Code

git.git
18 years agoUse git-update-ref to delete a tag instead of rm()ing the ref file.
Christian Couder [Sun, 1 Oct 2006 20:16:22 +0000 (22:16 +0200)]
Use git-update-ref to delete a tag instead of rm()ing the ref file.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix refs.c;:repack_without_ref() clean-up path
Junio C Hamano [Sun, 1 Oct 2006 18:41:00 +0000 (11:41 -0700)]
Fix refs.c;:repack_without_ref() clean-up path

The function repack_without_ref() passes a lock-file structure
on the stack to hold_lock_file_for_update(), which in turn
registers it to be cleaned up via atexit().  This is a big
no-no.

This is the same bug James Bottomley fixed with commit
31f584c242e7af28018ff920b6c8d1952beadbd4.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoClean up "git-branch.sh" and add remove recursive dir test cases.
Christian Couder [Sun, 1 Oct 2006 12:38:18 +0000 (14:38 +0200)]
Clean up "git-branch.sh" and add remove recursive dir test cases.

Now that directory recursive remove works in the core C code, we
don't need to do it in "git-branch.sh".

Also add test cases to check that directory recursive remove will
continue to work.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix a remove_empty_dir_recursive problem.
Christian Couder [Sun, 1 Oct 2006 12:36:49 +0000 (14:36 +0200)]
Fix a remove_empty_dir_recursive problem.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoshow-ref --hash=len, --abbrev=len, and --abbrev
Junio C Hamano [Sun, 1 Oct 2006 07:27:27 +0000 (00:27 -0700)]
show-ref --hash=len, --abbrev=len, and --abbrev

This teaches show-ref to abbreviate the object name.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-branch: remove D/F check done by hand.
Junio C Hamano [Sat, 30 Sep 2006 22:21:28 +0000 (15:21 -0700)]
git-branch: remove D/F check done by hand.

Now ref creation codepath in lock_ref_sha1() and friends notices
the directory/file conflict situation, we do not do this by hand
in git-branch anymore.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agodelete_ref(): delete packed ref
Junio C Hamano [Sat, 30 Sep 2006 22:02:00 +0000 (15:02 -0700)]
delete_ref(): delete packed ref

This implements deletion of a packed ref.  Since it is a very
rare event to delete a ref compared to looking up, creating and
updating, this opts to remove the ref from the packed-ref file
instead of doing any of the filesystem based "negative ref" trick
to optimize the deletion path.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agolock_ref_sha1(): check D/F conflict with packed ref when creating.
Junio C Hamano [Sat, 30 Sep 2006 21:19:25 +0000 (14:19 -0700)]
lock_ref_sha1(): check D/F conflict with packed ref when creating.

This makes the ref locking codepath to notice if an existing ref
overlaps with the ref we are creating.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agolock_ref_sha1(): do not sometimes error() and sometimes die().
Junio C Hamano [Sat, 30 Sep 2006 21:14:31 +0000 (14:14 -0700)]
lock_ref_sha1(): do not sometimes error() and sometimes die().

This cleans up the error path in the function so it does not
die() itself sometimes while signalling an error with NULL some
other times which was inconsistent and confusing.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agorefs: minor restructuring of cached refs data.
Junio C Hamano [Sat, 30 Sep 2006 19:37:37 +0000 (12:37 -0700)]
refs: minor restructuring of cached refs data.

Once we read packed and loose refs, for_each_ref() and friends
kept using them even after write_ref_sha1() and delete_ref()
changed the refs.  This adds invalidate_cached_refs() as a way
to flush the cached information.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoref locking: allow 'foo' when 'foo/bar' used to exist but not anymore.
Junio C Hamano [Sat, 30 Sep 2006 09:25:30 +0000 (02:25 -0700)]
ref locking: allow 'foo' when 'foo/bar' used to exist but not anymore.

It is normal to have .git/refs/heads/foo directory which is
empty after the last branch whose name starts with foo/ is
removed.  Make sure we notice this case and allow creation of
branch foo by removing the empty directory.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoUncomment test case: git branch c/d should barf if branch c exists.
Christian Couder [Thu, 28 Sep 2006 05:02:00 +0000 (07:02 +0200)]
Uncomment test case: git branch c/d should barf if branch c exists.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoWhen creating branch c/d check that branch c does not already exists.
Christian Couder [Thu, 28 Sep 2006 05:00:38 +0000 (07:00 +0200)]
When creating branch c/d check that branch c does not already exists.

With packed refs, there may not be a ".git/refs/heads/c" file
when branch c exists. And currently in this case, there is no check
to prevent creation of branch c/d.

This should probably be rewritten in C and done after the ref lock
has been taken to make sure no race exists though.

This is mainly to make all test cases in "t3210-pack-refs.sh" work.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAdd pack-refs and show-ref test cases.
Christian Couder [Thu, 28 Sep 2006 04:58:03 +0000 (06:58 +0200)]
Add pack-refs and show-ref test cases.

Some of these test cases are from Junio.
One test case is commented out because it doesn't work right now.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'master' into lj/refs
Junio C Hamano [Thu, 28 Sep 2006 05:23:12 +0000 (22:23 -0700)]
Merge branch 'master' into lj/refs

* master: (72 commits)
  runstatus: do not recurse into subdirectories if not needed
  grep: fix --fixed-strings combined with expression.
  grep: free expressions and patterns when done.
  Corrected copy-and-paste thinko in ignore executable bit test case.
  An illustration of rev-list --parents --pretty=raw
  Allow git-checkout when on a non-existant branch.
  gitweb: Decode long title for link tooltips
  git-svn: Fix fetch --no-ignore-externals with GIT_SVN_NO_LIB=1
  Ignore executable bit when adding files if filemode=0.
  Remove empty ref directories that prevent creating a ref.
  Use const for interpolate arguments
  git-archive: update documentation
  Deprecate merge-recursive.py
  gitweb: fix over-eager application of esc_html().
  Allow '(no author)' in git-svn's authors file.
  Allow 'svn fetch' on '(no date)' revisions in Subversion.
  git-repack: allow git-repack to run in subdirectory
  Remove upload-tar and make git-tar-tree a thin wrapper to git-archive
  git-tar-tree: Move code for git-archive --format=tar to archive-tar.c
  git-tar-tree: Remove duplicate git_config() call
  ...

18 years agoMerge branch 'cc/branch-test'
Junio C Hamano [Thu, 28 Sep 2006 05:17:01 +0000 (22:17 -0700)]
Merge branch 'cc/branch-test'

* cc/branch-test:
  Remove empty ref directories that prevent creating a ref.

18 years agoMerge branch 'jc/for-each-ref' into jc/ref-locking
Junio C Hamano [Thu, 28 Sep 2006 05:00:54 +0000 (22:00 -0700)]
Merge branch 'jc/for-each-ref' into jc/ref-locking

* jc/for-each-ref:
  git-for-each-ref: improve the documentation on scripting modes

18 years agoMerge branch 'sp/void'
Junio C Hamano [Thu, 28 Sep 2006 04:51:29 +0000 (21:51 -0700)]
Merge branch 'sp/void'

* sp/void:
  Allow git-checkout when on a non-existant branch.

18 years agoMerge branch 'jc/whitespace'
Junio C Hamano [Thu, 28 Sep 2006 04:50:42 +0000 (21:50 -0700)]
Merge branch 'jc/whitespace'

* jc/whitespace:
  git-apply: second war on whitespace.
  diff.c: second war on whitespace.

18 years agoMerge branch 'jc/repack'
Junio C Hamano [Thu, 28 Sep 2006 04:46:07 +0000 (21:46 -0700)]
Merge branch 'jc/repack'

* jc/repack:
  git-repack: allow git-repack to run in subdirectory
  repack: use only pack-objects, not rev-list.

18 years agoMerge branch 'jc/deprecate-recursive'
Junio C Hamano [Thu, 28 Sep 2006 04:41:24 +0000 (21:41 -0700)]
Merge branch 'jc/deprecate-recursive'

* jc/deprecate-recursive:
  Deprecate merge-recursive.py

18 years agorunstatus: do not recurse into subdirectories if not needed
Johannes Schindelin [Thu, 28 Sep 2006 00:44:30 +0000 (02:44 +0200)]
runstatus: do not recurse into subdirectories if not needed

This speeds up the case when you run git-status, having an untracked
subdirectory containing huge amounts of files.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogrep: fix --fixed-strings combined with expression.
Junio C Hamano [Wed, 27 Sep 2006 23:42:53 +0000 (16:42 -0700)]
grep: fix --fixed-strings combined with expression.

"git grep --fixed-strings -e GIT --and -e VERSION .gitignore"
misbehaved because we did not notice this needs to grab lines
that have the given two fixed strings at the same time.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogrep: free expressions and patterns when done.
Junio C Hamano [Wed, 27 Sep 2006 23:27:10 +0000 (16:27 -0700)]
grep: free expressions and patterns when done.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoreceive-pack: call setup_ident before git_config
Junio C Hamano [Wed, 27 Sep 2006 20:47:21 +0000 (13:47 -0700)]
receive-pack: call setup_ident before git_config

Otherwise we would end up getting values from Gecos which is often not
what people would want.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoCorrected copy-and-paste thinko in ignore executable bit test case.
Shawn Pearce [Wed, 27 Sep 2006 15:18:49 +0000 (11:18 -0400)]
Corrected copy-and-paste thinko in ignore executable bit test case.

This test should be testing update-index --add, not git-add as the
latter is implemented in terms of the former.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAn illustration of rev-list --parents --pretty=raw
Junio C Hamano [Wed, 27 Sep 2006 19:34:37 +0000 (12:34 -0700)]
An illustration of rev-list --parents --pretty=raw

This script creates two separate histories, A and B, each of
which does:

      (A0, B0): create fileA and subdir/fileB
      (A1, B1): modify fileA
      (A2, B2): modify subdir/fileB

and then grafts them together to make B0 a child of A2.  So
the final history looks like (time flows from top to bottom):

true parent touches subdir?

A0 none yes (creates it)
        A1      A0 no
        A2 A1 yes
        B0 none yes (different from what's in A2)
        B1 B0 no
        B2 B1 yes

"git rev-list --parents --pretty=raw B2" would give "fake"
parents on the "commit " header lines while "parent " header
lines show the parent as recorded in the commit object (i.e. B0
appears to have A2 as its parent on "commit " header but there
is no "parent A2" header line in it).

When you have path limiters, we simplify history to omit
commits that do not affect the specified paths.

So "git rev-list --parents --pretty=raw B2 subdir" would return
"B2 B0 A2 A0" (because B1 and A1 do not touch the path).  When
it does so, the "commit " header lines have "fake" parents
(i.e. B2 appears to have B0 as its parent on "commit " header),
but you can still get the true parents by looking at "parent "
header.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoTeach receive-pack about ref-log
Junio C Hamano [Wed, 27 Sep 2006 09:40:06 +0000 (02:40 -0700)]
Teach receive-pack about ref-log

This converts receive-pack to use the standard ref locking code
instead of its own.  As a side effect, it automatically records
the "push" event to ref-log if enabled.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoupdate a few Porcelain-ish for ref lock safety.
Junio C Hamano [Wed, 27 Sep 2006 09:06:31 +0000 (02:06 -0700)]
update a few Porcelain-ish for ref lock safety.

This updates the use of git-update-ref in git-branch, git-tag
and git-commit to make them safer in a few corner cases as
demonstration.

 - git-tag makes sure that the named tag does not exist, allows
   you to edit tag message and then creates the tag.  If a tag
   with the same name was created by somebody else in the
   meantime, it used to happily overwrote it.  Now it notices
   the situation.

 - git-branch -d and git-commit (for the initial commit) had the
   same issue but with smaller race window, which is plugged
   with this.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoupdate-ref: -d flag and ref creation safety.
Junio C Hamano [Wed, 27 Sep 2006 08:58:57 +0000 (01:58 -0700)]
update-ref: -d flag and ref creation safety.

This adds -d flag to update-ref to allow safe deletion of ref.
Before deleting it, the command checks if the given <oldvalue>
still matches the value the caller thought the ref contained.

Similarly, it also accepts 0{40} or an empty string as <oldvalue>
to allow safe creation of a new ref.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoClean-up lock-ref implementation
Junio C Hamano [Wed, 27 Sep 2006 08:09:18 +0000 (01:09 -0700)]
Clean-up lock-ref implementation

This drops "mustexist" parameter lock_ref_sha1() and lock_any_ref_forupdate()
functions take.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/lt-ref2-with-lt-refs' into jc/ref-locking
Junio C Hamano [Wed, 27 Sep 2006 07:53:29 +0000 (00:53 -0700)]
Merge branch 'jc/lt-ref2-with-lt-refs' into jc/ref-locking

* jc/lt-ref2-with-lt-refs:
  Fix show-ref usage for --dereference.
  Document git-show-ref [-s|--hash] option.
  Add man page for git-show-ref
  gitignore: git-show-ref is a generated file.
  Use Linus' show ref in "git-branch.sh".
  Add [-s|--hash] option to Linus' show-ref.
  Teach "git checkout" to use git-show-ref
  Add "git show-ref" builtin command

18 years agoMerge branch 'jc/for-each-ref-with-lt-refs' into jc/ref-locking
Junio C Hamano [Wed, 27 Sep 2006 07:53:16 +0000 (00:53 -0700)]
Merge branch 'jc/for-each-ref-with-lt-refs' into jc/ref-locking

* jc/for-each-ref-with-lt-refs:
  Add git-for-each-ref: helper for language bindings

18 years agoAllow git-checkout when on a non-existant branch.
Shawn Pearce [Mon, 25 Sep 2006 05:24:38 +0000 (01:24 -0400)]
Allow git-checkout when on a non-existant branch.

I've seen some users get into situtations where their HEAD
symbolic-ref is pointing at a non-existant ref.  (Sometimes this
happens during clone when the remote repository lacks a 'master'
branch.)  If this happens the user is unable to use git-checkout
to switch branches as there is no prior commit to merge from.

So instead of giving the user low-level errors about how HEAD
can't be resolved and how not a single revision was given change
the type of checkout to be a force and go through with the user's
request anyway.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Decode long title for link tooltips
Yasushi SHOJI [Wed, 27 Sep 2006 03:04:10 +0000 (12:04 +0900)]
gitweb: Decode long title for link tooltips

This is a simple one liner to decode long title string in perl's
internal form to utf-8 for link tooltips.

This is not crucial if the commit message is all in ASCII, however, if
you decide to use other encoding, such as UTF-8, tooltips ain't
readable any more.

Signed-off-by: Yasushi SHOJI <yashi@atmark-techno.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-svn: Fix fetch --no-ignore-externals with GIT_SVN_NO_LIB=1
Sergey Vlasov [Tue, 26 Sep 2006 09:42:55 +0000 (13:42 +0400)]
git-svn: Fix fetch --no-ignore-externals with GIT_SVN_NO_LIB=1

When using Subversion 1.3.1 without Perl bindings (GIT_SVN_NO_LIB=1),
"git-svn fetch --no-ignore-externals" fails with errors like:

  Tree (.../.git/svn/git-svn/tree) is not clean:
  X      directory_with_external

In this case the 'X' lines in the "svn status" output are not a sign
of unclean tree, and therefore should be ignored.

Signed-off-by: Sergey Vlasov <vsu@altlinux.ru>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoIgnore executable bit when adding files if filemode=0.
Shawn Pearce [Wed, 27 Sep 2006 05:21:19 +0000 (01:21 -0400)]
Ignore executable bit when adding files if filemode=0.

If the user has configured core.filemode=0 then we shouldn't set
the execute bit in the index when adding a new file as the user
has indicated that the local filesystem can't be trusted.

This means that when adding files that should be marked executable
in a repository with core.filemode=0 the user must perform a
'git update-index --chmod=+x' on the file before committing the
addition.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRemove empty ref directories that prevent creating a ref.
Christian Couder [Tue, 26 Sep 2006 05:23:37 +0000 (07:23 +0200)]
Remove empty ref directories that prevent creating a ref.

This patch also adds test cases from Linus and Junio.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoUse const for interpolate arguments
Alex Riesen [Mon, 25 Sep 2006 21:19:00 +0000 (23:19 +0200)]
Use const for interpolate arguments

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-archive: update documentation
Franck Bui-Huu [Mon, 25 Sep 2006 10:31:52 +0000 (12:31 +0200)]
git-archive: update documentation

This patch documents zip backend options.

It also adds git-archive command into the main git manual
page.

Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoDeprecate merge-recursive.py
Junio C Hamano [Mon, 25 Sep 2006 02:49:47 +0000 (19:49 -0700)]
Deprecate merge-recursive.py

This renames merge-recursive written in Python to merge-recursive-old,
and makes merge-recur as a synonym to merge-recursive.  We do not remove
merge-recur yet, but we will remove merge-recur and merge-recursive-old
in a few releases down the road.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: fix over-eager application of esc_html().
Petr Baudis [Sun, 24 Sep 2006 21:57:40 +0000 (14:57 -0700)]
gitweb: fix over-eager application of esc_html().

Contents of %diffinfo hash should be quoted upon output but kept
unquoted internally.  Later users of this hash expect filenames
to be filenames, not HTML gibberish.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAllow '(no author)' in git-svn's authors file.
Shawn Pearce [Mon, 25 Sep 2006 03:04:55 +0000 (23:04 -0400)]
Allow '(no author)' in git-svn's authors file.

When trying to import an SVN revision which has no author the Git
user may desire to relabel '(no author)' to another name and email
address with their svn.authorsfile.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAllow 'svn fetch' on '(no date)' revisions in Subversion.
Shawn Pearce [Mon, 25 Sep 2006 02:50:15 +0000 (22:50 -0400)]
Allow 'svn fetch' on '(no date)' revisions in Subversion.

Added --ignore-nodate to allow 'git svn fetch' to import revisions
from Subversion which have '(no date)' listed as the date of the
revision.  By default 'git svn fetch' will crash with an error
when encountering such a revision.  The user may restart the fetch
operation by adding --ignore-nodate if they want to continue tracking
that repository.

I'm not entirely sure why a centralized version control system such
as Subversion permits revisions to be created with absolutely no
date/time associated with it but it apparently is possible as one
of the Subversion repositories that I'm tracking with 'git svn'
created such a revision on '(no date)' and by '(no user)'.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-repack: allow git-repack to run in subdirectory
Jeff King [Mon, 25 Sep 2006 02:31:11 +0000 (22:31 -0400)]
git-repack: allow git-repack to run in subdirectory

Now that we explicitly create all tmpfiles below $GIT_DIR, there's no reason
to care about which directory we're in.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'sk/svnimport'
Junio C Hamano [Mon, 25 Sep 2006 03:00:59 +0000 (20:00 -0700)]
Merge branch 'sk/svnimport'

* sk/svnimport:
  git-svnimport: Parse log message for Signed-off-by: lines

18 years agoMerge early parts of branch 'np/pack'
Junio C Hamano [Mon, 25 Sep 2006 02:58:30 +0000 (19:58 -0700)]
Merge early parts of branch 'np/pack'

18 years agoRemove upload-tar and make git-tar-tree a thin wrapper to git-archive
Junio C Hamano [Sun, 24 Sep 2006 21:42:01 +0000 (14:42 -0700)]
Remove upload-tar and make git-tar-tree a thin wrapper to git-archive

The command now issues a big deprecation warning message and runs
git-archive command with appropriate arguments.

git-tar-tree $tree_ish $base always forces $base to be the leading
directory name, so the --prefix parameter passed internally to
git-archive is a slash appended to it, i.e. "--prefix=$base/".

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-tar-tree: Move code for git-archive --format=tar to archive-tar.c
Rene Scharfe [Sun, 24 Sep 2006 15:31:10 +0000 (17:31 +0200)]
git-tar-tree: Move code for git-archive --format=tar to archive-tar.c

This patch doesn't change any functionality, it only moves code around.  It
makes seeing the few remaining lines of git-tar-tree code easier. ;-)

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-tar-tree: Remove duplicate git_config() call
Rene Scharfe [Sun, 24 Sep 2006 15:30:44 +0000 (17:30 +0200)]
git-tar-tree: Remove duplicate git_config() call

generate_tar() eventually calls write_tar_archive() which does all the
"real" work and which also calls git_config(git_tar_config).  We only
need to do this once.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/filter-commit'
Junio C Hamano [Mon, 25 Sep 2006 02:53:52 +0000 (19:53 -0700)]
Merge branch 'jc/filter-commit'

* jc/filter-commit:
  git log: Unify header_filter and message_filter into one.
  Update grep internal for grepping only in head/body
  git-log --author and --committer are not left-anchored by default
  rev-list: fix segfault with --{author,committer,grep}
  revision traversal: --author, --committer, and --grep.
  revision traversal: prepare for commit log match.
  builtin-grep: make pieces of it available as library.

18 years agoMerge branch 'sb/branch-attributes'
Junio C Hamano [Mon, 25 Sep 2006 02:20:26 +0000 (19:20 -0700)]
Merge branch 'sb/branch-attributes'

* sb/branch-attributes:
  Add test for the default merges in fetch.
  fetch: get the remote branches to merge from the branch properties
  Add t5510 to test per branch configuration affecting git-fetch.
  Fetch: default remote repository from branch properties

18 years agoMerge branch 'jl/daemon'
Junio C Hamano [Sun, 24 Sep 2006 23:58:34 +0000 (16:58 -0700)]
Merge branch 'jl/daemon'

* jl/daemon:
  Add virtualization support to git-daemon

18 years agoAdd test for the default merges in fetch.
Santi Béjar [Sat, 23 Sep 2006 20:55:35 +0000 (22:55 +0200)]
Add test for the default merges in fetch.

[jc: with minor fix-ups]

Signed-off-by: Santi Béjar <sbejar@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agofetch: get the remote branches to merge from the branch properties
Santi Béjar [Sat, 23 Sep 2006 20:53:04 +0000 (22:53 +0200)]
fetch: get the remote branches to merge from the branch properties

If in branch "foo" and this in config:

[branch "foo"]
      merge=bar

"git fetch": fetch from the default repository and program the "bar"
             branch to be merged with pull.

Signed-off-by: Santi Béjar <sbejar@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAdd t5510 to test per branch configuration affecting git-fetch.
Junio C Hamano [Sat, 23 Sep 2006 10:40:17 +0000 (03:40 -0700)]
Add t5510 to test per branch configuration affecting git-fetch.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFetch: default remote repository from branch properties
Santi Béjar [Sat, 23 Sep 2006 10:05:43 +0000 (12:05 +0200)]
Fetch: default remote repository from branch properties

If in branch "foo" and this in config:

[branch "foo"]
       remote=bar

"git fetch" = "git fetch bar"
"git  pull" = "git pull  bar"

Signed-off-by: Santi Béjar <sbejar@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-apply: second war on whitespace.
Junio C Hamano [Sat, 23 Sep 2006 07:37:19 +0000 (00:37 -0700)]
git-apply: second war on whitespace.

This makes --whitespace={warn,error,strip} option to also notice
the leading whitespace errors in addition to the trailing
whitespace errors.  Spaces that are followed by a tab in indent
are detected as errors, and --whitespace=strip option fixes them.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agodiff.c: second war on whitespace.
Junio C Hamano [Sat, 23 Sep 2006 05:48:39 +0000 (22:48 -0700)]
diff.c: second war on whitespace.

This adds DIFF_WHITESPACE color class (default = reverse red) to
colored diff output to let you catch common whitespace errors.

 - trailing whitespaces at the end of line
 - a space followed by a tab in the indent

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agobuiltin-upload-archive.c broken on openbsd
Randal L. Schwartz [Sun, 24 Sep 2006 00:20:45 +0000 (17:20 -0700)]
builtin-upload-archive.c broken on openbsd

Looks like ctype again. Gotta be careful with that on BSD releases:

    $ gmake prefix=/opt/git all
    GIT_VERSION = 1.4.2.GIT
    gcc -o builtin-upload-archive.o -c -g -O2 -Wall -I/usr/local/include -DSHA1_HEADER='<openssl/sha.h>' -DNO_STRCASESTR builtin-upload-archive.c
    In file included from /usr/include/sys/poll.h:54,
                     from builtin-upload-archive.c:11:
    /usr/include/ctype.h:68: error: syntax error before ']' token
    /usr/include/ctype.h:69: error: syntax error before ']' token
    ...
    /usr/include/sys/poll.h:53:1: unterminated #ifndef
    /usr/include/sys/poll.h:28:1: unterminated #ifndef
    gmake: *** [builtin-upload-archive.o] Error 1

This fixes it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Consolidate escaping/validation of query string
Petr Baudis [Sat, 23 Sep 2006 22:18:41 +0000 (00:18 +0200)]
gitweb: Consolidate escaping/validation of query string

Consider:

http://repo.or.cz/?p=glibc-cvs.git;a=tree;h=2609cb0411389325f4ee2854cc7159756eb0671e;hb=2609cb0411389325f4ee2854cc7159756eb0671e

(click on the funny =__ify file)

We ought to handle anything in filenames and I actually see no reason why
we don't, modulo very little missing escaping that this patch hopefully
also fixes.

I have also made esc_param() escape [?=&;]. Not escaping [&;] was downright
buggy and [?=] just feels better escaped. ;-) YMMV.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoDeprecate git-resolve.sh
Petr Baudis [Sat, 23 Sep 2006 22:25:19 +0000 (00:25 +0200)]
Deprecate git-resolve.sh

Seriously, is anyone still using this thing? It's collecting dust and
blocking the name for something potentially useful like a tool for
user-friendly marking of resolved conflicts or resolving index conflicts.

We've loved you when Git was young, now thank you and please go away. ;-)

This makes git-resolve.sh print a big deprecation warning and sleep a bit
for extra annoyance. It should be removed completely after the next release.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRename builtin-zip-tree.c to archive-zip.c
Rene Scharfe [Sat, 23 Sep 2006 15:06:35 +0000 (17:06 +0200)]
Rename builtin-zip-tree.c to archive-zip.c

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRemove git-zip-tree
Rene Scharfe [Sat, 23 Sep 2006 15:06:11 +0000 (17:06 +0200)]
Remove git-zip-tree

git-zip-tree can be safely removed because it was never part of a formal
release.  This patch makes 'git-archive --format=zip' the one and only git
ZIP file creation command.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agomany cleanups to sha1_file.c
Nicolas Pitre [Thu, 21 Sep 2006 04:05:37 +0000 (00:05 -0400)]
many cleanups to sha1_file.c

Those cleanups are mainly to set the table for the support of deltas
with base objects referenced by offsets instead of sha1.  This means
that many pack lookup functions are converted to take a pack/offset
tuple instead of a sha1.

This eliminates many struct pack_entry usages since this structure
carried redundent information in many cases, and it increased stack
footprint needlessly for a couple recursively called functions that used
to declare a local copy of it for every recursion loop.

In the process, packed_object_info_detail() has been reorganized as well
so to look much saner and more amenable to deltas with offset support.

Finally the appropriate adjustments have been made to functions that
depend on the above changes.  But there is no functionality changes yet
simply some code refactoring at this point.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agolock_ref_sha1_basic: remove unused parameter "plen".
Junio C Hamano [Sat, 23 Sep 2006 04:41:49 +0000 (21:41 -0700)]
lock_ref_sha1_basic: remove unused parameter "plen".

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agopack-refs: fix git_path() usage.
Junio C Hamano [Sat, 23 Sep 2006 04:31:40 +0000 (21:31 -0700)]
pack-refs: fix git_path() usage.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agomove pack creation to version 3
Nicolas Pitre [Thu, 21 Sep 2006 04:11:59 +0000 (00:11 -0400)]
move pack creation to version 3

It's been quite a while now that GIT is able to read version 3 packs.
Let's create them at last.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-svnimport: Parse log message for Signed-off-by: lines
Sasha Khapyorsky [Tue, 5 Sep 2006 18:46:11 +0000 (21:46 +0300)]
git-svnimport: Parse log message for Signed-off-by: lines

This add '-S' option. When specified svn-import will try to parse
commit message for 'Signed-off-by: ...' line, and if found will use
the name and email address extracted at first occurrence as this commit
author name and author email address. Committer name and email are
extracted in usual way.

Signed-off-by: Sasha Khapyorsky <sashak@voltaire.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoQuote arguments to tr in test-lib
Robin Rosenberg [Fri, 22 Sep 2006 22:35:20 +0000 (00:35 +0200)]
Quote arguments to tr in test-lib

When there are single-character filenames in the test directory,
the shell tries to expand regexps meant for tr.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix snapshot link in tree view
Petr Baudis [Fri, 22 Sep 2006 23:21:20 +0000 (01:21 +0200)]
Fix snapshot link in tree view

It would just give HEAD snapshot instead of one of the particular tree.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Fix @git_base_url_list usage
Petr Baudis [Fri, 22 Sep 2006 23:15:18 +0000 (01:15 +0200)]
gitweb: Fix @git_base_url_list usage

As it is now, that array was never used because the customurl accessor was
broken and ''unless @url_list'' never happenned.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Fix tree link associated with each commit log entry.
Petr Baudis [Fri, 22 Sep 2006 23:56:43 +0000 (16:56 -0700)]
gitweb: Fix tree link associated with each commit log entry.

The link forgot to have hb parameter and the resulting tree view
failed to show the navbar for that commit.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix buggy ref recording
Petr Baudis [Fri, 22 Sep 2006 23:08:45 +0000 (01:08 +0200)]
Fix buggy ref recording

There is a format string vulnerability introduced with the packed refs
file format.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake path in tree view look nicer
Petr Baudis [Fri, 22 Sep 2006 23:00:12 +0000 (01:00 +0200)]
Make path in tree view look nicer

Based on talk on the IRC with Junio some evenings ago, I've updated the
path showing in tree view to look better and sent updated patches
privately, but it seems the old version ended up being used, so here's
the new one again.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Relabel "head" as "HEAD"
Petr Baudis [Fri, 22 Sep 2006 01:19:53 +0000 (03:19 +0200)]
gitweb: Relabel "head" as "HEAD"

"head" is a reference in refs/heads/, while those labels mean HEAD,
the latest revision of the default branch.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Rename "plain" labels to "raw"
Petr Baudis [Fri, 22 Sep 2006 01:19:50 +0000 (03:19 +0200)]
gitweb: Rename "plain" labels to "raw"

I don't have much preference either way and as far as I'm concerned, it may
go the other way as well. Consistency is what is important.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Link to associated tree from a particular log item in full log view
Petr Baudis [Fri, 22 Sep 2006 01:19:48 +0000 (03:19 +0200)]
gitweb: Link to associated tree from a particular log item in full log view

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Link to latest tree from the head line in heads list
Petr Baudis [Fri, 22 Sep 2006 01:19:46 +0000 (03:19 +0200)]
gitweb: Link to latest tree from the head line in heads list

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Link to tree instead of snapshot in shortlog
Petr Baudis [Fri, 22 Sep 2006 01:19:44 +0000 (03:19 +0200)]
gitweb: Link to tree instead of snapshot in shortlog

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: More per-view navigation bar links
Petr Baudis [Fri, 22 Sep 2006 01:19:41 +0000 (03:19 +0200)]
gitweb: More per-view navigation bar links

Navigation bars in various views were empty or missed important items that
should have been there, e.g. getting a snapshot in tree view or log of
ancestry in commit view...

This feeble patch attempts to consolidate that.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Link (HEAD) tree for each project from projects list
Petr Baudis [Thu, 21 Sep 2006 23:58:40 +0000 (01:58 +0200)]
gitweb: Link (HEAD) tree for each project from projects list

Current projects list is oriented on easily getting "what's new"
information. But when already using gitweb as an interface to something,
I personally find myself to _much_ more frequently wanting to rather
see "what's in" (or "what's new in") and it's quite annoying to have to
go through the summary page (which is also rather expensive to generate)
just to get there.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoshow-branch: mark active branch with a '*' again
Johannes Schindelin [Thu, 21 Sep 2006 22:07:01 +0000 (00:07 +0200)]
show-branch: mark active branch with a '*' again

This was lost in the packed-ref updates. The original test was a bit
dubious, so I cleaned that up, too. It fixes the case when the current HEAD
is refs/heads/bla/master: the original test was true for both bla/master
_and_ master.

However, it shares a hard-to-fix bug with the original test: if the current
HEAD is refs/heads/master, and there is a branch refs/heads/heads/master,
then both are marked active.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix git-update-index --again
Johannes Schindelin [Thu, 21 Sep 2006 21:29:59 +0000 (23:29 +0200)]
Fix git-update-index --again

It called read_ref(git_path(..)..), where read_ref does the git_path()
stuff itself.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Make git_get_hash_by_path check type if provided
Jakub Narebski [Thu, 21 Sep 2006 16:09:12 +0000 (18:09 +0200)]
gitweb: Make git_get_hash_by_path check type if provided

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-for-each-ref: improve the documentation on scripting modes
Andy Whitcroft [Thu, 21 Sep 2006 09:19:17 +0000 (10:19 +0100)]
git-for-each-ref: improve the documentation on scripting modes

When reading the synopsis for git-for-each-ref it is easy to miss
the obvious power of --shell and family.  Call this feature out in
the primary paragragh.  Also add more description to the examples
to indicate which features we are demonstrating.  Finally add a
very simple eval based example in addition to the very complex one
to give a gentler introduction.

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix showing of path in tree view
Petr Baudis [Thu, 21 Sep 2006 00:05:50 +0000 (02:05 +0200)]
Fix showing of path in tree view

This patch fixes two things - links to all path elements except the last
one were broken since gitweb does not like the trailing slash in them, and
the root tree was not reachable from the subdirectory view.

To compensate for the one more slash in the front, the trailing slash is
not there anymore. ;-) I don't care if it stays there though.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: fix display of trees via PATH_INFO.
Martin Waitz [Thu, 21 Sep 2006 07:48:21 +0000 (09:48 +0200)]
gitweb: fix display of trees via PATH_INFO.

When adding a / to the URL, git should display the corresponding
tree object, but it has to remove the / first.

Signed-off-by: Martin Waitz <tali@admingilde.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'lt/refs' into jc/lt-ref2-with-lt-refs
Junio C Hamano [Thu, 21 Sep 2006 07:40:28 +0000 (00:40 -0700)]
Merge branch 'lt/refs' into jc/lt-ref2-with-lt-refs

* lt/refs: (58 commits)
  git-pack-refs --prune
  pack-refs: do not pack symbolic refs.
  Tell between packed, unpacked and symbolic refs.
  Add callback data to for_each_ref() family.
  symbolit-ref: fix resolve_ref conversion.
  Fix broken sha1 locking
  fsck-objects: adjust to resolve_ref() clean-up.
  gitignore: git-pack-refs is a generated file.
  wt-status: use simplified resolve_ref to find current branch
  Fix t1400-update-ref test minimally
  Enable the packed refs file format
  Make ref resolution saner
  Add support for negative refs
  Start handling references internally as a sorted in-memory list
  gitweb fix validating pg (page) parameter
  git-repack(1): document --window and --depth
  git-apply(1): document --unidiff-zero
  gitweb: fix warnings in PATH_INFO code and add export_ok/strict_export
  upload-archive: monitor child communication even more carefully.
  gitweb: export options
  ...

18 years agoMerge branch 'lt/refs' into jc/for-each-ref-with-lt-refs
Junio C Hamano [Thu, 21 Sep 2006 07:29:37 +0000 (00:29 -0700)]
Merge branch 'lt/refs' into jc/for-each-ref-with-lt-refs

* lt/refs: (58 commits)
  git-pack-refs --prune
  pack-refs: do not pack symbolic refs.
  Tell between packed, unpacked and symbolic refs.
  Add callback data to for_each_ref() family.
  symbolit-ref: fix resolve_ref conversion.
  Fix broken sha1 locking
  fsck-objects: adjust to resolve_ref() clean-up.
  gitignore: git-pack-refs is a generated file.
  wt-status: use simplified resolve_ref to find current branch
  Fix t1400-update-ref test minimally
  Enable the packed refs file format
  Make ref resolution saner
  Add support for negative refs
  Start handling references internally as a sorted in-memory list
  gitweb fix validating pg (page) parameter
  git-repack(1): document --window and --depth
  git-apply(1): document --unidiff-zero
  gitweb: fix warnings in PATH_INFO code and add export_ok/strict_export
  upload-archive: monitor child communication even more carefully.
  gitweb: export options
  ...

18 years agogit-pack-refs --prune
Junio C Hamano [Thu, 21 Sep 2006 07:06:06 +0000 (00:06 -0700)]
git-pack-refs --prune

"git pack-refs --prune", after successfully packing the existing
refs, removes the loose ref files.  It tries to protect against
race by doing the usual lock_ref_sha1() which makes sure the
contents of the ref has not changed since we last looked at.

Also we do not bother trying to prune what was already packed, and
we do not try pruning symbolic refs.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agopack-refs: do not pack symbolic refs.
Junio C Hamano [Thu, 21 Sep 2006 07:06:05 +0000 (00:06 -0700)]
pack-refs: do not pack symbolic refs.

Now we can tell which one is symbolic and which one is not, it
is easy to do so.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoDocument receive.denyNonFastforwards
Johannes Schindelin [Thu, 21 Sep 2006 00:10:30 +0000 (02:10 +0200)]
Document receive.denyNonFastforwards

[jc: with a fix to config handling in t5400 test, which took
 annoyingly long to diagnose.]

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoreceive-pack: plug memory leak in fast-forward checking code.
Johannes Schindelin [Thu, 21 Sep 2006 00:07:19 +0000 (02:07 +0200)]
receive-pack: plug memory leak in fast-forward checking code.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoTell between packed, unpacked and symbolic refs.
Junio C Hamano [Thu, 21 Sep 2006 05:02:01 +0000 (22:02 -0700)]
Tell between packed, unpacked and symbolic refs.

This adds a "int *flag" parameter to resolve_ref() and makes
for_each_ref() family to call callback function with an extra
"int flag" parameter.  They are used to give two bits of
information (REF_ISSYMREF and REF_ISPACKED) about the ref.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAdd callback data to for_each_ref() family.
Junio C Hamano [Thu, 21 Sep 2006 04:47:42 +0000 (21:47 -0700)]
Add callback data to for_each_ref() family.

This is a long overdue fix to the API for for_each_ref() family
of functions.  It allows the callers to specify a callback data
pointer, so that the caller does not have to use static
variables to communicate with the callback funciton.

The updated for_each_ref() family takes a function of type

int (*fn)(const char *, const unsigned char *, void *)

and a void pointer as parameters, and calls the function with
the name of the ref and its SHA-1 with the caller-supplied void
pointer as parameters.

The commit updates two callers, builtin-name-rev.c and
builtin-pack-refs.c as an example.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agosymbolit-ref: fix resolve_ref conversion.
Junio C Hamano [Thu, 21 Sep 2006 04:10:17 +0000 (21:10 -0700)]
symbolit-ref: fix resolve_ref conversion.

An earlier conversion accidentally hardcoded "HEAD" to be passed to
resolve_ref(), thereby causing git-symbolic-ref command to always
report where the HEAD points at, ignoring the command line parameter.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoadd receive.denyNonFastforwards config variable
Johannes Schindelin [Wed, 20 Sep 2006 23:07:54 +0000 (01:07 +0200)]
add receive.denyNonFastforwards config variable

If receive.denyNonFastforwards is set to true, git-receive-pack will deny
non fast-forwards, i.e. forced updates. Most notably, a push to a repository
which has that flag set will fail.

As a first user, 'git-init-db --shared' sets this flag, since in a shared
setup, you are most unlikely to want forced pushes to succeed.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agosha1_name.c: understand "describe" output as a valid object name
Junio C Hamano [Wed, 20 Sep 2006 23:11:08 +0000 (16:11 -0700)]
sha1_name.c: understand "describe" output as a valid object name

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake hexval() available to others.
Junio C Hamano [Wed, 20 Sep 2006 23:04:46 +0000 (16:04 -0700)]
Make hexval() available to others.

builtin-mailinfo.c has its own hexval implementaiton but it can
share the table-lookup one recently implemented in sha1_file.c

Signed-off-by: Junio C Hamano <junkio@cox.net>