIT_SVN_LC_ALL variable, default to the $LANG
environment variable, when the $LC_ALL override is not set.

Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agot9113-*.sh: provide user feedback when test skipped
Ramsay Jones [Tue, 8 Jul 2008 23:18:26 +0000 (00:18 +0100)] 
t9113-*.sh: provide user feedback when test skipped

Currently this test simply exits without providing any
feedback at all.  Tell user if the test is being skipped
and provide a hint as to how the test may be enabled.

Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'sp/maint-pack-memuse'
Junio C Hamano [Wed, 9 Jul 2008 21:46:46 +0000 (14:46 -0700)] 
Merge branch 'sp/maint-pack-memuse'

* sp/maint-pack-memuse:
  Correct pack memory leak causing git gc to try to exceed ulimit

Conflicts:

sha1_file.c

16 years agoCorrect pack memory leak causing git gc to try to exceed ulimit
Shawn O. Pearce [Wed, 9 Jul 2008 07:10:07 +0000 (07:10 +0000)] 
Correct pack memory leak causing git gc to try to exceed ulimit

When recursing to unpack a delta base we must unuse_pack() so that
the pack window for the current object does not remain pinned in
memory while the delta base is itself being unpacked and materialized
for our use.

On a long delta chain of 50 objects we may need to access 6 different
windows from a very large (>3G) pack file in order to obtain all
of the delta base content.  If the process ulimit permits us to
map/allocate only 1.5G we must release windows during this recursion
to ensure we stay within the ulimit and transition memory from pack
cache to standard malloc, or other mmap needs.

Inserting an unuse_pack() call prior to the recursion allows us to
avoid pinning the current window, making it available for garbage
collection if memory runs low.

This has been broken since at least before 1.5.1-rc1, and very
likely earlier than that.  Its fixed now.  :)

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'maint'
Junio C Hamano [Wed, 9 Jul 2008 07:19:50 +0000 (00:19 -0700)] 
Merge branch 'maint'

* maint:
  Start preparing release notes for 1.5.6.3
  git-submodule - Fix bugs in adding an existing repo as a module
  bash: offer only paths after '--'
  Remove unnecessary pack-*.keep file after successful git-clone
  make deleting a missing ref more quiet

16 years agoStart preparing release notes for 1.5.6.3
Junio C Hamano [Wed, 9 Jul 2008 06:57:14 +0000 (23:57 -0700)] 
Start preparing release notes for 1.5.6.3

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: update sections on naming revisions and revision ranges
Junio C Hamano [Mon, 7 Jul 2008 21:58:58 +0000 (14:58 -0700)] 
Documentation: update sections on naming revisions and revision ranges

Various *_HEAD pseudo refs were not documented in any central place.
Especially since we may be teaching rebase and am to record ORIG_HEAD,
it would be a good time to do so.

While at it, reword the explanation on r1..r2 notation to reduce
confusion.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoTone down warning about GNU Interactive Tools
Junio C Hamano [Mon, 7 Jul 2008 02:10:00 +0000 (19:10 -0700)] 
Tone down warning about GNU Interactive Tools

The mention of 1997 was correct when it was made, and it still is true
to some extent (http://savannah.gnu.org/forum/forum.php?forum_id=5189
says it has not been actively maintained for quite some time).  However,
because its name changed not to conflict with us, it is no longer
relevant whether many users use gnuit or have moved away to graphical
file managers.

The only people possibly affected are people who have older version of
gnuit installed as "git".

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoAvoid apache complaining about lack of server's FQDN
Mike Hommey [Mon, 7 Jul 2008 20:22:15 +0000 (22:22 +0200)] 
Avoid apache complaining about lack of server's FQDN

On some setups, apache will say:
apache2: Could not reliably determine the server's fully qualified
domain name, using $(IP_address) for ServerName

Avoid this message polluting tests output by setting a ServerName in
apache configuration.

Signed-off-by: Mike Hommey <mh@glandium.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoSkip t5540-http-push test when USE_CURL_MULTI is undefined
Mike Hommey [Mon, 7 Jul 2008 19:02:50 +0000 (21:02 +0200)] 
Skip t5540-http-push test when USE_CURL_MULTI is undefined

When USE_CURL_MULTI is undefined, git http-push doesn't work, so it's
useless to test it.

Signed-off-by: Mike Hommey <mh@glandium.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix http-push test
Mike Hommey [Mon, 7 Jul 2008 19:02:37 +0000 (21:02 +0200)] 
Fix http-push test

http-push test has been broken by 4a7aaccd adding a space character
in the place where the test is being run.

Signed-off-by: Mike Hommey <mh@glandium.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoCatch failures from t5540-http-push
Mike Hommey [Mon, 7 Jul 2008 21:06:46 +0000 (23:06 +0200)] 
Catch failures from t5540-http-push

git http-push doesn't handle packed-refs, and now the new builtin-clone
created packed refs, the http-push test fails.

Mark the current failure as such, and also catch third test's failure
that went unreported because git push doesn't return an error code when
it says:
 No refs in common and none specified; doing nothing.
Which it does when http-push can't get a list of refs recursively from
$URL/refs/.

Signed-off-by: Mike Hommey <mh@glandium.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoGit.pm: Add remote_refs() git-ls-remote frontend
Petr Baudis [Tue, 8 Jul 2008 17:48:04 +0000 (19:48 +0200)] 
Git.pm: Add remote_refs() git-ls-remote frontend

This patch also converts the good ole' git-remote.perl to use it.
It is otherwise used in the repo.or.cz machinery and I guess other
scripts might find it useful too.

Unfortunately,

git-ls-remote --heads .

is subtly different from

git-ls-remote . refs/heads/

(since the second matches anywhere in the string, not just at the
beginning) so we have to provide interface for both.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agot7701-repack-unpack-unreachable.sh: check timestamp of unpacked objects
Brandon Casey [Sun, 29 Jun 2008 00:25:05 +0000 (19:25 -0500)] 
t7701-repack-unpack-unreachable.sh: check timestamp of unpacked objects

Unpacked objects should receive the timestamp of the pack they were
unpacked from. Check.

Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agocompletion: add branch options --contains --merged --no-merged
Eric Raible [Mon, 7 Jul 2008 20:41:54 +0000 (13:41 -0700)] 
completion: add branch options --contains --merged --no-merged

Signed-off-by: Eric Raible <raible@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-submodule - Fix bugs in adding an existing repo as a module
Mark Levedahl [Tue, 8 Jul 2008 02:36:40 +0000 (22:36 -0400)] 
git-submodule - Fix bugs in adding an existing repo as a module

git-submodule add would trip if path to the submodule included a space,
or if its .git was a gitdir: link to a GIT_DIR kept elsewhere. Fix both.

Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoinstall-doc-quick - use git --exec-path to find git-sh-setup
Mark Levedahl [Tue, 8 Jul 2008 02:37:38 +0000 (22:37 -0400)] 
install-doc-quick - use git --exec-path to find git-sh-setup

This is needed as git-sh-setup is no longer in the path.

Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agobash: offer only paths after '--'
SZEDER Gábor [Tue, 8 Jul 2008 16:56:14 +0000 (18:56 +0200)] 
bash: offer only paths after '--'

Many git commands use '--' to separate subcommands, options, and refs
from paths.  However, the programmable completion for several of these
commands does not respect the '--', and offer subcommands, options, or
refs after a '--', although only paths are permitted.  e.g. 'git bisect
-- <TAB>' offers subcommands, 'git log -- --<TAB>' offers options and
'git log -- git<TAB>' offers all gitgui tags.

The completion for the following commands share this wrong behaviour:
  am add bisect commit diff log reset shortlog submodule gitk.

To avoid this, we check the presence of a '--' on the command line first
and let the shell do filename completion, if one is found.

Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoRemove unnecessary pack-*.keep file after successful git-clone
Shawn O. Pearce [Tue, 8 Jul 2008 04:46:06 +0000 (04:46 +0000)] 
Remove unnecessary pack-*.keep file after successful git-clone

Once a clone is successful we no longer need to hold onto the
.keep file created by the transport.  Delete the file so we
can later repack the complete repository.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoavoid null SHA1 in oldest reflog
Jeff King [Tue, 8 Jul 2008 04:38:54 +0000 (00:38 -0400)] 
avoid null SHA1 in oldest reflog

When the user specifies a ref by a reflog entry older than
one we have (e.g., "HEAD@{20 years ago"}), we issue a
warning and give them the "from" value of the oldest reflog
entry. That is, we say "we don't know what happened before
this entry, but before this we know we had some particular
SHA1".

However, the oldest reflog entry is often a creation event
such as clone or branch creation. In this case, the entry
claims that the ref went from "00000..." (the null sha1) to
the new value, and the reflog lookup returns the null sha1.

While this is technically correct (the entry tells us that
the ref didn't exist at the specified time) it is not
terribly useful to the end user. What they probably want
instead is "the oldest useful sha1 that this ref ever had".
This patch changes the behavior such that if the oldest ref
would return the null sha1, it instead returns the first
value the ref ever had.

We never discovered this problem in the test scripts because
we created "fake" reflogs that had only a specified segment
of history. This patch updates the tests with a creation
event at the beginning of history.

Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomake deleting a missing ref more quiet
Jeff King [Tue, 8 Jul 2008 04:08:02 +0000 (00:08 -0400)] 
make deleting a missing ref more quiet

If git attempts to delete a ref, but the unlink of the ref
file fails, we print a message to stderr. This is usually a
good thing, but if the error is ENOENT, then it indicates
that the ref has _already_ been deleted. And since that's
our goal, it doesn't make sense to complain to the user.

This harmonizes the error reporting behavior for the
unpacked and packed cases; the packed case already printed
nothing on ENOENT, but the unpacked printed unconditionally.

Additionally, send-pack would, when deleting the tracking
ref corresponding to a remote delete, print "Failed to
delete" on any failure. This can be a misleading
message, since we actually _did_ delete at the remote side,
but we failed to delete locally. Rather than make the
message more precise, let's just eliminate it entirely; the
delete_ref routine already takes care of printing out a much
more specific message about what went wrong.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'qq/maint' into maint
Junio C Hamano [Tue, 8 Jul 2008 20:05:06 +0000 (13:05 -0700)] 
Merge branch 'qq/maint' into maint

* qq/maint:
  run_command(): respect GIT_TRACE

16 years agoDocumentation: fix broken "linkgit" links
Eric Hanchrow [Tue, 8 Jul 2008 20:02:11 +0000 (13:02 -0700)] 
Documentation: fix broken "linkgit" links

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agouser-manual: typo and grammar fixes
Eric Hanchrow [Tue, 8 Jul 2008 20:00:30 +0000 (13:00 -0700)] 
user-manual: typo and grammar fixes

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'lt/racy-empty' into maint
Junio C Hamano [Tue, 8 Jul 2008 07:19:17 +0000 (00:19 -0700)] 
Merge branch 'lt/racy-empty' into maint

* lt/racy-empty:
  racy-git: an empty blob has a fixed object name

16 years agoMerge branch 'qq/maint'
Junio C Hamano [Mon, 7 Jul 2008 23:42:08 +0000 (16:42 -0700)] 
Merge branch 'qq/maint'

* qq/maint:
  run_command(): respect GIT_TRACE

Conflicts:

run-command.c

16 years agogitweb: Describe projects_index format in more detail
Jakub Narebski [Mon, 7 Jul 2008 22:07:53 +0000 (00:07 +0200)] 
gitweb: Describe projects_index format in more detail

Update and extend information about $projects_list file format in
gitweb/README and in gitweb/INSTALL.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'js/maint-daemon-syslog'
Junio C Hamano [Mon, 7 Jul 2008 23:32:46 +0000 (16:32 -0700)] 
Merge branch 'js/maint-daemon-syslog'

* js/maint-daemon-syslog:
  git daemon: avoid calling syslog() from a signal handler

16 years agoMerge branch 'maint'
Junio C Hamano [Mon, 7 Jul 2008 23:31:55 +0000 (16:31 -0700)] 
Merge branch 'maint'

* maint:
  git-svn.perl: workaround assertions in svn library 1.5.0

16 years agoMerge branch 'qq/maint' (early part) into maint
Junio C Hamano [Mon, 7 Jul 2008 23:09:17 +0000 (16:09 -0700)] 
Merge branch 'qq/maint' (early part) into maint

* 'qq/maint' (early part):
  git-svn.perl: workaround assertions in svn library 1.5.0
  mailinfo: feed the correct line length to decode_transfer_encoding()
  git-clone: remove leftover debugging fprintf().
  Fix "config_error_nonbool" used with value instead of key
  clone -q: honor "quiet" option over native transports.
  attribute documentation: keep EXAMPLE at end
  builtin-commit.c: Use 'git_config_string' to get 'commit.template'
  http.c: Use 'git_config_string' to clean up SSL config.
  diff.c: Use 'git_config_string' to get 'diff.external'
  convert.c: Use 'git_config_string' to get 'smudge' and 'clean'
  builtin-log.c: Use 'git_config_string' to get 'format.subjectprefix' and 'format.suffix'
  Documentation cvs: Clarify when a bare repository is needed
  Documentation: be precise about which date --pretty uses

16 years agoAllow cherry-picking root commits
Johannes Schindelin [Fri, 4 Jul 2008 15:19:52 +0000 (16:19 +0100)] 
Allow cherry-picking root commits

A root commit couldn't be cherry-picked.  But its semantics can be
defined as simply merging two trees by overlaying disjoint parts
and merging overlapping files without any common ancestor.  You
should be able to rebase originally independent branches on top of
another branch by using this.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agorun_command(): respect GIT_TRACE
Johannes Schindelin [Mon, 7 Jul 2008 13:41:34 +0000 (14:41 +0100)] 
run_command(): respect GIT_TRACE

When GIT_TRACE is set, the user is most likely wanting to see an external
command that is about to be executed.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-svn.perl: workaround assertions in svn library 1.5.0
Gerrit Pape [Sun, 6 Jul 2008 19:28:50 +0000 (19:28 +0000)] 
git-svn.perl: workaround assertions in svn library 1.5.0

With subversion 1.5.0 (C and perl libraries) the git-svn selftest
t9101-git-svn-props.sh fails at test 25 and 26.  The following commands
cause assertions in the svn library

 $ cd deeply
 $ git-svn propget svn:ignore .
 perl: /build/buildd/subversion-1.5.0dfsg1/subversion/libsvn_ra/ra_loader.c:674: svn_ra_get_dir: Assertion `*path != '/'' failed.
 Aborted

 $ git-svn propget svn:ignore ..
 perl: /build/buildd/subversion-1.5.0dfsg1/subversion/libsvn_subr/path.c:120: svn_path_join: Assertion `is_canonical(component, clen)' failed.

With this commit, git-svn makes sure the path doesn't start with a
slash, and is not a dot, working around these assertions.

The breakage was reported by Lucas Nussbaum through
 http://bugs.debian.org/489108

Signed-off-by: Gerrit Pape <pape@smarden.org>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoUpdate draft release notes for 1.6.0
Junio C Hamano [Mon, 7 Jul 2008 08:39:28 +0000 (01:39 -0700)] 
Update draft release notes for 1.6.0

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'jc/rerere'
Junio C Hamano [Mon, 7 Jul 2008 09:17:28 +0000 (02:17 -0700)] 
Merge branch 'jc/rerere'

* jc/rerere:
  rerere.autoupdate
  t4200: fix rerere test
  rerere: remove dubious "tail_optimization"
  git-rerere: detect unparsable conflicts
  rerere: rerere_created_at() and has_resolution() abstraction

16 years agoMerge branch 'dr/ceiling'
Junio C Hamano [Mon, 7 Jul 2008 09:17:23 +0000 (02:17 -0700)] 
Merge branch 'dr/ceiling'

* dr/ceiling:
  Eliminate an unnecessary chdir("..")
  Add support for GIT_CEILING_DIRECTORIES
  Fold test-absolute-path into test-path-utils
  Implement normalize_absolute_path

Conflicts:

cache.h
setup.c

16 years agoMerge branch 'db/no-git-config'
Junio C Hamano [Mon, 7 Jul 2008 09:17:14 +0000 (02:17 -0700)] 
Merge branch 'db/no-git-config'

* db/no-git-config:
  Only use GIT_CONFIG in "git config", not other programs

Conflicts:

Documentation/RelNotes-1.6.0.txt

16 years agoMerge branch 'js/import-zip'
Junio C Hamano [Mon, 7 Jul 2008 09:16:55 +0000 (02:16 -0700)] 
Merge branch 'js/import-zip'

* js/import-zip:
  Add another fast-import example, this time for .zip files

16 years agoMerge branch 'maint'
Junio C Hamano [Mon, 7 Jul 2008 09:11:28 +0000 (02:11 -0700)] 
Merge branch 'maint'

* maint:
  Fix grammar in git-rev-parse(1).

16 years agoMerge branch 'qq/maint'
Junio C Hamano [Mon, 7 Jul 2008 09:09:38 +0000 (02:09 -0700)] 
Merge branch 'qq/maint'

* qq/maint:
  mailinfo: feed the correct line length to decode_transfer_encoding()
  git-clone: remove leftover debugging fprintf().

16 years agoTeach git-bundle to read revision arguments from stdin like git-rev-list.
Adam Brewster [Sat, 5 Jul 2008 21:26:40 +0000 (17:26 -0400)] 
Teach git-bundle to read revision arguments from stdin like git-rev-list.

This patch allows the caller to feed the revision parameters to git-bundle
from its standard input.  This way, a script do not have to worry about
limitation of the length of command line.

Documentation/git-bundle.txt says that git-bundle takes arguments acceptable
to git-rev-list.  Obviously some arguments that git-rev-list handles don't
make sense for git-bundle (e.g. --bisect) but --stdin is pretty reasonable.

Signed-off-by: Adam Brewster <adambrewster@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agocompletion.bash: add 'skip' and 'run' to git-bisect
Dmitry Potapov [Wed, 2 Jul 2008 13:29:50 +0000 (17:29 +0400)] 
completion.bash: add 'skip' and 'run' to git-bisect

Signed-off-by: Dmitry Potapov <dpotapov@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomailinfo: feed the correct line length to decode_transfer_encoding()
Junio C Hamano [Mon, 7 Jul 2008 05:26:45 +0000 (22:26 -0700)] 
mailinfo: feed the correct line length to decode_transfer_encoding()

When handling a MIME multipart message, multi-part boundary lines are eaten
by a call to handle_boundary() function from the main loop of handle_body(),
and after that happens, we should update the line length correctly, because
handle_boundary() udpates line[] with new data.

This was caused by a thinko in 9aa2309 (mailinfo: apply the same fix not
to lose NULs in BASE64 and QP codepaths, 2008-05-25).

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-apply --directory: make --root more similar to GNU diff
Junio C Hamano [Mon, 7 Jul 2008 01:36:01 +0000 (18:36 -0700)] 
git-apply --directory: make --root more similar to GNU diff

Applying a patch in the directory that is different from what the patch
records is done with --directory option in GNU diff.  The --root option we
introduced previously does the same, and we can call it the same way to
give users more familiar feel.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-clone: remove leftover debugging fprintf().
Alex Riesen [Sun, 6 Jul 2008 22:56:49 +0000 (00:56 +0200)] 
git-clone: remove leftover debugging fprintf().

Acked-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix grammar in git-rev-parse(1).
Mikael Magnusson [Sun, 6 Jul 2008 20:34:21 +0000 (22:34 +0200)] 
Fix grammar in git-rev-parse(1).

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit daemon: avoid calling syslog() from a signal handler
Johannes Schindelin [Thu, 3 Jul 2008 15:27:24 +0000 (16:27 +0100)] 
git daemon: avoid calling syslog() from a signal handler

Signal handlers should never call syslog(), as that can raise signals
of its own.

Instead, call the syslog() from the master process.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge git://repo.or.cz/git-gui
Junio C Hamano [Sun, 6 Jul 2008 19:55:34 +0000 (12:55 -0700)] 
Merge git://repo.or.cz/git-gui

* git://repo.or.cz/git-gui:
  git-gui: Implement "Stage/Unstage Line"
  git-gui: Don't select the wrong file if the last listed file is staged.
  git-gui: Fix accidental staged state toggle when clicking top pixel row
  git-gui: Move on to the next filename after staging/unstaging a change

16 years agoAllow per-command pager config
Jeff King [Thu, 3 Jul 2008 11:46:57 +0000 (07:46 -0400)] 
Allow per-command pager config

There is great debate over whether some commands should set
up a pager automatically. This patch allows individuals to
set their own pager preferences for each command, overriding
the default. For example, to disable the pager for git
status:

  git config pager.status false

If "--pager" or "--no-pager" is specified on the command
line, it takes precedence over the config option.

There are two caveats:

  - you can turn on the pager for plumbing commands.
    Combined with "core.pager = always", this will probably
    break a lot of things. Don't do it.

  - This only works for builtin commands. The reason is
    somewhat complex:

    Calling git_config before we do setup_git_directory
    has bad side effects, because it wants to know where
    the git_dir is to find ".git/config". Unfortunately,
    we cannot call setup_git_directory indiscriminately,
    because some builtins (like "init") break if we do.

    For builtins, this is OK, since we can just wait until
    after we call setup_git_directory. But for aliases, we
    don't know until we expand (recursively) which command
    we're doing. This should not be a huge problem for
    aliases, which can simply use "--pager" or "--no-pager"
    in the alias as appropriate.

    For external commands, however, we don't know we even
    have an external command until we exec it, and by then
    it is too late to check the config.

    An alternative approach would be to have a config mode
    where we don't bother looking at .git/config, but only
    at the user and system config files. This would make the
    behavior consistent across builtins, aliases, and
    external commands, at the cost of not allowing per-repo
    pager config for at all.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'qq/maint'
Junio C Hamano [Sun, 6 Jul 2008 07:35:13 +0000 (00:35 -0700)] 
Merge branch 'qq/maint'

* qq/maint:
  Fix "config_error_nonbool" used with value instead of key

16 years agohg-to-git: use git init instead of git init-db
Miklos Vajna [Sun, 6 Jul 2008 03:15:20 +0000 (05:15 +0200)] 
hg-to-git: use git init instead of git init-db

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agohg-to-git: rewrite "git-frotz" to "git frotz"
Miklos Vajna [Sun, 6 Jul 2008 03:15:19 +0000 (05:15 +0200)] 
hg-to-git: rewrite "git-frotz" to "git frotz"

This is not just nice but necessary since git-frotz is no longer in
PATH.

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agohg-to-git: abort if the project directory is not a hg repo
Miklos Vajna [Sun, 6 Jul 2008 03:15:18 +0000 (05:15 +0200)] 
hg-to-git: abort if the project directory is not a hg repo

Check the exit code of the first hg command, and abort to avoid a later
ValueError exception.

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agohg-to-git: avoid raising a string exception
Miklos Vajna [Sun, 6 Jul 2008 03:15:17 +0000 (05:15 +0200)] 
hg-to-git: avoid raising a string exception

This fixes the following warning:
hg-to-git.py:92: DeprecationWarning: raising a string exception is deprecated

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoINSTALL: Update section about git-frotz form.
Miklos Vajna [Sun, 6 Jul 2008 04:28:41 +0000 (06:28 +0200)] 
INSTALL: Update section about git-frotz form.

The old text stated that 'git-frotz' can be always used instead of 'git
frotz' which is no longer true.

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix "config_error_nonbool" used with value instead of key
Christian Couder [Sun, 6 Jul 2008 04:10:04 +0000 (06:10 +0200)] 
Fix "config_error_nonbool" used with value instead of key

The function "config_error_nonbool", that is defined in "config.c",
is used to report an error when a config key in the config file
should have a corresponding value but it hasn't.

So the parameter to this function should be the key and not the
value, because the value is undefined. And it could crash if the
value is used.

This patches fixes two occurences where the value was passed
instead of the key.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix "config_error_nonbool" used with value instead of key
Christian Couder [Sun, 6 Jul 2008 04:10:04 +0000 (06:10 +0200)] 
Fix "config_error_nonbool" used with value instead of key

The function "config_error_nonbool", that is defined in "config.c",
is used to report an error when a config key in the config file
should have a corresponding value but it hasn't.

So the parameter to this function should be the key and not the
value, because the value is undefined. And it could crash if the
value is used.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'qq/maint'
Junio C Hamano [Sun, 6 Jul 2008 01:33:16 +0000 (18:33 -0700)] 
Merge branch 'qq/maint'

* qq/maint:
  clone -q: honor "quiet" option over native transports.
  attribute documentation: keep EXAMPLE at end
  builtin-commit.c: Use 'git_config_string' to get 'commit.template'
  http.c: Use 'git_config_string' to clean up SSL config.
  diff.c: Use 'git_config_string' to get 'diff.external'
  convert.c: Use 'git_config_string' to get 'smudge' and 'clean'
  builtin-log.c: Use 'git_config_string' to get 'format.subjectprefix' and 'format.suffix'
  Documentation cvs: Clarify when a bare repository is needed
  Documentation: be precise about which date --pretty uses

Conflicts:

Documentation/gitattributes.txt

16 years agoclone -q: honor "quiet" option over native transports.
Junio C Hamano [Sun, 6 Jul 2008 00:58:50 +0000 (17:58 -0700)] 
clone -q: honor "quiet" option over native transports.

The earlier built-in conversion seems to have broken "git-clone"; this
teaches the command to honor the "-q" option again when talking to the
remote end over native transports (file://, git:// and ssh://).

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoattribute documentation: keep EXAMPLE at end
Junio C Hamano [Sun, 6 Jul 2008 01:14:27 +0000 (18:14 -0700)] 
attribute documentation: keep EXAMPLE at end

The document gives overall definition of states in DESCRIPTION, describes
various aspects of git operations that can be influenced in EFFECTS, and
finally gives examples in the EXAMPLE section.  Archive creation however
was somehow documented after the EXAMPLE section, not insode EFFECTS.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agobuiltin-commit.c: Use 'git_config_string' to get 'commit.template'
Brian Hetro [Sat, 5 Jul 2008 05:24:40 +0000 (01:24 -0400)] 
builtin-commit.c: Use 'git_config_string' to get 'commit.template'

Signed-off-by: Brian Hetro <whee@smaertness.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agohttp.c: Use 'git_config_string' to clean up SSL config.
Brian Hetro [Sat, 5 Jul 2008 05:24:44 +0000 (01:24 -0400)] 
http.c: Use 'git_config_string' to clean up SSL config.

Signed-off-by: Brian Hetro <whee@smaertness.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agodiff.c: Use 'git_config_string' to get 'diff.external'
Brian Hetro [Sat, 5 Jul 2008 05:24:43 +0000 (01:24 -0400)] 
diff.c: Use 'git_config_string' to get 'diff.external'

Signed-off-by: Brian Hetro <whee@smaertness.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoconvert.c: Use 'git_config_string' to get 'smudge' and 'clean'
Brian Hetro [Sat, 5 Jul 2008 05:24:42 +0000 (01:24 -0400)] 
convert.c: Use 'git_config_string' to get 'smudge' and 'clean'

Signed-off-by: Brian Hetro <whee@smaertness.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agobuiltin-log.c: Use 'git_config_string' to get 'format.subjectprefix' and 'format...
Brian Hetro [Sat, 5 Jul 2008 05:24:41 +0000 (01:24 -0400)] 
builtin-log.c: Use 'git_config_string' to get 'format.subjectprefix' and 'format.suffix'

Signed-off-by: Brian Hetro <whee@smaertness.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation cvs: Clarify when a bare repository is needed
Matthew Ogilvie [Sat, 5 Jul 2008 04:43:41 +0000 (22:43 -0600)] 
Documentation cvs: Clarify when a bare repository is needed

New users sometimes import a project and then immediately
try to use the imported repository as a central shared repository.
This provides pointers about setting up a bare repository for that
in the parts of the documentation dealing with CVS migration.

Signed-off-by: Matthew Ogilvie <mmogilvi_git@miniinfo.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMove read_revisions_from_stdin from builtin-rev-list.c to revision.c
Adam Brewster [Sat, 5 Jul 2008 21:26:39 +0000 (17:26 -0400)] 
Move read_revisions_from_stdin from builtin-rev-list.c to revision.c

Reading rev-list parameters from the command line can be reused by
commands other than rev-list.  Move this function to more "library-ish"
place to promote code reuse.

Signed-off-by: Adam Brewster <asb@bu.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix some warnings (on cygwin) to allow -Werror
Ramsay Jones [Thu, 3 Jul 2008 15:52:09 +0000 (16:52 +0100)] 
Fix some warnings (on cygwin) to allow -Werror

When printing valuds of type uint32_t, we should use PRIu32, and should
not assume that it is unsigned int.  On 32-bit platforms, it could be
defined as unsigned long. The same caution applies to ntohl().

Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: be precise about which date --pretty uses
Nikolaus Schulz [Sat, 5 Jul 2008 00:00:13 +0000 (02:00 +0200)] 
Documentation: be precise about which date --pretty uses

This makes it explicit that the --pretty formats 'medium' and 'email' use the
author date (and ignore the committer date).

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomanpages: use teletype font for sample command lines
Jonathan Nieder [Thu, 3 Jul 2008 06:06:23 +0000 (01:06 -0500)] 
manpages: use teletype font for sample command lines

I think that some of these uses of italics were meant to be
rendered in quotation marks, anyway.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomanpages: italicize git subcommand names (which were in teletype font)
Jonathan Nieder [Thu, 3 Jul 2008 05:59:09 +0000 (00:59 -0500)] 
manpages: italicize git subcommand names (which were in teletype font)

Italicize those git subcommand names already in teletype we missed.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomanpages: italicize nongit command names (if they are in teletype font)
Jonathan Nieder [Thu, 3 Jul 2008 05:55:07 +0000 (00:55 -0500)] 
manpages: italicize nongit command names (if they are in teletype font)

Some manual pages use teletype font to set command names. We
change them to use italics, instead.  This creates a visual
distinction between names of commands and command lines that
can be typed at the command line. It is also more consistent
with other man pages outside Git.

In this patch, the commands named are non-git commands like bash.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomanpages: italicize gitk's name (where it was in teletype font)
Jonathan Nieder [Thu, 3 Jul 2008 05:49:55 +0000 (00:49 -0500)] 
manpages: italicize gitk's name (where it was in teletype font)

The name `gitk` is sometimes meant to be entered at the command
prompt, but most uses are just referring to the program with that
name (not the incantation to start it).

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomanpages: italicize git command names (which were in teletype font)
Jonathan Nieder [Thu, 3 Jul 2008 05:41:41 +0000 (00:41 -0500)] 
manpages: italicize git command names (which were in teletype font)

The names of git commands are not meant to be entered at the
commandline; they are just names. So we render them in italics,
as is usual for command names in manpages.

Using

doit () {
  perl -e 'for (<>) { s/\`(git-[^\`.]*)\`/'\''\1'\''/g; print }'
}
for i in git*.txt config.txt diff*.txt blame*.txt fetch*.txt i18n.txt \
        merge*.txt pretty*.txt pull*.txt rev*.txt urls*.txt
do
  doit <"$i" >"$i+" && mv "$i+" "$i"
done
git diff

.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomanpages: italicize command names
Jonathan Nieder [Thu, 3 Jul 2008 05:37:18 +0000 (00:37 -0500)] 
manpages: italicize command names

This includes nongit commands like RCS 'merge'.  This patch only
italicizes names of commands if they had no formatting before.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomanpages: italicize command names in synopses
Jonathan Nieder [Thu, 3 Jul 2008 05:36:04 +0000 (00:36 -0500)] 
manpages: italicize command names in synopses

To tell command names from options in a glance.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogitdiffcore(7): fix awkward wording
Jonathan Nieder [Thu, 3 Jul 2008 05:30:25 +0000 (00:30 -0500)] 
gitdiffcore(7): fix awkward wording

The phrase "diff outputs" sounds awkward to my ear (I think
"output" is meant to be used as a substantive noun.)

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: more "git-" versus "git " changes
Jonathan Nieder [Thu, 3 Jul 2008 05:28:15 +0000 (00:28 -0500)] 
Documentation: more "git-" versus "git " changes

With git-commands moving out of $(bindir), it is useful to make a
clearer distinction between the git subcommand 'git-whatever' and
the command you type, `git whatever <options>`.  So we use a dash
after "git" when referring to the former and not the latter.

I already sent a patch doing this same thing, but I missed some
spots.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: rewrap to prepare for "git-" vs "git " change
Jonathan Nieder [Thu, 3 Jul 2008 05:20:21 +0000 (00:20 -0500)] 
Documentation: rewrap to prepare for "git-" vs "git " change

Rewrap lines in preparation for added dashes.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-commit(1): depersonalize description
Jonathan Nieder [Thu, 3 Jul 2008 05:13:45 +0000 (00:13 -0500)] 
git-commit(1): depersonalize description

The intent is to make git-commit(1) feel more like a manual page.  The
change also makes the page four words shorter.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit(1): add comma
Jonathan Nieder [Thu, 3 Jul 2008 05:08:12 +0000 (00:08 -0500)] 
git(1): add comma

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agomanpages: fix bogus whitespace
Jonathan Nieder [Thu, 3 Jul 2008 05:03:54 +0000 (00:03 -0500)] 
manpages: fix bogus whitespace

It's distracting.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoDocumentation: fix gitlinks
Jonathan Nieder [Thu, 3 Jul 2008 04:54:38 +0000 (23:54 -0500)] 
Documentation: fix gitlinks

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agogit-format-patch(1): fix stray \ in output
Jonathan Nieder [Thu, 3 Jul 2008 04:47:05 +0000 (23:47 -0500)] 
git-format-patch(1): fix stray \ in output

In listing blocks (set off by rows of dashes), the usual
formatting characters of asciidoc are instead rendered verbatim.
When the escaped double-hyphen of olden days is moved into such a
block along with other formatting improvements, it becomes
backslash-dash-dash.

So we remove the backslash.

Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoRetire 'stupid' merge strategy
Miklos Vajna [Sat, 5 Jul 2008 14:43:51 +0000 (16:43 +0200)] 
Retire 'stupid' merge strategy

As pointed out by Linus, this strategy tries to take the best merge
base, but 'recursive' just does it better. If one needs something more
than 'resolve' then he/she should really use 'recursive' and not
'stupid'.

Cf. Message-ID: <alpine.LFD.1.10.0807030947360.18105@woody.linux-foundation.org>

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix apply --recount handling of no-EOL line
Thomas Rast [Fri, 4 Jul 2008 19:10:14 +0000 (21:10 +0200)] 
Fix apply --recount handling of no-EOL line

If a patch modifies the last line of a file that previously had no
terminating '\n', it looks like

    -old text
    \ No newline at end of file
    +new text

Hence, a '\' line does not signal the end of the hunk.  This modifies
'git apply --recount' to take this into account.

Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoMerge branch 'maint'
Junio C Hamano [Fri, 4 Jul 2008 08:59:57 +0000 (01:59 -0700)] 
Merge branch 'maint'

* maint:
  GIT 1.5.6.2
  Fix executable bits in t/ scripts
  Work around gcc warnings from curl headers

16 years agoGIT 1.5.6.2 v1.5.6.2
Junio C Hamano [Fri, 4 Jul 2008 08:12:54 +0000 (01:12 -0700)] 
GIT 1.5.6.2

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoFix executable bits in t/ scripts
Junio C Hamano [Fri, 4 Jul 2008 08:38:34 +0000 (01:38 -0700)] 
Fix executable bits in t/ scripts

Pointed out by Ramsay Jones.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agot4127-apply-same-fn: Avoid sed -i
Johannes Sixt [Fri, 4 Jul 2008 06:43:19 +0000 (08:43 +0200)] 
t4127-apply-same-fn: Avoid sed -i

Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
16 years agoWork around gcc warnings from curl headers
Junio C Hamano [Fri, 4 Jul 2008 07:37:40 +0000 (00:37 -0700)] 
Work around gcc warnings from curl headers

After master.k.org upgrade, I started seeing these warning messages:

    transport.c: In function 'get_refs_via_curl':
    transport.c:458: error: call to '_curl_easy_setopt_err_write_callback' declared with attribute warning: curl_easy_setopt expects a curl_write_callback argument for this option

It appears that the curl header wants to enforce the function signature
for callback function given to curl_easy_setopt() to be compatible with
that of (*curl_write_callback) or fwrite.  This patch seems to work the
issue around.

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