Code

git.git
17 years agoDocumentation/git-branch: new -r to delete remote-tracking branches.
Quy Tonthat [Thu, 21 Dec 2006 04:18:07 +0000 (15:18 +1100)]
Documentation/git-branch: new -r to delete remote-tracking branches.

Signed-off-by: Quy Tonthat <qtonthat@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoFix system header problems on Mac OS X
Terje Sten Bjerkseth [Thu, 21 Dec 2006 01:32:21 +0000 (17:32 -0800)]
Fix system header problems on Mac OS X

For Mac OS X 10.4, _XOPEN_SOURCE defines _POSIX_C_SOURCE which
hides many symbols from the program.

Breakage noticed and initial analysis provided by Randal
L. Schwartz.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'ew/svn-pm'
Junio C Hamano [Thu, 21 Dec 2006 01:20:45 +0000 (17:20 -0800)]
Merge branch 'ew/svn-pm'

* ew/svn-pm:
  git-svn: rename 'commit' command to 'set-tree'
  git-svn: remove support for the svn command-line client
  git-svn: convert to using Git.pm

17 years agodiff documentation: mostly talk about <commit>
Junio C Hamano [Wed, 20 Dec 2006 22:41:54 +0000 (14:41 -0800)]
diff documentation: mostly talk about <commit>

This corrects minor remaining bits that still talked about <tree-ish>;
the Porcelain users (as opposed to plumbers) are mostly interested in
commits so use <commit> consistently and keep a sentence that mentions
that <tree-ish> can be used in place of them.

17 years agoMerge branch 'jc/leftright'
Junio C Hamano [Wed, 20 Dec 2006 22:20:17 +0000 (14:20 -0800)]
Merge branch 'jc/leftright'

* jc/leftright:
  Revert "Make left-right automatic."
  Make left-right automatic.
  Teach all of log family --left-right output.
  rev-list --left-right

17 years agoMerge branch 'jc/blame'
Junio C Hamano [Wed, 20 Dec 2006 21:58:10 +0000 (13:58 -0800)]
Merge branch 'jc/blame'

* jc/blame:
  blame: -b (blame.blankboundary) and --root (blame.showroot)

17 years agoMerge branch 'jc/branch-remove-remote'
Junio C Hamano [Wed, 20 Dec 2006 21:57:59 +0000 (13:57 -0800)]
Merge branch 'jc/branch-remove-remote'

* jc/branch-remove-remote:
  git-branch -d: do not stop at the first failure.
  Teach git-branch to delete tracking branches with -r -d

17 years agoMerge branch 'jc/clone'
Junio C Hamano [Wed, 20 Dec 2006 21:56:14 +0000 (13:56 -0800)]
Merge branch 'jc/clone'

* jc/clone:
  Move "no merge candidate" warning into git-pull
  Use preprocessor constants for environment variable names.
  Do not create $GIT_DIR/remotes/ directory anymore.
  Introduce GIT_TEMPLATE_DIR
  Revert "fix testsuite: make sure they use templates freshly built from the source"
  fix testsuite: make sure they use templates freshly built from the source
  git-clone: lose the traditional 'no-separate-remote' layout
  git-clone: lose the artificial "first" fetch refspec
  git-pull: refuse default merge without branch.*.merge
  git-clone: use wildcard specification for tracking branches

17 years agocompat/inet_ntop: do not use u_int
Junio C Hamano [Wed, 20 Dec 2006 21:45:55 +0000 (13:45 -0800)]
compat/inet_ntop: do not use u_int

It is pointless.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-add: error out when given no arguments.
Junio C Hamano [Wed, 20 Dec 2006 21:06:46 +0000 (13:06 -0800)]
git-add: error out when given no arguments.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'maint'
Junio C Hamano [Wed, 20 Dec 2006 19:24:56 +0000 (11:24 -0800)]
Merge branch 'maint'

* maint:
  GIT 1.4.4.3

17 years agoGIT 1.4.4.3 v1.4.4.3
Junio C Hamano [Wed, 20 Dec 2006 19:23:22 +0000 (11:23 -0800)]
GIT 1.4.4.3

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoclarify some error messages wrt unknown object types
Nicolas Pitre [Wed, 20 Dec 2006 18:34:05 +0000 (13:34 -0500)]
clarify some error messages wrt unknown object types

If ever new object types are added for future extensions then better
have current git version report them as "unknown" instead of
"corrupted".

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agosimplify inclusion of system header files.
Junio C Hamano [Tue, 19 Dec 2006 22:34:12 +0000 (14:34 -0800)]
simplify inclusion of system header files.

This is a mechanical clean-up of the way *.c files include
system header files.

 (1) sources under compat/, platform sha-1 implementations, and
     xdelta code are exempt from the following rules;

 (2) the first #include must be "git-compat-util.h" or one of
     our own header file that includes it first (e.g. config.h,
     builtin.h, pkt-line.h);

 (3) system headers that are included in "git-compat-util.h"
     need not be included in individual C source files.

 (4) "git-compat-util.h" does not have to include subsystem
     specific header files (e.g. expat.h).

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoindex-pack usage of mmap() is unacceptably slower on many OSes other than Linux
Nicolas Pitre [Tue, 19 Dec 2006 15:53:08 +0000 (10:53 -0500)]
index-pack usage of mmap() is unacceptably slower on many OSes other than Linux

It was reported by Randal L. Schwartz <merlyn@stonehenge.com> that
indexing the Linux repository ~150MB pack takes about an hour on OS x
while it's a minute on Linux.  It seems that the OS X mmap()
implementation is more than 2 orders of magnitude slower than the Linux
one.

Linus proposed a patch replacing mmap() with pread() bringing index-pack
performance on OS X in line with the Linux one.  The performances on
Linux also improved by a small margin.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Show '...' links in "summary" view only if there are more items
Robert Fitzsimons [Tue, 19 Dec 2006 12:08:54 +0000 (12:08 +0000)]
gitweb: Show '...' links in "summary" view only if there are more items

Show "..." links in "summary" view to shortlog, heads (if there are
any), and tags (if there are any) only if there are more items to show
than shown already.

This means that "..." link is shown below shortened shortlog if there
are more than 16 commits, "..." link below shortened heads list if
there are more than 16 heads refs (16 branches), "..." link below
shortened tags list if there are more than 16 tags.

Modified patch from Jakub to to apply cleanly to master, also preform
the same "..." link logic to the forks list.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Robert Fitzsimons <robfitz@273k.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoconfig_rename_section: fix FILE* leak
Junio C Hamano [Wed, 20 Dec 2006 05:55:27 +0000 (21:55 -0800)]
config_rename_section: fix FILE* leak

Noticed by SungHyun Nam.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoRemove COLLISION_CHECK from Makefile since it's not used.
Brian Gernhardt [Wed, 20 Dec 2006 03:25:13 +0000 (22:25 -0500)]
Remove COLLISION_CHECK from Makefile since it's not used.

It's rather misleading to have configuration options that don't do
anything.  If someone adds collision checking they might also want to
restore this option.

Signed-off-by: Brian Gernhardt <benji@silverinsanity.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agofix populate-filespec
Junio C Hamano [Wed, 20 Dec 2006 02:26:48 +0000 (18:26 -0800)]
fix populate-filespec

I hand munged the original patch when committing 1510fea78, and
screwed up the conversion.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-cvsserver: fix breakage when calling git merge-file
Eric Wong [Tue, 19 Dec 2006 22:58:20 +0000 (14:58 -0800)]
git-cvsserver: fix breakage when calling git merge-file

In the same vein as 8336afa563fbeff35e531396273065161181f04c,
this fixes the the RCS merge to git-merge-file conversion in
commit e2b70087.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoRevert "Make left-right automatic."
Junio C Hamano [Tue, 19 Dec 2006 10:28:16 +0000 (02:28 -0800)]
Revert "Make left-right automatic."

This reverts commit 5761231975ceffa531d86d9bab0f9a9a370674f6.

Feeding symmetric difference to gitk is so useful, and it is the
same for other graphical Porcelains.  Rather than forcing them
to pass --no-left-right, making it optional.

Noticed and reported by Jeff King.

17 years agoMove "no merge candidate" warning into git-pull
Josef Weidendorfer [Tue, 19 Dec 2006 00:39:07 +0000 (01:39 +0100)]
Move "no merge candidate" warning into git-pull

The warning triggered even when running "git fetch" only
when resulting .git/FETCH_HEAD only contained
branches marked as 'not-for-merge'.

Signed-off-by: Josef Weidendorfer <weidendo@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoUse preprocessor constants for environment variable names.
Junio C Hamano [Tue, 19 Dec 2006 09:28:15 +0000 (01:28 -0800)]
Use preprocessor constants for environment variable names.

We broke the discipline Linus set up to allow compiler help us
avoid typos in environment names in the early days of git over
time.  This defines a handful preprocessor constants for
environment variable names used in relatively core parts of the
system.

I've left out variable names specific to subsystems such as HTTP
and SSL as I do not think they are big problems.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoDo not create $GIT_DIR/remotes/ directory anymore.
Junio C Hamano [Tue, 19 Dec 2006 09:50:37 +0000 (01:50 -0800)]
Do not create $GIT_DIR/remotes/ directory anymore.

Because we do not use --no-separate-remote anymore, there is no
reason to create that directory from the template.

t5510 test is updated to test both $GIT_DIR/remotes/ based
configuration and $GIT_DIR/config variable (credits to
Johannes).

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'jc/test-clone' into jc/clone
Junio C Hamano [Tue, 19 Dec 2006 09:38:18 +0000 (01:38 -0800)]
Merge branch 'jc/test-clone' into jc/clone

* jc/test-clone: (35 commits)
  Introduce GIT_TEMPLATE_DIR
  Revert "fix testsuite: make sure they use templates freshly built from the source"
  fix testsuite: make sure they use templates freshly built from the source
  rerere: fix breakage of resolving.
  Add config example with respect to branch
  Add documentation for show-branch --topics
  make git a bit less cryptic on fetch errors
  make patch_delta() error cases a bit more verbose
  racy-git: documentation updates.
  show-ref: fix --exclude-existing
  parse-remote::expand_refs_wildcard()
  vim syntax: follow recent changes to commit template
  show-ref: fix --verify --hash=length
  show-ref: fix --quiet --verify
  avoid accessing _all_ loose refs in git-show-ref --verify
  git-fetch: Avoid reading packed refs over and over again
  Teach show-branch how to show ref-log data.
  markup fix in svnimport documentation.
  Documentation: new option -P for git-svnimport
  Fix mis-mark-up in git-merge-file.txt documentation
  ...

17 years agoIntroduce GIT_TEMPLATE_DIR
Johannes Schindelin [Tue, 19 Dec 2006 08:18:09 +0000 (09:18 +0100)]
Introduce GIT_TEMPLATE_DIR

Instead of passing --template explicitely to init-db and clone, you can
just set the environment variable GIT_TEMPLATE_DIR.

Also make use of it in the tests, to make sure that the templates are
copied.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoRevert "fix testsuite: make sure they use templates freshly built from the source"
Junio C Hamano [Tue, 19 Dec 2006 09:14:35 +0000 (01:14 -0800)]
Revert "fix testsuite: make sure they use templates freshly built from the source"

This reverts commit 74d20040cafdced657efbf49795183d209a3a07b.
Version from Johannes to introduce GIT_TEMPLATE_DIR is simpler,
although I unconsciously stayed away from introducing yet another
environment variable.

17 years agoblame: -b (blame.blankboundary) and --root (blame.showroot)
Junio C Hamano [Mon, 18 Dec 2006 22:04:38 +0000 (14:04 -0800)]
blame: -b (blame.blankboundary) and --root (blame.showroot)

When blame.blankboundary is set (or -b option is given), commit
object names are blanked out in the "human readable" output
format for boundary commits.

When blame.showroot is not set (or --root is not given), the
root commits are treated as boundary commits.  The code still
attributes the lines to them, but with -b their object names are
not shown.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-branch -d: do not stop at the first failure.
Quy Tonthat [Mon, 18 Dec 2006 22:42:16 +0000 (09:42 +1100)]
git-branch -d: do not stop at the first failure.

If there are more than one branches to be deleted, failure on
one will no longer stop git-branch to process the next ones.
The command still reports failures by exitting non-zero status.

Signed-off-by: Quy Tonthat <qtonthat@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoTeach git-branch to delete tracking branches with -r -d
Junio C Hamano [Mon, 18 Dec 2006 07:58:16 +0000 (23:58 -0800)]
Teach git-branch to delete tracking branches with -r -d

Because -r already means "remote" when listing, you can say:

$ git branch -d -r origin/todo origin/html origin/man

I just twisted it not to check fast-forwardness with the current
branch when you are removing a tracking branch.  Most likely,
removal of a tracking branch is not because you are "done with"
it (for a local branch, it usually means "you merged it up"),
but because you are not even interested in it.  In other words,
remote tracking branches are more like tags than branches.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agofix testsuite: make sure they use templates freshly built from the source
Junio C Hamano [Mon, 18 Dec 2006 23:27:43 +0000 (15:27 -0800)]
fix testsuite: make sure they use templates freshly built from the source

The initial t/trash repository for testing was created properly
but over time we gained many tests that create secondary test
repositories with init-db or clone and they were not careful
enough.

This fixes it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agorerere: fix breakage of resolving.
Junio C Hamano [Tue, 19 Dec 2006 00:20:53 +0000 (16:20 -0800)]
rerere: fix breakage of resolving.

commit e2b70087 botched the RCS merge to git-merge-file conversion.
There is no command called "git merge-file" (yes, we are using safer
variant of Perl's system(3)).

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAdd config example with respect to branch
Aneesh Kumar K.V [Thu, 7 Dec 2006 06:36:55 +0000 (12:06 +0530)]
Add config example with respect to branch

Update config.txt with example with respect to branch
config variable. This give a better idea regarding
how branch names are expected.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAdd documentation for show-branch --topics
Brian Gernhardt [Mon, 18 Dec 2006 15:48:33 +0000 (10:48 -0500)]
Add documentation for show-branch --topics

Add a quick paragraph explaining the --topics option for show-branch.
The explanation is an abbreviated version of the commit message from
d320a5437f8304cf9ea3ee1898e49d643e005738.

Signed-off-by: Brian Gernhardt <benji@silverinsanity.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agomake git a bit less cryptic on fetch errors
Nicolas Pitre [Mon, 18 Dec 2006 20:16:58 +0000 (15:16 -0500)]
make git a bit less cryptic on fetch errors

The remote server might not want to tell why it doesn't like us for
security reasons, but let's make the client report such error in a bit
less confusing way.  The remote failure remains a mystery, but the local
message might be a bit less so.

[jc: with a gentle wording updates from Andy Parkins]

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agomake patch_delta() error cases a bit more verbose
Nicolas Pitre [Mon, 18 Dec 2006 21:06:50 +0000 (16:06 -0500)]
make patch_delta() error cases a bit more verbose

It is especially important to distinguish between a malloc() failure
from all the other cases.  An out of memory condition is much less
worrisome than a compatibility/corruption problem.

Also make test-delta compilable again.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoracy-git: documentation updates.
Junio C Hamano [Mon, 18 Dec 2006 22:18:54 +0000 (14:18 -0800)]
racy-git: documentation updates.

We've removed the workaround for runtime penalty that did not
exist in practice some time ago, but the technical paper that
proposed that change still said "we probably should do so".

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoshow-ref: fix --exclude-existing
Junio C Hamano [Mon, 18 Dec 2006 21:33:47 +0000 (13:33 -0800)]
show-ref: fix --exclude-existing

Do not falsely document --filter-invalid which does not even exist.
Also make sure the line is long enough to have ^{} suffix before
checking for it.

Pointed out by Dscho.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoparse-remote::expand_refs_wildcard()
Junio C Hamano [Mon, 18 Dec 2006 20:13:46 +0000 (12:13 -0800)]
parse-remote::expand_refs_wildcard()

Work around dash incompatibility by not using "${name%'^{}'}".

Noticed by Jeff King; dash seems to mistake the closing brace
inside the single quote as the terminating brace for parameter
expansion.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agovim syntax: follow recent changes to commit template
Jeff King [Mon, 18 Dec 2006 07:26:42 +0000 (02:26 -0500)]
vim syntax: follow recent changes to commit template

This patch changes the syntax highlighting to correctly match the new
text of the commit message introduced by
  82dca84871637ac9812c0dec27f56d07cfba524c

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoshow-ref: fix --verify --hash=length
Junio C Hamano [Mon, 18 Dec 2006 03:27:49 +0000 (19:27 -0800)]
show-ref: fix --verify --hash=length

An earlier optimization for --verify broke a lot of stuff
because it did not take interaction with other flags into
account.

This also fixes an unrelated argument parsing error; --hash=8
should mean the same as "--hash --abbrev=8".

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoshow-ref: fix --quiet --verify
Junio C Hamano [Mon, 18 Dec 2006 02:53:24 +0000 (18:53 -0800)]
show-ref: fix --quiet --verify

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'jc/blame-boundary'
Junio C Hamano [Mon, 18 Dec 2006 02:34:51 +0000 (18:34 -0800)]
Merge branch 'jc/blame-boundary'

* jc/blame-boundary:
  git-blame: show lines attributed to boundary commits differently.

17 years agoMerge branch 'jc/reflog' (early part)
Junio C Hamano [Mon, 18 Dec 2006 02:31:24 +0000 (18:31 -0800)]
Merge branch 'jc/reflog' (early part)

* 'jc/reflog' (early part):
  Teach show-branch how to show ref-log data.

17 years agoMerge branch 'js/branch-config'
Junio C Hamano [Mon, 18 Dec 2006 02:27:17 +0000 (18:27 -0800)]
Merge branch 'js/branch-config'

* js/branch-config:
  git-branch: rename config vars branch.<branch>.*, too
  add a function to rename sections in the config

17 years agoMerge branch 'jn/web' (early part)
Junio C Hamano [Mon, 18 Dec 2006 02:27:17 +0000 (18:27 -0800)]
Merge branch 'jn/web' (early part)

* 'jn/web' (early part):
  gitweb: Add "next" link to commit view
  gitweb: Add title attribute to ref marker with full ref name
  gitweb: Do not show difftree for merges in "commit" view
  gitweb: SHA-1 in commit log message links to "object" view
  gitweb: Hyperlink target of symbolic link in "tree" view (if possible)
  gitweb: Add generic git_object subroutine to display object of any type
  gitweb: Show target of symbolic link in "tree" view
  gitweb: Don't use Content-Encoding: header in git_snapshot

17 years agoavoid accessing _all_ loose refs in git-show-ref --verify
Junio C Hamano [Mon, 18 Dec 2006 02:08:52 +0000 (18:08 -0800)]
avoid accessing _all_ loose refs in git-show-ref --verify

If you want to verify a ref, it is overkill to first read all loose refs
into a linked list, and then check if the desired ref is there.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
17 years agogit-fetch: Avoid reading packed refs over and over again
Junio C Hamano [Mon, 18 Dec 2006 01:57:19 +0000 (17:57 -0800)]
git-fetch: Avoid reading packed refs over and over again

When checking which tags to fetch, the old code used to call
git-show-ref --verify for each remote tag. Since reading even
packed refs is not a cheap operation when there are a lot of
local refs, the code became quite slow.

This fixes it by teaching git-show-ref to filter out existing
refs using a new mode of operation of git-show-ref.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoTeach show-branch how to show ref-log data.
Junio C Hamano [Thu, 14 Dec 2006 23:58:56 +0000 (15:58 -0800)]
Teach show-branch how to show ref-log data.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMake left-right automatic.
Junio C Hamano [Sun, 17 Dec 2006 00:07:20 +0000 (16:07 -0800)]
Make left-right automatic.

When using symmetric differences, I think the user almost always
would want to know which side of the symmetry each commit came
from.  So this removes --left-right option from the command
line, and turns it on automatically when a symmetric difference
is used ("git log --merge" counts as a symmetric difference
between HEAD and MERGE_HEAD).

Just in case, a new option --no-left-right is provided to defeat
this, but I do not know if it would be useful.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoTeach all of log family --left-right output.
Junio C Hamano [Sat, 16 Dec 2006 23:31:25 +0000 (15:31 -0800)]
Teach all of log family --left-right output.

This makes reviewing

     git log --left-right --merge --no-merges -p

a lot more pleasant.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agorev-list --left-right
Junio C Hamano [Mon, 23 Oct 2006 00:32:47 +0000 (17:32 -0700)]
rev-list --left-right

The output from "symmetric diff", i.e. A...B, does not
distinguish between commits that are reachable from A and the
ones that are reachable from B.  In this picture, such a
symmetric diff includes commits marked with a and b.

         x---b---b  branch B
        / \ /
       /   .
      /   / \
     o---x---a---a  branch A

However, you cannot tell which ones are 'a' and which ones are
'b' from the output.  Sometimes this is frustrating.  This adds
an output option, --left-right, to rev-list.

        rev-list --left-right A...B

would show ones reachable from A prefixed with '<' and the ones
reachable from B prefixed with '>'.

When combined with --boundary, boundary commits (the ones marked
with 'x' in the above picture) are shown with prefix '-', so you
would see list that looks like this:

    git rev-list --left-right --boundary --pretty=oneline A...B

    >bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb 3rd on b
    >bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb 2nd on b
    <aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa 3rd on a
    <aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa 2nd on a
    -xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 1st on b
    -xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 1st on a

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agomarkup fix in svnimport documentation.
Junio C Hamano [Sun, 17 Dec 2006 18:33:24 +0000 (10:33 -0800)]
markup fix in svnimport documentation.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoDocumentation: new option -P for git-svnimport
Quy Tonthat [Sun, 17 Dec 2006 14:50:12 +0000 (01:50 +1100)]
Documentation: new option -P for git-svnimport

Documentation: new option -P for git-svnimport.

Signed-off-by: Quy Tonthat <qtonthat@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoFix mis-mark-up in git-merge-file.txt documentation
Junio C Hamano [Sun, 17 Dec 2006 08:46:35 +0000 (00:46 -0800)]
Fix mis-mark-up in git-merge-file.txt documentation

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoDefault GIT_COMMITTER_NAME to login name in recieve-pack.
Shawn O. Pearce [Sun, 17 Dec 2006 08:15:14 +0000 (03:15 -0500)]
Default GIT_COMMITTER_NAME to login name in recieve-pack.

If GIT_COMMITTER_NAME is not available in receive-pack but reflogs
are enabled we would normally die out with an error message asking
the user to correct their environment settings.

Now that reflogs are enabled by default in (what we guessed to be)
non-bare Git repositories this may cause problems for some users
who don't have their full name in the gecos field and who don't
have access to the remote system to correct the problem.

So rather than die()'ing out in receive-pack when we try to log a
ref change and have no committer name we default to the username,
as obtained from the host's password database.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoFix check_file_directory_conflict().
Junio C Hamano [Sun, 17 Dec 2006 01:39:06 +0000 (17:39 -0800)]
Fix check_file_directory_conflict().

When replacing an existing file A with a directory A that has a
file A/B in it in the index, 'update-index --replace --add A/B'
did not properly remove the file to make room for the new
directory.

There was a trivial logic error, most likely a cut & paste one,
dating back to quite early days of git.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-add: remove conflicting entry when adding.
Junio C Hamano [Sun, 17 Dec 2006 09:09:41 +0000 (01:09 -0800)]
git-add: remove conflicting entry when adding.

When replacing an existing file A with a directory A that has a
file A/B in it in the index, 'git add' did not succeed because
it forgot to pass the allow-replace flag to add_cache_entry().

It might be safer to leave this as an error and require the user
to explicitly remove the existing A first before adding A/B
since it is an unusual case, but doing that automatically is
much easier to use.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoupdate-index: make D/F conflict error a bit more verbose.
Junio C Hamano [Sun, 17 Dec 2006 00:23:02 +0000 (16:23 -0800)]
update-index: make D/F conflict error a bit more verbose.

When you remove a directory D that has a tracked file D/F out of the
way to create a file D and try to "git update-index --add D", it used
to say "cannot add" which was not very helpful.  This issues an extra
error message to explain the situation before the final "fatal" message.

Since D/F conflicts are relatively rare event, extra verbosity would
not make things too noisy.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-branch: rename config vars branch.<branch>.*, too
Johannes Schindelin [Sat, 16 Dec 2006 14:15:02 +0000 (15:15 +0100)]
git-branch: rename config vars branch.<branch>.*, too

When renaming a branch, the corresponding config section should
be renamed, too.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoadd a function to rename sections in the config
Johannes Schindelin [Sat, 16 Dec 2006 14:14:14 +0000 (15:14 +0100)]
add a function to rename sections in the config

Given a config like this:

# A config
[very.interesting.section]
not

The command

$ git repo-config --rename-section very.interesting.section bla.1

will lead to this config:

# A config
[bla "1"]
not

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-clone: lose the traditional 'no-separate-remote' layout
Junio C Hamano [Sat, 16 Dec 2006 09:53:10 +0000 (01:53 -0800)]
git-clone: lose the traditional 'no-separate-remote' layout

Finally.

The separate-remote layout is so much more organized than
traditional and easier to work with especially when you need to
deal with remote repositories with multiple branches and/or you
need to deal with more than one remote repositories, and using
traditional layout for new repositories simply does not make
much sense.

Internally we still have code for 1:1 mappings to create a bare
clone; that is a good thing and will not go away.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-clone: lose the artificial "first" fetch refspec
Junio C Hamano [Sat, 16 Dec 2006 09:41:51 +0000 (01:41 -0800)]
git-clone: lose the artificial "first" fetch refspec

Now we lost the "first refspec is the one that is merged by default"
rule, there is no reason for clone to list the remote primary branch
in the config file explicitly anymore.

We still need it for the traditional layout for other reasons,
though.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-pull: refuse default merge without branch.*.merge
Junio C Hamano [Sat, 16 Dec 2006 09:36:32 +0000 (01:36 -0800)]
git-pull: refuse default merge without branch.*.merge

Everybody hated the pull behaviour of merging the first branch
listed on remotes/* file (or remote.*.fetch config) into the
current branch.  This finally corrects that UI wart by
forbidding "git pull" without an explicit branch name on the
command line or branch.$current.merge for the current branch.

The matching change to git-clone was made to prepare the default
branch.*.merge entry for the primary branch some time ago.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-clone: use wildcard specification for tracking branches
Junio C Hamano [Sat, 16 Dec 2006 09:14:39 +0000 (01:14 -0800)]
git-clone: use wildcard specification for tracking branches

This stops enumerating the set of branches found on the remote
side when a clone was made in the configuration file.  Instead,
a single entry that maps each remote branch to the local
tracking branch for the remote under the same name is created.

Doing it this way not only shortens the configuration file, but
automatically adjusts to a new branch added on the remote side
after the clone is made.

Unfortunately this cannot be done for the traditional layout,
where we always need to special case the 'master' to 'origin'
mapping within the local branch namespace.  But that is Ok; it
will be going away before v1.5.0.

We could also lose the "primary branch" mapping at the
beginning, but that has to wait until we implement the "forbid
'git pull' when we do not have branch.$current.merge for the
current branch" policy we earlier discussed.  That should also
be in v1.5.0

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agomerge: give a bit prettier merge message to "merge branch~$n"
Junio C Hamano [Sat, 16 Dec 2006 20:31:45 +0000 (12:31 -0800)]
merge: give a bit prettier merge message to "merge branch~$n"

This hacks the input to fmt-merge-msg to make the message for
merging early part of a branch a little easier to read.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Add "next" link to commit view
Jakub Narebski [Fri, 15 Dec 2006 20:57:16 +0000 (21:57 +0100)]
gitweb: Add "next" link to commit view

Add a kind of "next" view in the bottom part of navigation bar for
"commit" view, similar to what was added for "commitdiff" view in
commit 151602df00b8e5c5b4a8193f59a94b85f9b5aebc
  'gitweb: Add "next" link to commitdiff view'

For "commit" view for single parent commit:
  (parent: _commit_)
For "commit" view for merge (multi-parent) commit:
  (merge: _commit_ _commit_ ...)
For "commit" view for root (parentless) commit
  (initial)
where _link_ denotes hyperlink.  SHA1 of commit is shortened
to 7 characters on display.

While at it, remove leftovers from commit cae1862a by Petr Baudis:
  'gitweb: More per-view navigation bar links'
namely the "blame" link if there exist $file_name and commit has a
parent; it was added in git_commit probably by mistake.  The rest
of what mentioned commit added for git_commit was removed in
commit 6e0e92fda893311ff5af91836e5007bf6bbd4a21 by Luben Tuikov:
  'gitweb: Do not print "log" and "shortlog" redundantly in commit view'
(which should have probably removed also this "blame" link removed now).

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Add title attribute to ref marker with full ref name
Jakub Narebski [Fri, 15 Dec 2006 22:49:12 +0000 (23:49 +0100)]
gitweb: Add title attribute to ref marker with full ref name

Add title attribute, which will be shown as popup on mouseover in
graphical web browsers, with full name of ref, including part (type)
removed from the name of ref itself. This is useful to see that this
strange ref is StGIT ref, or it is remote branch, or it is lightweigh
tag (with branch-like name).

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Do not show difftree for merges in "commit" view
Jakub Narebski [Fri, 15 Dec 2006 16:53:45 +0000 (17:53 +0100)]
gitweb: Do not show difftree for merges in "commit" view

Do not show difftree against first parent for merges (commits with
more than one parent) in "commit" view, because it usually is
misleading.  git-show and git-whatchanged doesn't show diff for merges
either.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-svn: rename 'commit' command to 'set-tree'
Eric Wong [Sat, 16 Dec 2006 07:58:08 +0000 (23:58 -0800)]
git-svn: rename 'commit' command to 'set-tree'

'set-tree' probably accurately describes what the command
formerly known as 'commit' does.

I'm not entirely sure that 'dcommit' should be renamed to 'commit'
just yet...  Perhaps 'push' or 'push-changes'?

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-svn: remove support for the svn command-line client
Eric Wong [Sat, 16 Dec 2006 07:58:07 +0000 (23:58 -0800)]
git-svn: remove support for the svn command-line client

Using the command-line client was great for prototyping and
getting something working quickly.  Eventually I found time
to study the library documentation and add support for using
the libraries which are much faster and more flexible when
it comes to supporting new features.

Note that we require version 1.1 of the SVN libraries, whereas
we supported the command-line svn client down to version 1.0.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoDocument git-merge-file
Johannes Schindelin [Sat, 16 Dec 2006 11:22:18 +0000 (12:22 +0100)]
Document git-merge-file

Most of this is derived from the documentation of RCS merge.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-clone documentation
Quy Tonthat [Thu, 14 Dec 2006 14:17:53 +0000 (01:17 +1100)]
git-clone documentation

When --use-separate-remote is used on git-clone, the remote
heads are saved under $GIT_DIR/refs/remotes/origin/, not
"$GIT_DIR/remotes/origin/"

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-status always says what branch it's on
Andy Parkins [Thu, 14 Dec 2006 15:25:23 +0000 (15:25 +0000)]
git-status always says what branch it's on

If the current branch was "master" then git-status wouldn't say

 # On branch XXXX

In its output.  This patch makes it so that this message is always
output; regardless of branch name.

Signed-off-by: Andy Parkins <andyparkins@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-svn: convert to using Git.pm
Eric Wong [Fri, 15 Dec 2006 18:59:54 +0000 (10:59 -0800)]
git-svn: convert to using Git.pm

Thanks to Git.pm, I've been able to greatly reduce the amount
of extra work that needs to be done to manage input/output
pipes in Perl.

chomp usage has also been greatly reduced, too.

All tests (including full-svn-test) still pass, but this has
not been tested extensively in the real-world.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAlign section headers of 'git status' to new 'git add'.
Shawn O. Pearce [Sat, 16 Dec 2006 02:53:13 +0000 (21:53 -0500)]
Align section headers of 'git status' to new 'git add'.

Now that 'git add' is considered a first-class UI for 'update-index'
and that the 'git add' documentation states "Even modified files
must be added to the set of changes about to be committed" we should
make the output of 'git status' align with that documentation and
common usage.

So now we see a status output such as:

  # Added but not yet committed:
  #   (will commit)
  #
  #       new file: x
  #
  # Changed but not added:
  #   (use "git add file1 file2" to include for commit)
  #
  #       modified:   x
  #
  # Untracked files:
  #   (use "git add" on files to include for commit)
  #
  #       y

which just reads better in the context of using 'git add' to
manipulate a commit (and not a checkin, whatever the heck that is).

We also now support 'color.status.added' as an alias for the existing
'color.status.updated', as this alias more closely aligns with the
current output and documentation.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoSuggest use of "git add file1 file2" when there is nothing to commit.
Shawn O. Pearce [Sat, 16 Dec 2006 02:53:09 +0000 (21:53 -0500)]
Suggest use of "git add file1 file2" when there is nothing to commit.

If a user modifies files and runs 'git commit' (without the very
useful -a option) and they have not yet updated the index they
are probably coming from another SCM-like tool which would perform
the same as 'git commit -a' in this case.  Showing the user their
current status and a final line of "nothing to commit" is not very
reassuring, as the user might believe that Git did not recognize
their files were modified.

Instead we can suggest as part of the 'nothing to commit' message
that the user invoke 'git add' to add files to their next commit.

Suggested by Andy Parkins' Git 'niggles' list
(<200612132237.10051.andyparkins@gmail.com>).

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMake git-diff documentation use [--] when it should.
Brian Gernhardt [Fri, 15 Dec 2006 12:39:33 +0000 (07:39 -0500)]
Make git-diff documentation use [--] when it should.

Two of the cases has "[--] [<path>...]" and two had "-- [<path>...]".
Not terribly consistent and potentially confusing.  Also add "[--]" to
the synopsis so that it's obvious you can use it from the very
beginning.

Signed-off-by: Brian Gernhardt <benji@silverinsanity.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAdd --add option to git-repo-config
Brian Gernhardt [Fri, 15 Dec 2006 12:39:04 +0000 (07:39 -0500)]
Add --add option to git-repo-config

For multivars, the "git-repo-config name value ^$" is useful but
nonintuitive and troublesome to do repeatedly (since the value is not
at the end of the command line).  This commit simply adds an --add
option that adds a new value to a multivar.  Particularly useful for
tracking a new branch on a remote:

git-repo-config --add remote.origin.fetch +next:origin/next

Includes documentation and test.

Signed-off-by: Brian Gernhardt <benji@silverinsanity.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoTeach bash the new features of 'git show'.
Shawn O. Pearce [Fri, 15 Dec 2006 07:20:03 +0000 (02:20 -0500)]
Teach bash the new features of 'git show'.

Now that 'git show' accepts ref:path as an argument to specify a
tree or blob we should use the same completion logic as we support
for cat-file's object identifier.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoExport PERL_PATH
Luben Tuikov [Fri, 15 Dec 2006 07:03:03 +0000 (23:03 -0800)]
Export PERL_PATH

PERL_PATH is used by perl/Makefile so export it.

Signed-off-by: Luben Tuikov <ltuikov@yahoo.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoEnable reflogs by default in any repository with a working directory.
Shawn O. Pearce [Thu, 14 Dec 2006 22:41:17 +0000 (17:41 -0500)]
Enable reflogs by default in any repository with a working directory.

New and experienced Git users alike are finding out too late that
they forgot to enable reflogs in the current repository, and cannot
use the information stored within it to recover from an incorrectly
entered command such as `git reset --hard HEAD^^^` when they really
meant HEAD^^ (aka HEAD~2).

So enable reflogs by default in all future versions of Git, unless
the user specifically disables it with:

  [core]
    logAllRefUpdates = false

in their .git/config or ~/.gitconfig.

We only enable reflogs in repositories that have a working directory
associated with them, as shared/bare repositories do not have
an easy means to prune away old log entries, or may fail logging
entirely if the user's gecos information is not valid during a push.
This heuristic was suggested on the mailing list by Junio.

Documentation was also updated to indicate the new default behavior.
We probably should start to teach usuing the reflog to recover
from mistakes in some of the tutorial material, as new users are
likely to make a few along the way and will feel better knowing
they can recover from them quickly and easily, without fsck-objects'
lost+found features.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoProvide more meaningful output from 'git init-db'.
Shawn O. Pearce [Fri, 15 Dec 2006 05:44:58 +0000 (00:44 -0500)]
Provide more meaningful output from 'git init-db'.

Back in the old days of Git when people messed around with their
GIT_DIR environment variable more often it was nice to know whether
or not git-init-db created a .git directory or used GIT_DIR.
As most users at that time were rather technical UNIXy folk the
message "defaulting to local storage area" made sense to some and
seemed reasonable.

But it doesn't really convey any meaning to the new Git user,
as they don't know what a 'local storage area is' nor do they
know enough about Git to care.  It also really doesn't tell the
experienced Git user a whole lot about the command they just ran,
especially if they might be reinitializing an existing repository
(e.g. to update hooks).

So now we print out what we did ("Initialized empty" or
"Reinitialized existing"), what type of repository ("" or "shared"),
and what location the repository will be in ("$GIT_DIR").

Suggested in part by Andy Parkins in his Git 'niggles' list
(<200612132237.10051.andyparkins@gmail.com>).

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agomake commit message a little more consistent and conforting
Nicolas Pitre [Fri, 15 Dec 2006 04:15:44 +0000 (23:15 -0500)]
make commit message a little more consistent and conforting

It is nicer to let the user know when a commit succeeded all the time,
not only the first time.  Also the commit sha1 is much more useful than
the tree sha1 in this case.

This patch also introduces a -q switch to supress this message as well
as the summary of created/deleted files.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAvoid accessing a slow working copy during diffcore operations.
Shawn O. Pearce [Thu, 14 Dec 2006 11:15:57 +0000 (06:15 -0500)]
Avoid accessing a slow working copy during diffcore operations.

The Cygwin folks have done a fine job at creating a POSIX layer
on Windows That Just Works(tm).  However it comes with a penalty;
accessing files in the working tree by way of stat/open/mmap can
be slower for diffcore than inflating the data from a blob which
is stored in a packfile.

This performance problem is especially an issue in merge-recursive
when dealing with nearly 7000 added files, as we are loading
each file's content from the working directory to perform rename
detection.  I have literally seen (and sadly watched) paint dry in
less time than it takes for merge-recursive to finish such a merge.
On the other hand this very same merge runs very fast on Solaris.

If Git is compiled with NO_FAST_WORKING_DIRECTORY set then we will
avoid looking at the working directory when the blob in question
is available within a packfile and the caller doesn't need the data
unpacked into a temporary file.

We don't use loose objects as they have the same open/mmap/close
costs as the working directory file access, but have the additional
CPU overhead of needing to inflate the content before use.  So it
is still faster to use the working tree file over the loose object.

If the caller needs the file data unpacked into a temporary file
its likely because they are going to call an external diff program,
passing the file as a parameter.  In this case reusing the working
tree file will be faster as we don't need to inflate the data and
write it out to a temporary file.

The NO_FAST_WORKING_DIRECTORY feature is enabled by default on
Cygwin, as that is the platform which currently appears to benefit
the most from this option.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'js/show'
Junio C Hamano [Sat, 16 Dec 2006 05:38:26 +0000 (21:38 -0800)]
Merge branch 'js/show'

* js/show:
  git-show: grok blobs, trees and tags, too

17 years agogit-show: grok blobs, trees and tags, too
Johannes Schindelin [Thu, 14 Dec 2006 10:31:05 +0000 (11:31 +0100)]
git-show: grok blobs, trees and tags, too

Since git-show is pure Porcelain, it is the ideal candidate to
pretty print other things than commits, too.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-reset [--mixed] <tree> [--] <paths>...
Junio C Hamano [Thu, 14 Dec 2006 09:19:19 +0000 (01:19 -0800)]
git-reset [--mixed] <tree> [--] <paths>...

Sometimes it is asked on the list how to revert selected path in
the index from a tree, most often HEAD, without affecting the
files in the working tree.  A similar operation that also
affects the working tree files has been available in the form of
"git checkout <tree> -- <paths>...".

By definition --soft would never affect either the index nor the
working tree files, and --hard is the way to make the working
tree files as close to pristine, so this new option is available
only for the default --mixed case.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-reset: make it work from within a subdirectory.
Junio C Hamano [Thu, 14 Dec 2006 08:40:15 +0000 (00:40 -0800)]
git-reset: make it work from within a subdirectory.

If you typically sit in, say "src/", it's annoying to have to
change directory to do a reset.

This may need to be reworked when we add "git reset -- paths..."
to encapsulate the "ls-tree $tree | update-index --index-info"
pattern.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-fetch: make it work from within a subdirectory.
Junio C Hamano [Thu, 14 Dec 2006 08:36:23 +0000 (00:36 -0800)]
git-fetch: make it work from within a subdirectory.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoINSTALL: no need to have GNU diff installed
Johannes Schindelin [Thu, 14 Dec 2006 10:40:21 +0000 (11:40 +0100)]
INSTALL: no need to have GNU diff installed

Since a long time, we have inbuilt diff generation.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'maint'
Junio C Hamano [Thu, 14 Dec 2006 10:45:12 +0000 (02:45 -0800)]
Merge branch 'maint'

* maint:
  Bypass expensive content comparsion during rename detection.

17 years agoBypass expensive content comparsion during rename detection.
Shawn O. Pearce [Thu, 14 Dec 2006 10:07:46 +0000 (05:07 -0500)]
Bypass expensive content comparsion during rename detection.

When comparing file contents during the second loop through a rename
detection attempt we can skip the expensive byte-by-byte comparsion
if both source and destination files have valid SHA1 values.  This
improves performance by avoiding either an expensive open/mmap to
read the working tree copy, or an expensive inflate of a blob object.

Unfortunately we still have to at least initialize the sizes of the
source and destination files even if the SHA1 values don't match.
Failing to initialize the sizes causes a number of test cases to fail
and start reporting different copy/rename behavior than was expected.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoUpdate git-diff documentation
Junio C Hamano [Thu, 14 Dec 2006 08:03:18 +0000 (00:03 -0800)]
Update git-diff documentation

Porcelain documentation should talk in terms of end-user workflow, not
in terms of implementation details.  Do not suggest update-index, but
git-add instead.  Explain differences among 0-, 1- and 2-tree cases
not as differences of number of trees given to the command, but say
why user would want to give these number of trees to the command in
what situation.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'jc/diff--cached'
Junio C Hamano [Thu, 14 Dec 2006 07:55:42 +0000 (23:55 -0800)]
Merge branch 'jc/diff--cached'

* jc/diff--cached:
  Revert "git-diff: Introduce --index and deprecate --cached."

17 years agogit-svn: allow both diff.color and color.diff
Junio C Hamano [Wed, 13 Dec 2006 23:58:41 +0000 (15:58 -0800)]
git-svn: allow both diff.color and color.diff

The list concensus is to group color related configuration under
"color.*" so let's be consistent.

Inspired by Andy Parkins's patch to do the same for diff/log
family.  With fixes from Eric Wong.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agorepacked packs should be read-only
Nicolas Pitre [Wed, 13 Dec 2006 21:25:26 +0000 (16:25 -0500)]
repacked packs should be read-only

... just like the other pack creating tools do.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoconfig documentation: group color items together.
Junio C Hamano [Wed, 13 Dec 2006 20:11:03 +0000 (12:11 -0800)]
config documentation: group color items together.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-svn: correctly handle "(no author)" when using an authors file
Eric Wong [Tue, 12 Dec 2006 04:25:58 +0000 (20:25 -0800)]
git-svn: correctly handle "(no author)" when using an authors file

The low-level parts of the SVN library return NULL/undef for
author-less revisions, whereas "(no author)" is a (svn) client
convention.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-blame: show lines attributed to boundary commits differently.
Junio C Hamano [Sat, 2 Dec 2006 04:45:45 +0000 (20:45 -0800)]
git-blame: show lines attributed to boundary commits differently.

When blaming with revision ranges, often many lines are attributed
to different commits at the boundary, but they are not interesting
for the purpose of finding project history during that revision range.

This outputs the lines blamed on boundary commits differently. When
showing "human format" output, their SHA-1 are shown with '^' prefixed.
In "porcelain format", the commit will be shown with an extra attribute
line "boundary".

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