Code

git.git
14 years agoaccept "git grep -- pattern"
Jeff King [Sun, 7 Feb 2010 04:44:15 +0000 (23:44 -0500)]
accept "git grep -- pattern"

Currently the only way to "quote" a grep pattern that might
begin with a dash is to use "git grep -e pattern". This
works just fine, and is also the way right way to do it on
many traditional grep implemenations.

Some people prefer to use "git grep -- pattern", however, as
"--" is the usual "end of options" marker, and at least GNU
grep and Solaris 10 grep support this. This patch makes that
syntax work.

There is a slight behavior change, in that "git grep -- $X"
used to be interpreted as "grep for -- in $X". However, that
usage is questionable. "--" is usually the end-of-options
marker, so "git grep" was unlike many other greps in
treating it as a literal pattern (e.g., both GNU grep and
Solaris 10 grep will treat "grep --" as missing a pattern).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoRevert 30816237 and 7e62265
Junio C Hamano [Fri, 5 Feb 2010 17:20:01 +0000 (09:20 -0800)]
Revert 30816237 and 7e62265

It seems that we have bad interaction with the code related to
GIT_WORK_TREE and "grep --no-index", and broke running grep inside
the .git directory.  For now, just revert it and resurrect it after
1.7.0 ships.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogit-svn: persistent memoization
Andrew Myrick [Sat, 30 Jan 2010 03:14:22 +0000 (03:14 +0000)]
git-svn: persistent memoization

Make memoization of the svn:mergeinfo processing functions persistent with
Memoize::Storable so that the memoization tables don't need to be regenerated
every time the user runs git-svn fetch.

The Memoize::Storable hashes are stored in ENV{GIT_DIR}/svn/.caches.

[ew: changed caches path to avoid conflicts with old repos]
[ew: File::Path::{make_path => mkpath} for compatibility]
[ew: line-wrapped at 80 chars]

Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Andrew Myrick <amyrick@apple.com>
14 years agofast-import: count --max-pack-size in bytes
Junio C Hamano [Thu, 4 Feb 2010 19:10:44 +0000 (11:10 -0800)]
fast-import: count --max-pack-size in bytes

Similar in spirit to 07cf0f2 (make --max-pack-size argument to 'git
pack-object' count in bytes, 2010-02-03) which made the option by the same
name to pack-objects, this counts the pack size limit in bytes.

In order not to cause havoc with people used to the previous megabyte
scale an integer smaller than 8192 is interpreted in megabytes but the
user gets a warning.  Also a minimum size of 1 MiB is enforced to avoid an
explosion of pack files.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Acked-by: Nicolas Pitre <nico@fluxnic.net>
14 years agoupdate git-repack documentation wrt repack.UseDeltaBaseOffset
Nicolas Pitre [Thu, 4 Feb 2010 21:19:35 +0000 (16:19 -0500)]
update git-repack documentation wrt repack.UseDeltaBaseOffset

This default for repack.UseDeltaBaseOffset has been "true" since
Git v1.6.0.

Signed-off-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogit-clean: fix the description of the default behavior
Michael J Gruber [Thu, 4 Feb 2010 16:01:16 +0000 (17:01 +0100)]
git-clean: fix the description of the default behavior

Currently, when called without -n and -f, git clean issues

fatal: clean.requireForce not set and -n or -f not given; refusing to clean

which leaves the user wondering why force is required when requireForce
is not set. Looking up in git-clean(1) does not help because its
description is wrong.

Change it so that git clean issues

fatal: clean.requireForce defaults to true and -n or -f not given; refusing to clean

in this situation (and "...set to true..." when it is set) which makes
it clearer that an unset config means true here, and adjust the
documentation.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agofast-import.c: Fix big-file-threshold parsing bug
Junio C Hamano [Thu, 4 Feb 2010 02:27:08 +0000 (18:27 -0800)]
fast-import.c: Fix big-file-threshold parsing bug

Manual merge made at 844ad3d (Merge branch 'sp/maint-fast-import-large-blob'
into sp/fast-import-large-blob, 2010-02-01) did not correctly reflect the change
of unit in which this variable's value is counted from its previous version.

Now it counts in bytes, not in megabytes.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
14 years agomake --max-pack-size argument to 'git pack-object' count in bytes
Nicolas Pitre [Thu, 4 Feb 2010 03:48:28 +0000 (22:48 -0500)]
make --max-pack-size argument to 'git pack-object' count in bytes

The value passed to --max-pack-size used to count in MiB which was
inconsistent with the corresponding configuration variable as well as
other command arguments which are defined to count in bytes with an
optional unit suffix.  This brings --max-pack-size in line with the
rest of Git.

Also, in order not to cause havoc with people used to the previous
megabyte scale, and because this is a sane thing to do anyway, a
minimum size of 1 MiB is enforced to avoid an explosion of pack files.

Adjust and extend test suite accordingly.

Signed-off-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agopack-objects: fix pack generation when using pack_size_limit
Nicolas Pitre [Thu, 4 Feb 2010 03:48:27 +0000 (22:48 -0500)]
pack-objects: fix pack generation when using pack_size_limit

Current handling of pack_size_limit is quite suboptimal.  Let's consider
a list of objects to pack which contain alternatively big and small
objects (which pretty matches reality when big blobs are interlaced
with tree objects).  Currently, the code simply close the pack and opens
a new one when the next object in line breaks the size limit.

The current code may degenerate into:

  - small tree object => store into pack #1
  - big blob object busting the pack size limit => store into pack #2
  - small blob but pack #2 is over the limit already => pack #3
  - big blob busting the size limit => pack #4
  - small tree but pack #4 is over the limit => pack #5
  - big blob => pack #6
  - small tree => pack #7
  - ... and so on.

The reality is that the content of packs 1, 3, 5 and 7 could well be
stored more efficiently (and delta compressed) together in pack #1 if
the big blobs were not forcing an immediate transition to a new pack.

Incidentally this can be fixed pretty easily by simply skipping over
those objects that are too big to fit in the current pack while trying
the whole list of unwritten objects, and then that list considered from
the beginning again when a new pack is opened.  This creates much fewer
smallish pack files and help making more predictable test cases for the
test suite.

This change made one of the self sanity checks useless so it is removed
as well. That check was rather redundant already anyway.

Signed-off-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agofix multiple issues with t5300
Nicolas Pitre [Thu, 4 Feb 2010 03:48:26 +0000 (22:48 -0500)]
fix multiple issues with t5300

First of all, trying to run 'git verify-pack' on packs  produced by
the tests using pack.packSizeLimit always failed.  After lots of digging
and head scratching, it turns out that the preceeding test simulating
a SHA1 collision did leave the repository quite confused, impacting
subsequent tests.

So let's move that destructive test last, and add tests to run
verify-pack on the output from those packSizeLimit tests to catch such
goofage.

Finally, group those packSizeLimit tests together.

Signed-off-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogitweb: Simplify (and fix) chop_str
John 'Warthog9' Hawley [Wed, 3 Feb 2010 11:28:28 +0000 (12:28 +0100)]
gitweb: Simplify (and fix) chop_str

The chop_str subroutine is meant to be used on strings (such as commit
description / title) *before* HTML escaping, which means before
applying esc_html or equivalent.

Therefore get rid of the failed attempt to always remove full HTML
entities (like e.g. &amp; or &nbsp;).  It is not necessary (HTML
entities gets added later), and it can cause chop_str to chop a string
incorrectly.

Specifically:

     API & protocol: support option to force written data immediately to disk

from http://git.kernel.org/?p=daemon/distsrv/chunkd.git;a=commit;h=3b02f749df2cb1288f345a689d85e7061f507e54

The short version of the title gets chopped to

     API ...

where it should be

     API & protocol: support option to force written data...

Noticed-by: John 'Warthog9' Hawley <warthog9@kernel.org>
Signed-off-by: John 'Warthog9' Hawley <warthog9@kernel.org>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogrep: simplify assignment of ->fixed
René Scharfe [Wed, 3 Feb 2010 18:16:30 +0000 (19:16 +0100)]
grep: simplify assignment of ->fixed

After 885d211e, the value of the ->fixed pattern option only depends on
the grep option of the same name.  Regex flags don't matter any more,
because fixed mode and regex mode are strictly separated.  Thus we can
simply copy the value from struct grep_opt to struct grep_pat, as we do
already for ->word_regexp and ->ignore_case.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoMerge branch 'ms/filter-branch-submodule'
Junio C Hamano [Wed, 3 Feb 2010 05:48:34 +0000 (21:48 -0800)]
Merge branch 'ms/filter-branch-submodule'

* ms/filter-branch-submodule:
  filter-branch: Add tests for submodules in tree-filter
  filter-branch: Fix to allow replacing submodules with another content

14 years agoMerge branch 'jh/gitweb-caching' (early part)
Junio C Hamano [Wed, 3 Feb 2010 05:48:22 +0000 (21:48 -0800)]
Merge branch 'jh/gitweb-caching' (early part)

* 'jh/gitweb-caching' (early part):
  gitweb: Add optional extra parameter to die_error, for extended explanation
  gitweb: add a "string" variant of print_sort_th
  gitweb: add a "string" variant of print_local_time
  gitweb: Check that $site_header etc. are defined before using them
  gitweb: Makefile improvements
  gitweb: Load checking
  gitweb: Make running t9501 test with '--debug' reliable and usable

14 years agoMerge branch 'bw/no-python-autoconf'
Junio C Hamano [Wed, 3 Feb 2010 05:48:13 +0000 (21:48 -0800)]
Merge branch 'bw/no-python-autoconf'

* bw/no-python-autoconf:
  configure: Allow --without-python
  configure: Allow GIT_ARG_SET_PATH to handle --without-PROGRAM

14 years agoMerge branch 'sp/fast-import-large-blob'
Junio C Hamano [Wed, 3 Feb 2010 05:47:51 +0000 (21:47 -0800)]
Merge branch 'sp/fast-import-large-blob'

* sp/fast-import-large-blob:
  fast-import: Stream very large blobs directly to pack

14 years agoCorrect spelling of 'REUC' extension
Shawn O. Pearce [Tue, 2 Feb 2010 15:33:28 +0000 (07:33 -0800)]
Correct spelling of 'REUC' extension

The new dircache extension CACHE_EXT_RESOLVE_UNDO, whose value is
0x52455543, is actually the ASCII sequence 'REUC', not the ASCII
sequence 'REUN'.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agobuild: make code "-Wpointer-arith" clean
Junio C Hamano [Tue, 2 Feb 2010 06:04:03 +0000 (22:04 -0800)]
build: make code "-Wpointer-arith" clean

Recently introduced resolve_undo_read() expected arithmetic to (void *)
to work on byte-addresses.  Correct this.

Noticed by Brandon Casey.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoconfigure: Allow --without-python
Ben Walton [Mon, 1 Feb 2010 02:15:05 +0000 (21:15 -0500)]
configure: Allow --without-python

This patch allows someone to use configure to build git while at the
same time disabling the python remote helper code.  It leverages the
ability of GIT_ARG_SET_PATH to accept an optional second argument
indicating that --without-$PROGRAM is acceptable.

Signed-off-by: Ben Walton <bwalton@artsci.utoronto.ca>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoconfigure: Allow GIT_ARG_SET_PATH to handle --without-PROGRAM
Ben Walton [Mon, 1 Feb 2010 02:15:04 +0000 (21:15 -0500)]
configure: Allow GIT_ARG_SET_PATH to handle --without-PROGRAM

Add an optional second argument to both GIT_ARG_SET_PATH and
GIT_CONF_APPEND_PATH such that any value of the second argument will
enable configure to set NO_$PROGRAM in addition to an empty
$PROGRAM_PATH.  This is initially useful for allowing configure to
disable the use of python, as the remote helper code has nothing
leveraging it yet.

The Makefile already recognizes NO_PYTHON, but configure provided no
way to set it appropriately.

Signed-off-by: Ben Walton <bwalton@artsci.utoronto.ca>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoMerge branch 'sp/maint-fast-import-large-blob' into sp/fast-import-large-blob
Junio C Hamano [Mon, 1 Feb 2010 20:41:31 +0000 (12:41 -0800)]
Merge branch 'sp/maint-fast-import-large-blob' into sp/fast-import-large-blob

* sp/maint-fast-import-large-blob:
  fast-import: Stream very large blobs directly to pack
  bash: don't offer remote transport helpers as subcommands

Conflicts:
fast-import.c

14 years agofast-import: Stream very large blobs directly to pack
Shawn O. Pearce [Mon, 1 Feb 2010 17:27:35 +0000 (09:27 -0800)]
fast-import: Stream very large blobs directly to pack

If a blob is larger than the configured big-file-threshold, instead
of reading it into a single buffer obtained from malloc, stream it
onto the end of the current pack file.  Streaming the larger objects
into the pack avoids the 4+ GiB memory footprint that occurs when
fast-import is processing 2+ GiB blobs.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogit-p4: Fix sync errors due to new server version
Pal-Kristian Engstad [Fri, 22 Jan 2010 02:33:00 +0000 (18:33 -0800)]
git-p4: Fix sync errors due to new server version

Fix sync errors due to new Perforce servers.

The P4D/NTX64/2009.2/228098 (2009/12/16) server reports
'move/delete' instead of 'delete'. This causes the Perforce
depot and the git repo to get out of sync. Fixed by adding
the new status string.

Signed-off-by: Pal-Kristian Engstad <pal_engstad@naughtydog.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoUpdates for dirty submodules in release notes and user manual
Jens Lehmann [Mon, 1 Feb 2010 18:09:39 +0000 (19:09 +0100)]
Updates for dirty submodules in release notes and user manual

In the release notes "git status" was not mentioned, also shortly explain
the "-dirty" output generated by diff.

Added a paragraph to the "Pitfalls with submodules" section in
user-manual.txt describing new and old behavior of "git status" and "git
diff" for dirty submodules.

Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoUpdate draft release notes to 1.7.0
Junio C Hamano [Sun, 31 Jan 2010 20:20:30 +0000 (12:20 -0800)]
Update draft release notes to 1.7.0

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoMerge branch 'dm/make-threaded-simplify'
Junio C Hamano [Sun, 31 Jan 2010 20:09:35 +0000 (12:09 -0800)]
Merge branch 'dm/make-threaded-simplify'

* dm/make-threaded-simplify:
  Make NO_PTHREADS the sole thread configuration variable

14 years agoMake NO_PTHREADS the sole thread configuration variable
Dan McGee [Sat, 30 Jan 2010 01:22:19 +0000 (19:22 -0600)]
Make NO_PTHREADS the sole thread configuration variable

When the first piece of threaded code was introduced in commit 8ecce684, it
came with its own THREADED_DELTA_SEARCH Makefile option. Since this time,
more threaded code has come into the codebase and a NO_PTHREADS option has
also been added. Get rid of the original option as the newer, more generic
option covers everything we need.

Signed-off-by: Dan McGee <dpmcgee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoRPM packaging: use %global inside %{!?...}
Junio C Hamano [Sun, 31 Jan 2010 19:11:16 +0000 (11:11 -0800)]
RPM packaging: use %global inside %{!?...}

According to
https://www.redhat.com/archives/fedora-devel-list/2010-January/msg00093.html

scope of %define lasts until the end brace; earlier RPM up to Fedora 12
didn't necessarily honor the scope, but later versions corrected the bug.

Problem and solution both pointed out by Todd Zullinger.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agomention new shell execution behavior in release notes
Jeff King [Sun, 31 Jan 2010 09:14:46 +0000 (04:14 -0500)]
mention new shell execution behavior in release notes

This is already in the "bells and whistles" section, but it also has a
slight chance of breakage, so let's also mention it in the "changed
behaviors" section.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoFix memory leak in submodule.c
Jens Lehmann [Sun, 31 Jan 2010 16:43:49 +0000 (17:43 +0100)]
Fix memory leak in submodule.c

The strbuf used in add_submodule_odb() was never released. So for every
submodule - populated or not - we leaked its object directory name when
using "git diff*" with the --submodule option.

Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoFix typos in technical documentation.
Ralf Wildenhues [Sun, 31 Jan 2010 13:24:39 +0000 (14:24 +0100)]
Fix typos in technical documentation.

Signed-off-by: Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogitweb: Add optional extra parameter to die_error, for extended explanation
John 'Warthog9' Hawley [Sat, 30 Jan 2010 22:30:44 +0000 (23:30 +0100)]
gitweb: Add optional extra parameter to die_error, for extended explanation

Add a 3rd, optional, parameter to die_error to allow for extended error
information to be output along with what the error was.

Signed-off-by: John 'Warthog9' Hawley <warthog9@kernel.org>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogitweb: add a "string" variant of print_sort_th
John 'Warthog9' Hawley [Sat, 30 Jan 2010 22:30:43 +0000 (23:30 +0100)]
gitweb: add a "string" variant of print_sort_th

Add a function (named format_sort_th) that returns the string that
print_sort_th would print.

Signed-off-by: John 'Warthog9' Hawley <warthog9@kernel.org>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogitweb: add a "string" variant of print_local_time
John 'Warthog9' Hawley [Sat, 30 Jan 2010 22:30:42 +0000 (23:30 +0100)]
gitweb: add a "string" variant of print_local_time

Add a function (named format_local_time) that returns the string that
print_local_time would print.

Signed-off-by: John 'Warthog9' Hawley <warthog9@kernel.org>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogitweb: Check that $site_header etc. are defined before using them
John 'Warthog9' Hawley [Sat, 30 Jan 2010 22:30:41 +0000 (23:30 +0100)]
gitweb: Check that $site_header etc. are defined before using them

If one of $site_header, $site_footer or $home_text is not defined, you
get extraneous errors in the web logs, for example (line wrapped for
better readibility):

 [Wed Jan 13 16:55:42 2010] [error] [client ::1] [Wed Jan 13 16:55:42 2010]
 gitweb.cgi: Use of uninitialized value $site_header in -f at
 /var/www/gitweb/gitweb.cgi line 3287., referer: http://git/gitweb.cgi

This ensures that those variables are defined before trying to use it.

Note that such error can happen only because of an error in gitweb
config file; building gitweb.cgi can make mentioned variables holding
empty string (it is even the default), but they are still defined.

Signed-off-by: John 'Warthog9' Hawley <warthog9@kernel.org>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogitweb: Makefile improvements
John 'Warthog9' Hawley [Sat, 30 Jan 2010 22:30:40 +0000 (23:30 +0100)]
gitweb: Makefile improvements

Adjust the main Makefile so you can simply run

     make gitweb

which in turn calls gitweb/Makefile.  This means that in order to
generate gitweb, you can simply run 'make' from gitweb subdirectory:

     cd gitweb
     make

Targets gitweb/gitweb.cgi and (dependent on JSMIN being defined)
gitweb/gitweb.min.js in main Makefile are preserved for backward
compatibility.

Signed-off-by: John 'Warthog9' Hawley <warthog9@kernel.org>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogitweb: Load checking
John 'Warthog9' Hawley [Sat, 30 Jan 2010 22:30:39 +0000 (23:30 +0100)]
gitweb: Load checking

This changes slightly the behavior of gitweb, so that it verifies
that the box isn't inundated with before attempting to serve gitweb.
If the box is overloaded, it basically returns a 503 Server Unavailable
until the load falls below the defined threshold.  This helps dramatically
if you have a box that's I/O bound, reaches a certain load and you
don't want gitweb, the I/O hog that it is, increasing the pain the
server is already undergoing.

This behavior is controlled by $maxload configuration variable.
Default is a load of 300, which for most cases should never be hit.
Unset it (set it to undefined value, i.e. undef) to turn off checking.

Currently it requires that '/proc/loadavg' file exists, otherwise the
load check is bypassed (load is taken to be 0).  So platforms that do
not implement '/proc/loadavg' currently cannot use this feature
(provisions are included for additional checks to be added by others).

There is simple test in t/t9501-gitweb-standalone-http-status.sh to
check that it correctly returns "503 Service Unavailable" if load is
too high, and also if there are any Perl warnings or errors.

Signed-off-by: John 'Warthog9' Hawley <warthog9@kernel.org>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogitweb: Make running t9501 test with '--debug' reliable and usable
Jakub Narebski [Sat, 30 Jan 2010 22:30:38 +0000 (23:30 +0100)]
gitweb: Make running t9501 test with '--debug' reliable and usable

Remove test_debug lines after 'snapshots: tgz only default format
enabled' and 'snapshots: all enabled in default, use default disabled
value' tests.  Those tests constitute of multiple gitweb_run
invocation, therefore outputting gitweb.output for the last gitweb_run
wouldn't help much in debugging test failure, and can only confuse.

For snapshot tests which check for "200 OK" status, change
  test_debug 'cat gitweb.output'
to
  test_debug 'cat gitweb.headers'
Otherwise when running this test with '--debug' option,
t/t9501-gitweb-standalone-http-status.sh would dump *binary data* (the
snapshot itself) to standard output, which can mess up state of terminal
due to term control characters which can be embedded in output.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agois_submodule_modified(): fix breakage with external GIT_INDEX_FILE v1.7.0-rc1
Junio C Hamano [Sat, 30 Jan 2010 20:08:26 +0000 (12:08 -0800)]
is_submodule_modified(): fix breakage with external GIT_INDEX_FILE

Even when the environment was given for the top-level process, checking
in the submodule work tree should use the index file associated with the
work tree of the submodule.  Do not export it to the environment.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoRPM packaging: don't include foreign-scm-helper bits yet
Junio C Hamano [Sat, 30 Jan 2010 17:45:11 +0000 (09:45 -0800)]
RPM packaging: don't include foreign-scm-helper bits yet

The files in /usr/lib/python* are only the support infrastructure for
foreign scm interface yet to be written and/or shipped with git.  Don't
include them in the binary package (this will also free us from Python
dependency).

When we ship with foreign scm interface, we will need to package these
files with it in a separate subpackage, but we are not there yet.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogrep: Fix two memory leaks
Dan McGee [Sat, 30 Jan 2010 15:42:58 +0000 (09:42 -0600)]
grep: Fix two memory leaks

We duplicate the grep_opt structure when using grep threads, but didn't
later free either the patterns attached to this new structure or the
structure itself.

Signed-off-by: Dan McGee <dpmcgee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agorebase: don't invoke the pager for each commit summary
Markus Heidelberg [Sat, 30 Jan 2010 16:23:37 +0000 (17:23 +0100)]
rebase: don't invoke the pager for each commit summary

This regression was introduced by commit 0aa958d (rebase: replace
antiquated sed invocation, 2010-01-24), which changed the invocation of
"git rev-list | sed" to "git log".

It can be reproduced by something like this:
$ git rebase -s recursive origin/master

Signed-off-by: Markus Heidelberg <markus.heidelberg@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoUpdate draft release notes to 1.7.0
Junio C Hamano [Sat, 30 Jan 2010 07:38:31 +0000 (23:38 -0800)]
Update draft release notes to 1.7.0

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoMerge branch 'maint'
Junio C Hamano [Sat, 30 Jan 2010 07:36:17 +0000 (23:36 -0800)]
Merge branch 'maint'

* maint:
  fix memcpy of overlapping area

14 years agoMerge branch 'maint-1.6.5' into maint
Junio C Hamano [Sat, 30 Jan 2010 07:36:13 +0000 (23:36 -0800)]
Merge branch 'maint-1.6.5' into maint

* maint-1.6.5:
  fix memcpy of overlapping area

14 years agoImplement pthread_cond_broadcast on Windows
Johannes Sixt [Fri, 29 Jan 2010 23:54:05 +0000 (00:54 +0100)]
Implement pthread_cond_broadcast on Windows

See http://www.cse.wustl.edu/~schmidt/win32-cv-1.html, section "The
SignalObjectAndWait solution". But note that this implementation does not
use SignalObjectAndWait (which is needed to achieve fairness, but we do
not need fairness).

Note that our implementations of pthread_cond_broadcast and
pthread_cond_signal require that they are invoked with the mutex held that
is used in the pthread_cond_wait calls.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoIf deriving SVN_SSH from GIT_SSH on msys, also add quotes
Sebastian Schuberth [Sat, 23 Jan 2010 14:20:28 +0000 (15:20 +0100)]
If deriving SVN_SSH from GIT_SSH on msys, also add quotes

In contrast to GIT_SSH, SVN_SSH requires quotes for paths that contain
spaces. As GIT_SSH will not work if it contains quotes, it is safe to
assume it never contains quotes. Also, adding quotes to SVN_SSH for paths
that do not contain spaces does no harm. So we always add quotes when
deriving SVN_SSH from GIT_SSH on msys.

This fixes msysGit issue 385, see
http://code.google.com/p/msysgit/issues/detail?id=385

Signed-off-by: Sebastian Schuberth <sschuberth@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoMerge git://git.kernel.org/pub/scm/gitk/gitk
Junio C Hamano [Fri, 29 Jan 2010 20:57:44 +0000 (12:57 -0800)]
Merge git://git.kernel.org/pub/scm/gitk/gitk

* git://git.kernel.org/pub/scm/gitk/gitk:
  gitk: Update German translation
  gitk: Add French translation
  gitk: update Italian translation
  gitk: Update Swedish translation
  gitk: Adjust two equal strings which differed in whitespace
  gitk: Display submodule diffs with appropriate encoding
  gitk: Fix display of newly-created tags
  gitk: Enable gitk to create tags with messages
  gitk: Update Hungarian translation
  gitk: Add Hungarian translation
  gitk: Add "--no-replace-objects" option

14 years agoadd shebang line to git-mergetool--lib.sh
Jeff King [Fri, 29 Jan 2010 10:37:23 +0000 (05:37 -0500)]
add shebang line to git-mergetool--lib.sh

Even though this script is expected to be sourced instead of
executed on its own, the #!/bin/sh line provides simple
documentation about what format the file is in.

In particular, the lack of such a line was confusing the
valgrind support of our test scripts, which assumed that any
executable without a #!-line should be intercepted and run
through valgrind. So during valgrind-enabled tests, any
script sourcing this file actually sourced the valgrind
interception script instead.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agofix off-by-one allocation error
Jeff King [Fri, 29 Jan 2010 10:31:30 +0000 (05:31 -0500)]
fix off-by-one allocation error

Caught by valgrind in t5516. Reading the code shows we
malloc enough for our string, but not trailing NUL.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agofix memcpy of overlapping area
Jeff King [Fri, 29 Jan 2010 10:28:44 +0000 (05:28 -0500)]
fix memcpy of overlapping area

Caught by valgrind in t5500, but it is pretty obvious from
reading the code that this is shifting elements of an array
to the left, which needs memmove.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogitk: Update German translation
Christian Stimming [Fri, 29 Jan 2010 11:55:14 +0000 (22:55 +1100)]
gitk: Update German translation

Signed-off-by: Christian Stimming <stimming@tuhh.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
14 years agogitk: Add French translation
Emmanuel Trillaud [Tue, 12 Jan 2010 13:02:50 +0000 (14:02 +0100)]
gitk: Add French translation

Signed-off-by: Emmanuel Trillaud <etrillaud@gmail.com>
Signed-off-by: Thomas Moulard <thomas.moulard@gmail.com>
Signed-off-by: Guy Brand <gb@unistra.fr>
Signed-off-by: Nicolas Sebrecht <nicolas.s.dev@gmx.fr>
Signed-off-by: Paul Mackerras <paulus@samba.org>
14 years agogitk: update Italian translation
Michele Ballabio [Fri, 29 Jan 2010 11:51:28 +0000 (22:51 +1100)]
gitk: update Italian translation

Signed-off-by: Michele Ballabio <barra_cuda@katamail.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
14 years agogitk: Update Swedish translation
Peter Krefting [Fri, 29 Jan 2010 11:50:29 +0000 (22:50 +1100)]
gitk: Update Swedish translation

Signed-off-by: Peter Krefting <peter@softwolves.pp.se>
Signed-off-by: Paul Mackerras <paulus@samba.org>
14 years agoMerge branch 'maint'
Junio C Hamano [Thu, 28 Jan 2010 22:33:33 +0000 (14:33 -0800)]
Merge branch 'maint'

* maint:
  bash: don't offer remote transport helpers as subcommands

14 years agobash: support 'git notes' and its subcommands
SZEDER Gábor [Thu, 28 Jan 2010 01:05:55 +0000 (02:05 +0100)]
bash: support 'git notes' and its subcommands

... and it will offer refs unless after -m or -F, because these two
options require a non-ref argument.

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>
14 years agofilter-branch: Add tests for submodules in tree-filter
Michal Sojka [Thu, 28 Jan 2010 09:08:47 +0000 (10:08 +0100)]
filter-branch: Add tests for submodules in tree-filter

Add tests to make sure that:

1) a submodule can be removed and its content replaced with regular files
   ('rewrite submodule with another content'). This test passes only with
   the previous patch applied.

2) it is possible to replace submodule revision by direct index
   manipulation ('replace submodule revision'). Although it would be
   better to run such a filter in --index-filter, this test shows that
   this functionality is not broken by the previous patch. This succeeds
   both with and without the previous patch.

Signed-off-by: Michal Sojka <sojkam1@fel.cvut.cz>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agofilter-branch: Fix to allow replacing submodules with another content
Michal Sojka [Thu, 28 Jan 2010 09:08:46 +0000 (10:08 +0100)]
filter-branch: Fix to allow replacing submodules with another content

When git filter-branch is used to replace a submodule with another
content, it always fails on the first commit.

Consider a repository with submod directory containing a submodule.  The
following command to remove the submodule and replace it with a file fails:

    git filter-branch --tree-filter 'rm -rf submod &&
                                     git rm -q submod &&
                                     mkdir submod &&
                                     touch submod/file'

with an error:

    error: submod: is a directory - add files inside instead

The reason is that git diff-index, which generates the first part of the
list of files updated by the tree filter, emits also the removed submodule
even if it was replaced by a real directory.

Signed-off-by: Michal Sojka <sojkam1@fel.cvut.cz>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agobash: don't offer remote transport helpers as subcommands
SZEDER Gábor [Fri, 22 Jan 2010 10:54:51 +0000 (11:54 +0100)]
bash: don't offer remote transport helpers as subcommands

Since commits a2d725b7 (Use an external program to implement fetching
with curl, 2009-08-05) and c9e388bb (Make the
"traditionally-supported" URLs a special case, 2009-09-03) remote
transport helpers like 'remote-ftp' and 'remote-curl' are offered by the
completion script as available subcommands.  Not good, since they are
helpers, therefore should not be offered, so filter them out.

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>
14 years agogrep --quiet: finishing touches
Junio C Hamano [Thu, 28 Jan 2010 20:33:42 +0000 (12:33 -0800)]
grep --quiet: finishing touches

Name the option "--quiet" not "--quick", document it, and add tests.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoreject @{-1} not at beginning of object name
Jeff King [Thu, 28 Jan 2010 09:56:43 +0000 (04:56 -0500)]
reject @{-1} not at beginning of object name

Something like foo@{-1} is nonsensical, as the @{-N} syntax
is reserved for "the Nth last branch", and is not an actual
reflog selector. We should not feed such nonsense to
approxidate at all.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agofix parsing of @{-1}@{u} combination
Jeff King [Thu, 28 Jan 2010 09:52:22 +0000 (04:52 -0500)]
fix parsing of @{-1}@{u} combination

Previously interpret_branch_name would see @{-1} and stop
parsing, leaving the @{u} as cruft that provoked an error.
Instead, we should recurse if there is more to parse.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agotest combinations of @{} syntax
Jeff King [Thu, 28 Jan 2010 09:50:20 +0000 (04:50 -0500)]
test combinations of @{} syntax

Now that we have several different types of @{} syntax, it
is a good idea to test them together, which reveals some
failures.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agorerere: fix too-short initialization
Jeff King [Thu, 28 Jan 2010 14:52:16 +0000 (09:52 -0500)]
rerere: fix too-short initialization

This was caused by a typo in the sizeof parameter, and meant
we looked at uninitialized memory.  Caught by valgrind in
t2030.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agot0101: use absolute date
Jeff King [Thu, 28 Jan 2010 08:52:18 +0000 (03:52 -0500)]
t0101: use absolute date

The original version used relative approxidates, which don't
reproduce as reliably as absolute ones. Commit 6c647a fixed
this for one case, but missed the "silly" case.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoMerge branch 'fk/threaded-grep'
Junio C Hamano [Thu, 28 Jan 2010 08:46:45 +0000 (00:46 -0800)]
Merge branch 'fk/threaded-grep'

* fk/threaded-grep:
  Threaded grep
  grep: expose "status-only" feature via -q

14 years agoMerge branch 'gp/maint-cvsserver'
Junio C Hamano [Thu, 28 Jan 2010 08:46:33 +0000 (00:46 -0800)]
Merge branch 'gp/maint-cvsserver'

* gp/maint-cvsserver:
  git-cvsserver: allow regex metacharacters in CVSROOT

14 years agotests: update tests that used to fail
Junio C Hamano [Thu, 28 Jan 2010 08:41:52 +0000 (00:41 -0800)]
tests: update tests that used to fail

"diff --cc" output t4038 tests was fixed by b810cbb (diff --cc: a lost
line at the beginning of the file is shown incorrectly, 2009-07-22), which
was actually the commit that introduced this test..

An error in "git merge -s resolve" t6035 tests was fixed by 730f728
(unpack-trees.c: look ahead in the index, 2009-09-20).

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoMerge branch 'jc/upstream-reflog'
Junio C Hamano [Wed, 27 Jan 2010 22:58:21 +0000 (14:58 -0800)]
Merge branch 'jc/upstream-reflog'

* jc/upstream-reflog:
  Fix log -g this@{upstream}

14 years agoMerge branch 'jc/maint-reflog-bad-timestamp'
Junio C Hamano [Wed, 27 Jan 2010 22:57:37 +0000 (14:57 -0800)]
Merge branch 'jc/maint-reflog-bad-timestamp'

* jc/maint-reflog-bad-timestamp:
  t0101: use a fixed timestamp when searching in the reflog
  Update @{bogus.timestamp} fix not to die()
  approxidate_careful() reports errorneous date string

14 years agoMerge branch 'il/maint-colon-address'
Junio C Hamano [Wed, 27 Jan 2010 22:56:42 +0000 (14:56 -0800)]
Merge branch 'il/maint-colon-address'

* il/maint-colon-address:
  Allow use of []-wrapped addresses in git://
  Support addresses with ':' in git-daemon

14 years agoMerge branch 'il/maint-xmallocz'
Junio C Hamano [Wed, 27 Jan 2010 22:56:38 +0000 (14:56 -0800)]
Merge branch 'il/maint-xmallocz'

* il/maint-xmallocz:
  Fix integer overflow in unpack_compressed_entry()
  Fix integer overflow in unpack_sha1_rest()
  Fix integer overflow in patch_delta()
  Add xmallocz()

14 years agoMerge branch 'jh/maint-config-file-prefix'
Junio C Hamano [Wed, 27 Jan 2010 22:56:25 +0000 (14:56 -0800)]
Merge branch 'jh/maint-config-file-prefix'

* jh/maint-config-file-prefix:
  builtin-config: Fix crash when using "-f <relative path>" from non-root dir

14 years agotransport_get(): drop unnecessary check for !remote
Junio C Hamano [Wed, 27 Jan 2010 20:22:37 +0000 (12:22 -0800)]
transport_get(): drop unnecessary check for !remote

At the beginning of the function we make sure remote is not NULL, and
the remainder of the funciton already depends on it.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoFix remote.<remote>.vcs
Ilari Liusvaara [Wed, 27 Jan 2010 17:53:17 +0000 (19:53 +0200)]
Fix remote.<remote>.vcs

remote.<remote>.vcs causes remote->foreign_vcs to be set on entry to
transport_get(). Unfortunately, the code assumed that any such entry
is stale from previous round.

Fix this by making VCS set by URL to be volatile w.r.t. transport_get()
instead.

Signed-off-by: Ilari Liusvaara <ilari.liusvaara@elisanet.fi>
Acked-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agot0101: use a fixed timestamp when searching in the reflog
Junio C Hamano [Wed, 27 Jan 2010 18:54:22 +0000 (10:54 -0800)]
t0101: use a fixed timestamp when searching in the reflog

That will give us a better reproducibility during tests.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoUpdate @{bogus.timestamp} fix not to die()
Junio C Hamano [Wed, 27 Jan 2010 18:53:09 +0000 (10:53 -0800)]
Update @{bogus.timestamp} fix not to die()

The caller will say "It is not a valid object name" if it wants to, and
some callers may even try to see if it names an object and otherwise try to
see if it is a path.

Pointed out by Jeff King.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoMerge branch 'jl/diff-submodule-ignore'
Junio C Hamano [Wed, 27 Jan 2010 06:53:13 +0000 (22:53 -0800)]
Merge branch 'jl/diff-submodule-ignore'

* jl/diff-submodule-ignore:
  Teach diff --submodule that modified submodule directory is dirty
  git diff: Don't test submodule dirtiness with --ignore-submodules
  Make ce_uptodate() trustworthy again

14 years agowork around an obnoxious bash "safety feature" on OpenBSD
Junio C Hamano [Wed, 27 Jan 2010 00:29:30 +0000 (16:29 -0800)]
work around an obnoxious bash "safety feature" on OpenBSD

Bash (4.0.24) on OpenBSD 4.6 refuses to run this snippet:

    $ cat gomi.sh
    #!/bin/sh
    one="/var/tmp/1 1"
    rm -f /var/tmp/1 "/var/tmp/1 1"
    echo hello >$one
    $ sh gomi.sh; ls /var/tmp/1*
    /var/tmp/1 1
    $ bash gomi.sh; ls /var/tmp/1*
    gomi.sh: line 4: $one: ambiguous redirect
    ls: /var/tmp/1*: No such file or directory

Every competent shell programmer knows that a <$word in redirection is not
subject to field splitting (POSIX.1 "2.7 Redirection" explicitly lists the
kind of expansion performed: "... the word that follows the redirection
operator shall be subjected to ...", and "Field Splitting" is not among
them).

Some clueless folks apparently decided that users need to be protected in
the name of "security", however.

Output from "git grep -e '> *\$' -- '*.sh'" indicates that rebase-i
suffers from this bogus "safety".  Work it around by surrounding the
variable reference with a dq pair.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agofix portability issues with $ in double quotes
Stephen Boyd [Tue, 26 Jan 2010 23:08:31 +0000 (15:08 -0800)]
fix portability issues with $ in double quotes

Using a dollar sign in double quotes isn't portable. Escape them with
a backslash or replace the double quotes with single quotes.

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoapproxidate_careful() reports errorneous date string
Junio C Hamano [Tue, 26 Jan 2010 19:58:00 +0000 (11:58 -0800)]
approxidate_careful() reports errorneous date string

For a long time, the time based reflog syntax (e.g. master@{yesterday})
didn't complain when the "human readable" timestamp was misspelled, as
the underlying mechanism tried to be as lenient as possible.  The funny
thing was that parsing of "@{now}" even relied on the fact that anything
not recognized by the machinery returned the current timestamp.

Introduce approxidate_careful() that takes an optional pointer to an
integer, that gets assigned 1 when the input does not make sense as a
timestamp.

As I am too lazy to fix all the callers that use approxidate(), most of
the callers do not take advantage of the error checking, but convert the
code to parse reflog to use it as a demonstration.

Tests are mostly from Jeff King.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoFix log -g this@{upstream}
Junio C Hamano [Tue, 26 Jan 2010 21:48:28 +0000 (13:48 -0800)]
Fix log -g this@{upstream}

It showed the correct objects but walked a wrong reflog.
Again, tests are from Jeff King.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoFix integer overflow in unpack_compressed_entry()
Ilari Liusvaara [Tue, 26 Jan 2010 18:24:15 +0000 (20:24 +0200)]
Fix integer overflow in unpack_compressed_entry()

Signed-off-by: Ilari Liusvaara <ilari.liusvaara@elisanet.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoFix integer overflow in unpack_sha1_rest()
Ilari Liusvaara [Tue, 26 Jan 2010 18:24:14 +0000 (20:24 +0200)]
Fix integer overflow in unpack_sha1_rest()

[jc: later NUL termination by the caller becomes unnecessary]

Signed-off-by: Ilari Liusvaara <ilari.liusvaara@elisanet.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoFix integer overflow in patch_delta()
Ilari Liusvaara [Tue, 26 Jan 2010 18:24:13 +0000 (20:24 +0200)]
Fix integer overflow in patch_delta()

Signed-off-by: Ilari Liusvaara <ilari.liusvaara@elisanet.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoAdd xmallocz()
Ilari Liusvaara [Tue, 26 Jan 2010 18:24:12 +0000 (20:24 +0200)]
Add xmallocz()

Add routine for allocating NUL-terminated memory block without risking
integer overflow in addition of +1 for NUL byte.

[jc: with suggestion from Bill Lear]

Signed-off-by: Ilari Liusvaara <ilari.liusvaara@elisanet.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogit-cvsserver: allow regex metacharacters in CVSROOT
Gerrit Pape [Tue, 26 Jan 2010 14:47:16 +0000 (14:47 +0000)]
git-cvsserver: allow regex metacharacters in CVSROOT

When run in a repository with a path name containing regex metacharacters
(e.g. +), git-cvsserver failed to split the client request into CVSROOT
and module.  Now metacharacters are disabled for the value of CVSROOT in
the perl regex so that directory names containing metacharacters are
handled properly.

Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agobuiltin-config: Fix crash when using "-f <relative path>" from non-root dir
Johan Herland [Tue, 26 Jan 2010 15:02:16 +0000 (16:02 +0100)]
builtin-config: Fix crash when using "-f <relative path>" from non-root dir

When your current directory is not at the root of the working tree, and you
use the "-f" option with a relative path, the current code tries to read
from a wrong file, since argv[2] is now beyond the end of the rearranged
argument list.

This patch replaces the incorrect argv[2] with the variable holding the
given config file name.

The bug was introduced by d64ec16 (git config: reorganize to use parseopt).

[jc: added test]

Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoAllow use of []-wrapped addresses in git://
Ilari Liusvaara [Tue, 26 Jan 2010 18:24:42 +0000 (20:24 +0200)]
Allow use of []-wrapped addresses in git://

Allow using "["<host>"]":<port> and "["<host>"]" notations in git://
host addresses. This is needed to be able to connect to addresses
that contain ':' (e.g. numeric IPv6 addresses). Also send the host
header []-wrapped so it can actually be parsed by remote end.

Signed-off-by: Ilari Liusvaara <ilari.liusvaara@elisanet.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoSupport addresses with ':' in git-daemon
Ilari Liusvaara [Tue, 26 Jan 2010 18:24:41 +0000 (20:24 +0200)]
Support addresses with ':' in git-daemon

If host address could have ':' in it (e.g. numeric IPv6 address), then
host and port could not be uniquely parsed. Fix this by parsing the
"["<host>"]":<port> and "["<host>"]" notations. Currently the built-in
git:// client would send <host>:<port> or <host> for such thing, but
it doesn't matter as due to bugs, resolving address fails if <host>
contains ':'.

Signed-off-by: Ilari Liusvaara <ilari.liusvaara@elisanet.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogrep: use REG_STARTEND (if available) to speed up regexec
Benjamin Kramer [Tue, 26 Jan 2010 17:48:36 +0000 (18:48 +0100)]
grep: use REG_STARTEND (if available) to speed up regexec

BSD and glibc have an extension to regexec which takes a buffer + length pair
instead of a NUL-terminated string. Since we already have the length computed
this can save us a strlen call inside regexec.

Signed-off-by: Benjamin Kramer <benny.kra@googlemail.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoThreaded grep
Fredrik Kuivinen [Mon, 25 Jan 2010 22:51:39 +0000 (23:51 +0100)]
Threaded grep

Make git grep use threads when it is available.

The results below are best of five runs in the Linux repository (on a
box with two cores).

With the patch:

git grep qwerty
1.58user 0.55system 0:01.16elapsed 183%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+800outputs (0major+5774minor)pagefaults 0swaps

Without:

git grep qwerty
1.59user 0.43system 0:02.02elapsed 100%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+800outputs (0major+3716minor)pagefaults 0swaps

And with a pattern with quite a few matches:

With the patch:

$ /usr/bin/time git grep void
5.61user 0.56system 0:03.44elapsed 179%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+800outputs (0major+5587minor)pagefaults 0swaps

Without:

$ /usr/bin/time git grep void
5.36user 0.51system 0:05.87elapsed 100%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+800outputs (0major+3693minor)pagefaults 0swaps

In either case we gain about 40% by the threading.

Signed-off-by: Fredrik Kuivinen <frekui@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agoam: fix patch format detection for Thunderbird "Save As" emails
Stephen Boyd [Tue, 26 Jan 2010 00:33:59 +0000 (16:33 -0800)]
am: fix patch format detection for Thunderbird "Save As" emails

The patch detection wants to inspect all the headers of a rfc2822 message
and ensure that they look like header fields. The headers are always
separated from the message body with a blank line. When Thunderbird saves
the message the blank line separating the headers from the body includes a
CR. The patch detection is failing because a CRLF doesn't match /^$/. Fix
this by allowing a CR to exist on the separating line.

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agot0022: replace non-portable literal CR
Stephen Boyd [Tue, 26 Jan 2010 00:33:58 +0000 (16:33 -0800)]
t0022: replace non-portable literal CR

We shouldn't have literal CR's in tests as they aren't portable.

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agotests: consolidate CR removal/addition functions
Stephen Boyd [Tue, 26 Jan 2010 00:33:57 +0000 (16:33 -0800)]
tests: consolidate CR removal/addition functions

append_cr(), remove_cr(), q_to_nul() and q_to_cr() are defined in multiple
tests. Consolidate them into test-lib.sh so we can stop redefining them.
The use of remove_cr() in t0020 to test for a CR is replaced with a new
function has_cr() to accurately reflect what is intended (the output of
remove_cr() was being thrown away).

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agogrep: expose "status-only" feature via -q
Junio C Hamano [Mon, 25 Jan 2010 23:37:23 +0000 (15:37 -0800)]
grep: expose "status-only" feature via -q

Teach "git grep" a new "-q" option to report the presense of a match via
its exit status without showing any output, similar to how "grep -q"
works.  Internally "grep" engine already knew this "status-only" mode of
operation because it needed to grep inside log message to filter commits
when called from the "git log" machinery, and this patch only exposes it
to the command line tool.

A somewhat unfair benchmark in the Linux kernel directory shows a dramatic
improvement:

    (with patch)
    $ time ../git.git/git grep -q linux HEAD ; echo $?

    real    0m0.030s
    user    0m0.004s
    sys     0m0.004s
    0

    (without patch)
    $ time git grep linux HEAD >/dev/null; echo $?

    real    0m4.432s
    user    0m4.272s
    sys     0m0.076s
    0

This is "somewhat unfair" because I knew a file with such a string comes
very early in the tree traversal (namely, ".gitignore").

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agocommit-tree: remove unused #define
Stephen Boyd [Mon, 25 Jan 2010 07:05:43 +0000 (23:05 -0800)]
commit-tree: remove unused #define

Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agot5541-http-push: make grep expression check for one line only
Tay Ray Chuan [Mon, 25 Jan 2010 07:42:23 +0000 (15:42 +0800)]
t5541-http-push: make grep expression check for one line only

Don't feed a multiple-line pattern to grep and expect the them to match
with lines in order.

Simplify the grep expressions in the non-fast-forward tests to check
only for the first line of the non-fast-forward warning - having that
line should be enough assurance that the full warning is printed.

Signed-off-by: Tay Ray Chuan <rctay89@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
14 years agorebase: replace antiquated sed invocation
Stephen Boyd [Mon, 25 Jan 2010 07:06:31 +0000 (23:06 -0800)]
rebase: replace antiquated sed invocation

Use the modern form of printing a commit subject instead of piping
the output of rev-list to sed.

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>