git.git
18 years agoMerge branch 'lt/push' into next
Junio C Hamano [Tue, 2 May 2006 06:41:57 +0000 (23:41 -0700)]
Merge branch 'lt/push' into next

* lt/push:
  builtin-push: make it official.

18 years agobuiltin-push: make it official.
Junio C Hamano [Tue, 2 May 2006 06:40:37 +0000 (23:40 -0700)]
builtin-push: make it official.

Remove the shell script version, and hardlink the git binary to it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/show-branch-dense' into next
Junio C Hamano [Tue, 2 May 2006 06:37:47 +0000 (23:37 -0700)]
Merge branch 'jc/show-branch-dense' into next

* jc/show-branch-dense:
  show-branch: omit uninteresting merges.

18 years agoshow-branch: omit uninteresting merges.
Junio C Hamano [Tue, 2 May 2006 00:12:26 +0000 (17:12 -0700)]
show-branch: omit uninteresting merges.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/count' into next
Junio C Hamano [Tue, 2 May 2006 06:18:09 +0000 (23:18 -0700)]
Merge branch 'jc/count' into next

* jc/count:
  builtin-count-objects: make it official.

18 years agoMerge branch 'jc/logs' into next
Junio C Hamano [Tue, 2 May 2006 06:16:35 +0000 (23:16 -0700)]
Merge branch 'jc/logs' into next

* jc/logs:
  builtin-log/whatchanged/show: make them official.

18 years agobuiltin-log/whatchanged/show: make them official.
Junio C Hamano [Tue, 2 May 2006 06:14:58 +0000 (23:14 -0700)]
builtin-log/whatchanged/show: make them official.

Remove the shell script version, and hardlink the git binary to
them.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/diff' into next
Junio C Hamano [Tue, 2 May 2006 06:09:15 +0000 (23:09 -0700)]
Merge branch 'jc/diff' into next

* jc/diff:
  builtin-diff: call it "git-diff", really.

18 years agobuiltin-diff: call it "git-diff", really.
Junio C Hamano [Tue, 2 May 2006 05:58:18 +0000 (22:58 -0700)]
builtin-diff: call it "git-diff", really.

Call it "git diff" not "git diffn", remove the shell script
version, and hardlink the git binary to it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agobuiltin-count-objects: make it official.
Junio C Hamano [Tue, 2 May 2006 06:05:39 +0000 (23:05 -0700)]
builtin-count-objects: make it official.

Remove the shell-script version, make the hardlink from the git
binary, and update the documentation to describe a new option.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/xsha1-2'
Junio C Hamano [Tue, 2 May 2006 05:55:40 +0000 (22:55 -0700)]
Merge branch 'jc/xsha1-2'

* jc/xsha1-2:
  Extended SHA1 -- "rev^@" syntax to mean "all parents"

18 years agoMerge branch 'jc/pack'
Junio C Hamano [Tue, 2 May 2006 05:54:22 +0000 (22:54 -0700)]
Merge branch 'jc/pack'

18 years agoMerge branch 'jc/pathcheck'
Junio C Hamano [Tue, 2 May 2006 05:51:27 +0000 (22:51 -0700)]
Merge branch 'jc/pathcheck'

* jc/pathcheck:
  revision parsing: make "rev -- paths" checks stronger.

18 years agoMerge branch 'nh/fetch-http'
Junio C Hamano [Tue, 2 May 2006 05:42:01 +0000 (22:42 -0700)]
Merge branch 'nh/fetch-http'

* nh/fetch-http:
  git-fetch: resolve remote symrefs for HTTP transport

18 years agoMerge branch 'se/rebase'
Junio C Hamano [Tue, 2 May 2006 05:39:57 +0000 (22:39 -0700)]
Merge branch 'se/rebase'

* se/rebase:
  Add --continue and --abort options to git-rebase.

18 years agoMerge branch 'jc/cache-tree' into next
Junio C Hamano [Tue, 2 May 2006 05:28:33 +0000 (22:28 -0700)]
Merge branch 'jc/cache-tree' into next

* jc/cache-tree:
  fsck-objects: mark objects reachable from cache-tree
  cache-tree: replace a sscanf() by two strtol() calls

18 years agofsck-objects: mark objects reachable from cache-tree
Junio C Hamano [Tue, 2 May 2006 05:15:54 +0000 (22:15 -0700)]
fsck-objects: mark objects reachable from cache-tree

When fsck-objects scanned cache-tree, it forgot to mark the
trees it found reachable and in use.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agocache-tree: replace a sscanf() by two strtol() calls
Johannes Schindelin [Tue, 2 May 2006 01:31:02 +0000 (03:31 +0200)]
cache-tree: replace a sscanf() by two strtol() calls

On one of my systems, sscanf() first calls strlen() on the buffer. But
this buffer is not terminated by NUL. So git crashed.

strtol() does not share that problem, as it stops reading after the
first non-digit.

[jc: original patch was wrong and did not read the cache-tree
 structure correctly; this has been fixed up and tested minimally
 with fsck-objects. ]

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/grep' into next
Junio C Hamano [Mon, 1 May 2006 23:27:07 +0000 (16:27 -0700)]
Merge branch 'jc/grep' into next

* jc/grep:
  builtin-grep: do not use setup_revisions()
  builtin-grep: support '-l' option.
  builtin-grep: wildcard pathspec fixes

18 years agobuiltin-grep: do not use setup_revisions()
Junio C Hamano [Mon, 1 May 2006 22:58:29 +0000 (15:58 -0700)]
builtin-grep: do not use setup_revisions()

Grep may want to grok multiple revisions, but it does not make
much sense to walk revisions while doing so.  This stops calling
the code to parse parameters for the revision walker.  The
parameter parsing for the optional "-e" option becomes a lot
simpler with it as well.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agobuiltin-grep: support '-l' option.
Junio C Hamano [Mon, 1 May 2006 19:39:21 +0000 (12:39 -0700)]
builtin-grep: support '-l' option.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agobuiltin-grep: wildcard pathspec fixes
Junio C Hamano [Mon, 1 May 2006 19:27:56 +0000 (12:27 -0700)]
builtin-grep: wildcard pathspec fixes

This tweaks the pathspec wildcard used in builtin-grep to match
that of ls-files.  With this:

git grep -e DEBUG -- '*/Kconfig*'

would work like the shell script version, and you could even do:

git grep -e DEBUG --cached -- '*/Kconfig*' ;# from index
git grep -e DEBUG v2.6.12 -- '*/Kconfig*' ;# from rev

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/fmt-patch' into next
Junio C Hamano [Mon, 1 May 2006 08:55:02 +0000 (01:55 -0700)]
Merge branch 'jc/fmt-patch' into next

* jc/fmt-patch:
  Use RFC2822 dates from "git fmt-patch".

18 years agoMerge branch 'jc/xsha1' into next
Junio C Hamano [Mon, 1 May 2006 08:54:53 +0000 (01:54 -0700)]
Merge branch 'jc/xsha1' into next

* jc/xsha1:
  get_sha1(): :path and :[0-3]:path to extract from index.

18 years agoMerge branch 'jc/xsha1-2' into next
Junio C Hamano [Mon, 1 May 2006 08:54:27 +0000 (01:54 -0700)]
Merge branch 'jc/xsha1-2' into next

* jc/xsha1-2:
  Extended SHA1 -- "rev^@" syntax to mean "all parents"

18 years agoMerge branch 'jc/grep' into next
Junio C Hamano [Mon, 1 May 2006 08:54:02 +0000 (01:54 -0700)]
Merge branch 'jc/grep' into next

* jc/grep:
  built-in "git grep"

18 years agoUse RFC2822 dates from "git fmt-patch".
Junio C Hamano [Mon, 1 May 2006 08:44:33 +0000 (01:44 -0700)]
Use RFC2822 dates from "git fmt-patch".

Still Work-in-progress git fmt-patch (should it be known as
format-patch-ng?) is matched with the fix made by Huw Davies
in 262a6ef76a1dde97ab50d79fa5cd6d3f9f125765 commit to use
RFC2822 date format.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agobuilt-in "git grep"
Junio C Hamano [Mon, 1 May 2006 06:28:15 +0000 (23:28 -0700)]
built-in "git grep"

This attempts to set up built-in "git grep" to further reduce
our dependence on the shell, while at the same time optionally
allowing to run grep against object database.  You could do
funky things like these:

git grep --cached -e pattern ;# grep from index
git grep -e pattern master ;# or in a rev
git grep -e pattern master next ;# or in multiple revs
git grep -e pattern pu^@ ;# even like this with an
;# extension from another topic ;-)
git grep -e pattern master..next ;# or even from rev ranges
git grep -e pattern master~20:Documentation
;# or an arbitrary tree
git grep -e pattern next:git-commit.sh
         ;# or an arbitrary blob

Right now, it does not understand and/or obey many options grep
should accept, and the pattern must be given with -e option due
to the way the parameter parser is structured, both of which
obviously need to be fixed for usability.

But this is going in the right direction.  The shell script
version is one of the worst Portability offender in the git
barebone Porcelainish; it uses xargs -0 to pass paths around and
shell arrays to sift flags and parameters.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoget_sha1(): :path and :[0-3]:path to extract from index.
Junio C Hamano [Sat, 22 Apr 2006 00:31:04 +0000 (17:31 -0700)]
get_sha1(): :path and :[0-3]:path to extract from index.

Earlier patch to say <ent>:<path> by Linus was very useful, and
this extends the same idea to the current index.  An sha1
expression :<path> extracts the object name for the named path
from the current index.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoExtended SHA1 -- "rev^@" syntax to mean "all parents"
Junio C Hamano [Sun, 30 Apr 2006 07:54:29 +0000 (00:54 -0700)]
Extended SHA1 -- "rev^@" syntax to mean "all parents"

A short-hand "rev^@" is understood to be "all parents of the
named commit" with this patch.  So you can do

git show v1.0.0^@

to view the parents of a merge commit,

gitk ^v1.0.0^@ v1.0.4

to view the log between two revs (including the bottom one), and

git diff --cc v1.1.0 v1.0.0^@

to inspect what got changed from the merge parents of v1.0.0 to v1.1.0.

This might be just my shiny new toy that is not very useful in
practice.  I needed it to do the multi-tree diff on Len's
infamous 12-way Octopus; typing "diff --cc funmerge funmerge^1
funmerge^2 funmerge^3 ..." was too painful.

[jc: taking suggestions from Linus and Johannes to match expectations
from shell users who are used to see $@ or $* either of which makes
sense.  I tend to write "$@" more often so...]

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'lt/push' into next
Junio C Hamano [Sun, 30 Apr 2006 23:20:18 +0000 (16:20 -0700)]
Merge branch 'lt/push' into next

* lt/push:
  Fix builtin-push to honor Push: lines in remotes file.
  builtin-push: resurrect parsing of Push: lines

18 years agoFix builtin-push to honor Push: lines in remotes file.
Junio C Hamano [Sun, 30 Apr 2006 23:03:27 +0000 (16:03 -0700)]
Fix builtin-push to honor Push: lines in remotes file.

[jc: originally from Johannes Schindelin, but reworked to lift a
 hard limit of Push: lines]

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agobuiltin-push: resurrect parsing of Push: lines
Johannes Schindelin [Sun, 30 Apr 2006 12:05:55 +0000 (14:05 +0200)]
builtin-push: resurrect parsing of Push: lines

The C'ification of push left these behind.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/diff' into next
Junio C Hamano [Sun, 30 Apr 2006 08:01:24 +0000 (01:01 -0700)]
Merge branch 'jc/diff' into next

* jc/diff:
  builtin-diff.c: die() formatting type fix.

18 years agoMerge branch 'lt/push' into next
Junio C Hamano [Sun, 30 Apr 2006 08:00:55 +0000 (01:00 -0700)]
Merge branch 'lt/push' into next

* lt/push:
  git builtin "push"
  git-format-patch: Use rfc2822 compliant date.

18 years agogit builtin "push"
Linus Torvalds [Sun, 30 Apr 2006 04:22:49 +0000 (21:22 -0700)]
git builtin "push"

This adds a builtin "push" command, which is largely just a C'ification of
the "git-push.sh" script.

Now, the reason I did it as a built-in is partly because it's yet another
step on relying less on shell, but it's actually mostly because I've
wanted to be able to push to _multiple_ repositories, and the most obvious
and simplest interface for that would seem be to just have a "remotes"
file that has multiple URL entries.

(For "pull", having multiple entries should either just select the first
one, or you could fall back on the others on failure - your choice).

And quite frankly, it just became too damn messy to do that in shell.
Besides, we actually have a fair amount of infrastructure in C, so it just
wasn't that hard to do.

Of course, this is almost totally untested. It probably doesn't work for
anything but the one trial I threw at it. "Simple" doesn't necessarily
mean "obviously correct".

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'fix'
Junio C Hamano [Sun, 30 Apr 2006 07:59:10 +0000 (00:59 -0700)]
Merge branch 'fix'

* fix:
  git-format-patch: Use rfc2822 compliant date.

18 years agogit-format-patch: Use rfc2822 compliant date.
Huw Davies [Sat, 29 Apr 2006 15:50:28 +0000 (15:50 +0000)]
git-format-patch: Use rfc2822 compliant date.

Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agobuiltin-diff.c: die() formatting type fix.
Junio C Hamano [Sun, 30 Apr 2006 07:26:41 +0000 (00:26 -0700)]
builtin-diff.c: die() formatting type fix.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/diff' into next
Junio C Hamano [Sat, 29 Apr 2006 08:33:16 +0000 (01:33 -0700)]
Merge branch 'jc/diff' into next

* jc/diff:
  built-in diff: assorted updates.
  built-in diff.

18 years agobuilt-in diff: assorted updates.
Junio C Hamano [Sat, 29 Apr 2006 08:24:49 +0000 (01:24 -0700)]
built-in diff: assorted updates.

"git diff(n)" without --base, --ours, etc. defaults to --cc,
which usually is the same as -p unless you are in the middle of
a conflicted merge, just like the shell script version.

"git diff(n) blobA blobB path" complains and dies.

"git diff(n) tree0 tree1 tree2...treeN" does combined diff that
shows a merge of tree1..treeN to result in tree0.

Giving "-c" option to any command that defaults to "--cc" turns
off dense-combined flag.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agobuilt-in diff.
Junio C Hamano [Sat, 29 Apr 2006 06:20:52 +0000 (23:20 -0700)]
built-in diff.

This starts to replace the shell script version of "git diff".

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'np/delta' into next
Junio C Hamano [Sat, 29 Apr 2006 05:42:41 +0000 (22:42 -0700)]
Merge branch 'np/delta' into next

* np/delta:
  replace adler32 with Rabin's polynomial in diff-delta

18 years agoreplace adler32 with Rabin's polynomial in diff-delta
Nicolas Pitre [Sat, 29 Apr 2006 04:58:05 +0000 (00:58 -0400)]
replace adler32 with Rabin's polynomial in diff-delta

This brings another small repacking speedup for sensibly the same pack
size.  On the Linux kernel repo, git-repack -a -f is 3.7% faster for a
0.4% larger pack.

Credits to Geert Bosch who brought the Rabin's polynomial idea to my
attention.

This also eliminate the issue of adler32() reading past the data buffer,
as noticed by Johannes Schindelin.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'master' into next
Junio C Hamano [Sat, 29 Apr 2006 02:30:41 +0000 (19:30 -0700)]
Merge branch 'master' into next

* master:
  Fix trivial typo in git-log man page.
  Properly render asciidoc "callouts" in git man pages.
  Fix up remaining man pages that use asciidoc "callouts".
  Update the git-branch man page to include the "-r" option,
  annotate: display usage information if no filename was given
  annotate: fix warning about uninitialized scalar
  git-am --resolved: more usable error message.

18 years agoMerge branch 'fix'
Junio C Hamano [Fri, 28 Apr 2006 23:57:32 +0000 (16:57 -0700)]
Merge branch 'fix'

* fix:
  Fix trivial typo in git-log man page.
  Properly render asciidoc "callouts" in git man pages.
  Fix up remaining man pages that use asciidoc "callouts".
  Update the git-branch man page to include the "-r" option,
  annotate: display usage information if no filename was given
  annotate: fix warning about uninitialized scalar
  git-am --resolved: more usable error message.

18 years agoFix trivial typo in git-log man page.
Sean Estabrooks [Fri, 28 Apr 2006 13:15:03 +0000 (09:15 -0400)]
Fix trivial typo in git-log man page.

Signed-off-by: Sean Estabrooks <seanlkml@sympatico.ca>
18 years agoProperly render asciidoc "callouts" in git man pages.
Sean Estabrooks [Fri, 28 Apr 2006 13:15:06 +0000 (09:15 -0400)]
Properly render asciidoc "callouts" in git man pages.

Adds an xsl fragment to render docbook callouts when
converting to man page format.  Update the Makefile
to have "xmlto" use it when generating man pages.

Signed-off-by: Sean Estabrooks <seanlkml@sympatico.ca>
18 years agoFix up remaining man pages that use asciidoc "callouts".
Sean Estabrooks [Fri, 28 Apr 2006 13:15:05 +0000 (09:15 -0400)]
Fix up remaining man pages that use asciidoc "callouts".

Unfortunately docbook does not allow a callout to be
referenced from inside a callout list description.
Rewrite one paragraph in git-reset man page to work
around this limitation.

Signed-off-by: Sean Estabrooks <seanlkml@sympatico.ca>
18 years agoUpdate the git-branch man page to include the "-r" option,
Sean Estabrooks [Fri, 28 Apr 2006 13:15:04 +0000 (09:15 -0400)]
Update the git-branch man page to include the "-r" option,

and fix up asciidoc "callouts"

Signed-off-by: Sean Estabrooks <seanlkml@sympatico.ca>
18 years agoannotate: display usage information if no filename was given
Matthias Kestenholz [Fri, 28 Apr 2006 08:41:19 +0000 (10:41 +0200)]
annotate: display usage information if no filename was given

Signed-off-by: Matthias Kestenholz <matthias@spinlock.ch>
18 years agoannotate: fix warning about uninitialized scalar
Matthias Kestenholz [Fri, 28 Apr 2006 08:42:28 +0000 (10:42 +0200)]
annotate: fix warning about uninitialized scalar

Use of uninitialized value in scalar chomp at
./git-annotate.perl line 212, <$kid> chunk 4.

Signed-off-by: Matthias Kestenholz <matthias@spinlock.ch>
18 years agogit-am --resolved: more usable error message.
Junio C Hamano [Fri, 28 Apr 2006 09:32:44 +0000 (02:32 -0700)]
git-am --resolved: more usable error message.

After doing the hard work of hand resolving the conflicts in the
working tree, if the user forgets to run update-index to mark
the paths that have been resolved, the command gave an
unfriendly "fatal: git-write-tree: not able to write tree" error
message.  Catch the situation early and give more meaningful
message and suggestion.

Noticed and suggested by Len Brown.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/cache-tree' into next
Junio C Hamano [Fri, 28 Apr 2006 08:36:52 +0000 (01:36 -0700)]
Merge branch 'jc/cache-tree' into next

* jc/cache-tree:
  cache-tree.c: typefix

18 years agocache-tree.c: typefix
Junio C Hamano [Fri, 28 Apr 2006 05:48:27 +0000 (22:48 -0700)]
cache-tree.c: typefix

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/count' into next
Junio C Hamano [Fri, 28 Apr 2006 04:38:43 +0000 (21:38 -0700)]
Merge branch 'jc/count' into next

* jc/count:
  built-in count-objects.
  pack-objects: update size heuristucs.
  verify-pack: check integrity in a saner order.

18 years agobuilt-in count-objects.
Junio C Hamano [Thu, 27 Apr 2006 22:37:18 +0000 (15:37 -0700)]
built-in count-objects.

Also it learned to do -v (verbose) to report:

- number of loose objects
- disk occupied by loose objects
- number of objects in local packs
- number of loose objects that are also in pack
- unrecognised garbage in .git/objects/??/.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agopack-objects: update size heuristucs.
Junio C Hamano [Fri, 28 Apr 2006 02:31:46 +0000 (19:31 -0700)]
pack-objects: update size heuristucs.

We used to omit delta base candidates that is much bigger than
the target, but delta size does not grow when we delete more, so
that was not a very good heuristics.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'np/delta' into next
Junio C Hamano [Fri, 28 Apr 2006 01:10:09 +0000 (18:10 -0700)]
Merge branch 'np/delta' into next

* np/delta:
  use delta index data when finding best delta matches

18 years agoMerge branch 'jc/cache-tree' into next
Junio C Hamano [Thu, 27 Apr 2006 23:25:32 +0000 (16:25 -0700)]
Merge branch 'jc/cache-tree' into next

* jc/cache-tree:
  test-dump-cache-tree: validate the cached data as well.
  cache_tree_update: give an option to update cache-tree only.

18 years agotest-dump-cache-tree: validate the cached data as well.
Junio C Hamano [Thu, 27 Apr 2006 23:22:45 +0000 (16:22 -0700)]
test-dump-cache-tree: validate the cached data as well.

While dumping the cached data, try recomputing everything from
scratch to make sure things match.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agocache_tree_update: give an option to update cache-tree only.
Junio C Hamano [Thu, 27 Apr 2006 23:21:54 +0000 (16:21 -0700)]
cache_tree_update: give an option to update cache-tree only.

When the extra "dryrun" parameter is true, cache_tree_update()
recomputes the invalid entry but does not actually creates
new tree object.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'fix'
Junio C Hamano [Thu, 27 Apr 2006 22:42:32 +0000 (15:42 -0700)]
Merge branch 'fix'

* fix:
  verify-pack: check integrity in a saner order.

18 years agoverify-pack: check integrity in a saner order.
Junio C Hamano [Thu, 27 Apr 2006 22:42:01 +0000 (15:42 -0700)]
verify-pack: check integrity in a saner order.

Check internal integrity to report corrupt pack or idx, and
then check cross-integrity between idx and pack.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/pathcheck' into next
Junio C Hamano [Thu, 27 Apr 2006 19:51:08 +0000 (12:51 -0700)]
Merge branch 'jc/pathcheck' into next

* jc/pathcheck:
  revision parsing: make "rev -- paths" checks stronger.

18 years agoFix mismerged update-index from jc/cache-tree branch.
Junio C Hamano [Thu, 27 Apr 2006 19:50:01 +0000 (12:50 -0700)]
Fix mismerged update-index from jc/cache-tree branch.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRetire rabinpoly fingerprinting code
Junio C Hamano [Thu, 27 Apr 2006 19:48:26 +0000 (12:48 -0700)]
Retire rabinpoly fingerprinting code

For now let's retire this and reintroduce it as part of the updated
pack-objects series from Geert when it is ready.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'master' into next
Junio C Hamano [Thu, 27 Apr 2006 19:25:18 +0000 (12:25 -0700)]
Merge branch 'master' into next

* master:
  Fix "git help -a" terminal autosizing
  diff-index: fix compilation warnings.

18 years agoMerge branch 'pb/config'
Junio C Hamano [Thu, 27 Apr 2006 19:23:01 +0000 (12:23 -0700)]
Merge branch 'pb/config'

* pb/config:
  git-cvsserver: typofixes
  Deprecate usage of git-var -l for getting config vars list
  git-repo-config --list support

18 years agoMerge branch 'jc/diffstat'
Junio C Hamano [Thu, 27 Apr 2006 18:58:33 +0000 (11:58 -0700)]
Merge branch 'jc/diffstat'

* jc/diffstat:
  diff --stat: show complete rewrites consistently.

18 years agoFix "git help -a" terminal autosizing
Linus Torvalds [Thu, 27 Apr 2006 16:02:54 +0000 (09:02 -0700)]
Fix "git help -a" terminal autosizing

When I split out the builtin commands into their own files, I left the
include of <sys/ioctl.h> in git.c rather than moving it to the file that
needed it (builtin-help.c).

Nobody seems to have noticed, because everything still worked, but because
the TIOCGWINSZ macro was now no longer defined when compiling the
"term_columns()" function, it would no longer automatically notice the
terminal size unless your system used the ancient "COLUMNS" environment
variable approach.

Trivially fixed by just moving the header include to the file that
actually needs it.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agodiff-index: fix compilation warnings.
Junio C Hamano [Thu, 27 Apr 2006 08:59:00 +0000 (01:59 -0700)]
diff-index: fix compilation warnings.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/cache-tree' into next
Junio C Hamano [Thu, 27 Apr 2006 08:37:03 +0000 (01:37 -0700)]
Merge branch 'jc/cache-tree' into next

* jc/cache-tree:
  read-tree: teach 1-way merege and plain read to prime cache-tree.
  read-tree: teach 1 and 2 way merges about cache-tree.
  update-index: when --unresolve, smudge the relevant cache-tree entries.

18 years agoread-tree: teach 1-way merege and plain read to prime cache-tree.
Junio C Hamano [Thu, 27 Apr 2006 08:33:07 +0000 (01:33 -0700)]
read-tree: teach 1-way merege and plain read to prime cache-tree.

This teaches read-tree to fully populate valid cache-tree when
reading a tree from scratch, or reading a single tree into an
existing index, reusing only the cached stat information (i.e.
one-way merge).  We have already taught update-index about cache-tree,
so "git checkout" followed by updates to a few path followed by
a "git commit" would become very efficient.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoread-tree: teach 1 and 2 way merges about cache-tree.
Junio C Hamano [Thu, 27 Apr 2006 07:13:34 +0000 (00:13 -0700)]
read-tree: teach 1 and 2 way merges about cache-tree.

This teaches one-way and two-way "read-tree -m" (and its special
form, "read-tree --reset" as well) not to discard cache-tree but
invalidate only the changed parts of the tree.  When switching
between related branches, this helps the eventual commit
(i.e. write-tree) by keeping cache-tree valid as much as
possible.

This does not prime cache-tree yet, but we ought to be able to
do that for no-merge (i.e. reading from a tree object) case and,
and also perhaps 1 way merge case.

With this patch applied, switching between the tip of Linux 2.6
kernel tree and a branch that touches one path (fs/ext3/Makefile)
from it invalidates only 3 paths out of 1201 cache-tree entries
in the index, and subsequent write-tree takes about a half as
much time as before.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoupdate-index: when --unresolve, smudge the relevant cache-tree entries.
Junio C Hamano [Thu, 27 Apr 2006 05:05:05 +0000 (22:05 -0700)]
update-index: when --unresolve, smudge the relevant cache-tree entries.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agouse delta index data when finding best delta matches
Nicolas Pitre [Thu, 27 Apr 2006 03:58:00 +0000 (23:58 -0400)]
use delta index data when finding best delta matches

This patch allows for computing the delta index for each base object
only once and reuse it when trying to find the best delta match.

This should set the mark and pave the way for possibly better delta
generator algorithms.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/cache-tree' into next
Junio C Hamano [Thu, 27 Apr 2006 01:48:53 +0000 (18:48 -0700)]
Merge branch 'jc/cache-tree' into next

* jc/cache-tree:
  t0000-basic: more commit-tree tests.

18 years agoMerge branch 'master' into jc/cache-tree
Junio C Hamano [Thu, 27 Apr 2006 01:32:45 +0000 (18:32 -0700)]
Merge branch 'master' into jc/cache-tree

* master:
  t0000-basic: more commit-tree tests.
  commit-tree.c: check_valid() microoptimization.
  Fix filename verification when in a subdirectory
  rebase: typofix.
  socksetup: don't return on set_reuse_addr() error

18 years agot0000-basic: more commit-tree tests.
Junio C Hamano [Thu, 27 Apr 2006 01:25:15 +0000 (18:25 -0700)]
t0000-basic: more commit-tree tests.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'se/rebase' into next
Junio C Hamano [Thu, 27 Apr 2006 00:24:15 +0000 (17:24 -0700)]
Merge branch 'se/rebase' into next

* se/rebase:
  Add --continue and --abort options to git-rebase.

18 years agoMerge branch 'nh/fetch-http' into next
Junio C Hamano [Thu, 27 Apr 2006 00:23:51 +0000 (17:23 -0700)]
Merge branch 'nh/fetch-http' into next

* nh/fetch-http:
  git-fetch: resolve remote symrefs for HTTP transport
  commit-tree.c: check_valid() microoptimization.
  Fix filename verification when in a subdirectory
  rebase: typofix.
  socksetup: don't return on set_reuse_addr() error

18 years agogit-fetch: resolve remote symrefs for HTTP transport
Nick Hengeveld [Wed, 26 Apr 2006 16:10:01 +0000 (09:10 -0700)]
git-fetch: resolve remote symrefs for HTTP transport

git-fetch validates that a remote ref resolves to a SHA1 prior to calling
git-http-fetch.  This adds support for resolving a few levels of symrefs
to get to the SHA1.

Signed-off-by: Nick Hengeveld <nickh@reactrix.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAdd --continue and --abort options to git-rebase.
sean [Wed, 26 Apr 2006 14:49:38 +0000 (10:49 -0400)]
Add --continue and --abort options to git-rebase.

  git rebase [--onto <newbase>] <upstream> [<branch>]
  git rebase --continue
  git rebase --abort

Add "--continue" to restart the rebase process after
manually resolving conflicts.  The user is warned if
there are still differences between the index and the
working files.

Add "--abort" to restore the original branch, and
remove the .dotest working files.

Some minor additions to the git-rebase documentation.

[jc: fix that applies to the maintenance track has been dealt
 with separately.]

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agorevision parsing: make "rev -- paths" checks stronger.
Junio C Hamano [Wed, 26 Apr 2006 22:09:27 +0000 (15:09 -0700)]
revision parsing: make "rev -- paths" checks stronger.

If you don't have a "--" marker, then:

 - all of the arguments we are going to assume are pathspecs
   must exist in the working tree.

 - none of the arguments we parsed as revisions could be
   interpreted as a filename.

so that there really isn't any possibility of confusion in case
somebody does have a revision that looks like a pathname too.

The former rule has been in effect; this implements the latter.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'fix'
Junio C Hamano [Thu, 27 Apr 2006 00:08:00 +0000 (17:08 -0700)]
Merge branch 'fix'

* fix:
  commit-tree.c: check_valid() microoptimization.
  Fix filename verification when in a subdirectory
  rebase: typofix.
  socksetup: don't return on set_reuse_addr() error

18 years agocommit-tree.c: check_valid() microoptimization.
Junio C Hamano [Wed, 26 Apr 2006 23:55:25 +0000 (16:55 -0700)]
commit-tree.c: check_valid() microoptimization.

There is no point reading the whole object just to make sure it exists and
it is of the expected type.  We added sha1_object_info() for such need
after this code was written, so use it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix filename verification when in a subdirectory
Linus Torvalds [Wed, 26 Apr 2006 17:15:54 +0000 (10:15 -0700)]
Fix filename verification when in a subdirectory

When we are in a subdirectory of a git archive, we need to take the prefix
of that subdirectory into accoung when we verify filename arguments.

Noted by Matthias Lederhofer

This also uses the improved error reporting for all the other git commands
that use the revision parsing interfaces, not just git-rev-parse. Also, it
makes the error reporting for mixed filenames and argument flags clearer
(you cannot put flags after the start of the pathname list).

[jc: with fix to a trivial typo noticed by Timo Hirvonen]

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agorebase: typofix.
Junio C Hamano [Wed, 26 Apr 2006 19:07:42 +0000 (12:07 -0700)]
rebase: typofix.

Noticed by Sean.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/cache-tree' into next
Junio C Hamano [Wed, 26 Apr 2006 10:39:01 +0000 (03:39 -0700)]
Merge branch 'jc/cache-tree' into next

* jc/cache-tree:
  commit-tree: allow generic object name for the tree as well.
  Makefile: remove and create xdiff library from scratch.
  t0000-basic: Add ls-tree recursive test back.

18 years agoMerge branch 'master' into jc/cache-tree
Junio C Hamano [Wed, 26 Apr 2006 10:29:09 +0000 (03:29 -0700)]
Merge branch 'master' into jc/cache-tree

* master:
  commit-tree: allow generic object name for the tree as well.
  Makefile: remove and create xdiff library from scratch.
  t0000-basic: Add ls-tree recursive test back.
  Libified diff-index: backward compatibility fix.
  Libify diff-index.
  Libify diff-files.
  Makefile: remove and create libgit.a from scratch.
  Document the configuration file
  Document git-var -l listing also configuration variables
  rev-parse: better error message for ambiguous arguments
  make update-index --chmod work with multiple files and --stdin
  socksetup: don't return on set_reuse_addr() error
  Fix "git show --stat"
  git-update-index --unresolve
  Add git-unresolve <paths>...
  Add colordiff for git to contrib/colordiff.
  gitk: Let git-rev-list do the argument list parsing

18 years agocommit-tree: allow generic object name for the tree as well.
Junio C Hamano [Wed, 26 Apr 2006 10:18:51 +0000 (03:18 -0700)]
commit-tree: allow generic object name for the tree as well.

We use get_sha1() for -p (parent) objects, but still used
get_sha1_hex() for the tree.  Just to be consistent, allow
extended SHA1 expression for the tree object name.

Note that this is not to encourage funky things like this:

git-commit-tree HEAD^{tree} -p HEAD

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMakefile: remove and create xdiff library from scratch.
Junio C Hamano [Wed, 26 Apr 2006 10:12:58 +0000 (03:12 -0700)]
Makefile: remove and create xdiff library from scratch.

... in the same spirit as 71459c193d04870076efa0a387c317390b53e3e2.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agot0000-basic: Add ls-tree recursive test back.
Junio C Hamano [Wed, 26 Apr 2006 09:27:59 +0000 (02:27 -0700)]
t0000-basic: Add ls-tree recursive test back.

When we updated ls-tree recursive output to omit the tree nodes,
246cc52f388cae8ca99e5a12b8458c9bfa467765 adjusted the old test
so that we do not expect to see trees in its output.  Later,
with 0f8f45cb4a7e664b396f73c25891da46b953b8b8, we added back the
ability to show both with -t option, but we forgot to update the
test as well.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'pb/config' into next
Junio C Hamano [Wed, 26 Apr 2006 07:00:49 +0000 (00:00 -0700)]
Merge branch 'pb/config' into next

* pb/config:
  git-cvsserver: typofixes

18 years agogit-cvsserver: typofixes
Junio C Hamano [Wed, 26 Apr 2006 06:59:28 +0000 (23:59 -0700)]
git-cvsserver: typofixes

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/diffstat' into next
Junio C Hamano [Wed, 26 Apr 2006 06:50:48 +0000 (23:50 -0700)]
Merge branch 'jc/diffstat' into next

* jc/diffstat:
  diff --stat: show complete rewrites consistently.
  Makefile: remove and create libgit.a from scratch.

18 years agodiff --stat: show complete rewrites consistently.
Junio C Hamano [Wed, 26 Apr 2006 06:40:09 +0000 (23:40 -0700)]
diff --stat: show complete rewrites consistently.

The patch format shows complete rewrite as deletion of all old lines
followed by addition of all new lines.  Count lines consistenly with
that when doing diffstat.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/diff'
Junio C Hamano [Wed, 26 Apr 2006 06:12:21 +0000 (23:12 -0700)]
Merge branch 'jc/diff'

* jc/diff:
  Libified diff-index: backward compatibility fix.
  Libify diff-index.
  Libify diff-files.

18 years agoMakefile: remove and create libgit.a from scratch.
Junio C Hamano [Wed, 26 Apr 2006 06:11:17 +0000 (23:11 -0700)]
Makefile: remove and create libgit.a from scratch.

Foolishly I renamed diff.o around which caused an old diff.o
taken out of libgit.a and got linked into resulting binary and
exhibited mysterious breakage for many people.  This borrows
from the kernel Makefile (scripts/Makefile.build) to first remove
the target and then recreate.

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