git.git
17 years ago"color.diff = true" is not "always" anymore.
Junio C Hamano [Mon, 26 Nov 2007 22:30:28 +0000 (14:30 -0800)]
"color.diff = true" is not "always" anymore.

Too many people got burned by setting color.diff and color.status to
true when they really should have set it to "auto".

This makes only "always" to do the unconditional colorization, and
change the meaning of "true" to the same as "auto": colorize only when
we are talking to a terminal.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMerge branch 'maint'
Junio C Hamano [Thu, 29 Nov 2007 01:06:57 +0000 (17:06 -0800)]
Merge branch 'maint'

* maint:
  scripts: do not get confused with HEAD in work tree
  Improve description of git-branch -d and -D in man page.

17 years agoscripts: do not get confused with HEAD in work tree
Junio C Hamano [Thu, 29 Nov 2007 00:15:04 +0000 (16:15 -0800)]
scripts: do not get confused with HEAD in work tree

When you have a file called HEAD in your work tree, many commands that
our scripts feed "HEAD" to would complain about the rev vs path
ambiguity.  A solution is to form command line more carefully by
appending -- to them, which makes it clear that we mean HEAD rev not
HEAD file.

This patch would apply to maint.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agobash completion: add diff options
Johannes Schindelin [Fri, 23 Nov 2007 01:11:35 +0000 (01:11 +0000)]
bash completion: add diff options

I use "git diff" (the porcelain) really often, and am almost as often
annoyed that the completions do not know how to complete something simple
as --cached.  Now they do.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogit checkout's reflog: even when detaching the HEAD, say from where
Johannes Schindelin [Fri, 23 Nov 2007 00:20:35 +0000 (00:20 +0000)]
git checkout's reflog: even when detaching the HEAD, say from where

When checking out another ref, the reflogs already record from which
branch you switched.  Do that also when switching to a detached HEAD.

While at it, record also when coming _from_ a detached HEAD.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agosha1_file.c: Fix size_t related printf format warnings
Steffen Prohaska [Wed, 21 Nov 2007 20:27:19 +0000 (21:27 +0100)]
sha1_file.c: Fix size_t related printf format warnings

The old way of fixing warnings did not succeed on MinGW.  MinGW
does not support C99 printf format strings for size_t [1].  But
gcc on MinGW issues warnings if C99 printf format is not used.
Hence, the old stragegy to avoid warnings fails.

[1] http://www.mingw.org/MinGWiki/index.php/C99

This commits passes arguments of type size_t through a tiny
helper functions that casts to the type expected by the format
string.

Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoRevert "t5516: test update of local refs on push"
Jeff King [Wed, 21 Nov 2007 07:19:34 +0000 (02:19 -0500)]
Revert "t5516: test update of local refs on push"

This reverts commit 09fba7a59d38d1cafaf33eadaf1d409c4113b30c.

These tests are superseded by the ones in t5404 (added in
6fa92bf3 and 8736a848), which are more extensive and better
organized.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoImprove description of git-branch -d and -D in man page.
Jan Hudec [Sat, 17 Nov 2007 19:51:44 +0000 (20:51 +0100)]
Improve description of git-branch -d and -D in man page.

Some users expect that deleting a remote-tracking branch would prevent
fetch from creating it again, so be explcit about that it's not the case.
Also be a little more explicit about what fully merged means.

Signed-off-by: Jan Hudec <bulb@ucw.cz>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agofilter-branch: fix dirty way to provide the helpers to commit filters
Johannes Schindelin [Wed, 28 Nov 2007 15:50:38 +0000 (15:50 +0000)]
filter-branch: fix dirty way to provide the helpers to commit filters

The helper functions 'map' and 'skip_commit' were provided to commit
filters by sourcing filter-branch itself.  This was done with a certain
environment variable set to indicate that only the functions should be
defined, and the script should return then.

This was really hacky, and it did not work all that well, since the
full path to git-filter-branch was not known at all times.

Avoid that by putting the functions into a variable, and eval'ing
that variable.  The commit filter gets these functions by prepending
the variable to the specified commands.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogit-stash: do not get fooled with "color.diff = true"
Pascal Obry [Mon, 26 Nov 2007 22:04:28 +0000 (23:04 +0100)]
git-stash: do not get fooled with "color.diff = true"

When colors are set to "true" on the repository, the git log output
will contain control characters to set/reset the colors, even when
the output is to a pipe. This makes list_stash() fail as the
downstream sed does not see what it is expecting.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoUse is_absolute_path() in diff-lib.c, lockfile.c, setup.c, trace.c
Steffen Prohaska [Sun, 25 Nov 2007 22:29:03 +0000 (23:29 +0100)]
Use is_absolute_path() in diff-lib.c, lockfile.c, setup.c, trace.c

Using the helper function to test for absolute paths makes porting easier.

Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoFix typo in draft 1.5.4 release notes
Wincent Colaiuta [Mon, 26 Nov 2007 08:34:48 +0000 (09:34 +0100)]
Fix typo in draft 1.5.4 release notes

Signed-off-by: Wincent Colaiuta <win@wincent.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMerge branch 'maint'
Junio C Hamano [Mon, 26 Nov 2007 03:10:01 +0000 (19:10 -0800)]
Merge branch 'maint'

* maint:
  user-manual: recovering from corruption
  user-manual: clarify language about "modifying" old commits
  user-manual: failed push to public repository
  user-manual: define "branch" and "working tree" at start
  git-checkout: describe detached head correctly

17 years agoMerge branch 'maint' of git://linux-nfs.org/~bfields/git into maint
Junio C Hamano [Mon, 26 Nov 2007 03:04:27 +0000 (19:04 -0800)]
Merge branch 'maint' of git://linux-nfs.org/~bfields/git into maint

* 'maint' of git://linux-nfs.org/~bfields/git:
  user-manual: recovering from corruption
  user-manual: clarify language about "modifying" old commits
  user-manual: failed push to public repository
  user-manual: define "branch" and "working tree" at start

17 years agouser-manual: recovering from corruption
J. Bruce Fields [Sun, 4 Mar 2007 03:53:37 +0000 (22:53 -0500)]
user-manual: recovering from corruption

Some instructions on dealing with corruption of the object database.

Most of this text is from an example by Linus, identified by Nicolas
Pitre <nico@cam.org> with a little further editing by me.

Signed-off-by: "J. Bruce Fields" <bfields@citi.umich.edu>
17 years agoPrint the real filename that we failed to open.
André Goddard Rosa [Wed, 21 Nov 2007 20:59:14 +0000 (18:59 -0200)]
Print the real filename that we failed to open.

When we fail to open a temporary file to be renamed to something else,
we reported the final filename, not the temporary file we failed to
open.

Signed-off-by: André Goddard Rosa <andre.goddard@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agouser-manual: clarify language about "modifying" old commits
J. Bruce Fields [Mon, 26 Nov 2007 00:01:57 +0000 (19:01 -0500)]
user-manual: clarify language about "modifying" old commits

It's important to remember that git doesn't really allowing "editing" or
"modifying" commits, only replacing them by new commits.  Redo some of
the language to make this clearer.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
17 years agorevert/cherry-pick: do not mention the original ref
Junio C Hamano [Sun, 25 Nov 2007 23:15:48 +0000 (15:15 -0800)]
revert/cherry-pick: do not mention the original ref

When you cherry-pick or revert a commit, naming it with an annotated
tag, we added a comment, attempting to repeat what we got from the end
user, to the message.

But this was inconsistent.  When we got "cherry-pick branch", we
recorded the object name (40-letter SHA-1) without saying anything like
"original was 'branch'".  There was no need to.  Also recent rewrite to
use parse-options made it impossible to parrot the original command line
without "unparsing".

This removes the code that implements the misguided "we dereferenced the
tag so record that in the commit message" behaviour.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agouser-manual: failed push to public repository
J. Bruce Fields [Sun, 25 Nov 2007 22:54:19 +0000 (17:54 -0500)]
user-manual: failed push to public repository

More details on the case of a failed push to a public (non-shared)
repository.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
17 years agouser-manual: define "branch" and "working tree" at start
J. Bruce Fields [Sun, 25 Nov 2007 18:53:37 +0000 (13:53 -0500)]
user-manual: define "branch" and "working tree" at start

Some explanation here might help.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
17 years agogit-checkout: describe detached head correctly
Junio C Hamano [Sun, 25 Nov 2007 18:34:27 +0000 (10:34 -0800)]
git-checkout: describe detached head correctly

When you have a file called HEAD in the work tree, the code to report
where the HEAD is at when "git checkout $commit^0" is done triggered
unnecessary ambiguity checking.

Explicitly mark the command line with "--" and make it clear that we are
talking about a revision.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoUpdate draft release notes for 1.5.4
Junio C Hamano [Sun, 25 Nov 2007 01:42:36 +0000 (17:42 -0800)]
Update draft release notes for 1.5.4

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMerge branch 'maint'
Junio C Hamano [Sun, 25 Nov 2007 03:52:45 +0000 (19:52 -0800)]
Merge branch 'maint'

* maint:
  Fix rev-list when showing objects involving submodules
  test format-patch -s: make sure MIME content type is shown as needed
  format-patch -s: add MIME encoding header if signer's name requires so

17 years agoMerge branch 'rv/maint-index-commit' into maint
Junio C Hamano [Sun, 25 Nov 2007 02:03:29 +0000 (18:03 -0800)]
Merge branch 'rv/maint-index-commit' into maint

* rv/maint-index-commit:
  Make GIT_INDEX_FILE apply to git-commit

17 years agoMerge branch 'lt/maint-rev-list-gitlink' into maint
Junio C Hamano [Sun, 25 Nov 2007 02:03:20 +0000 (18:03 -0800)]
Merge branch 'lt/maint-rev-list-gitlink' into maint

* lt/maint-rev-list-gitlink:
  Fix rev-list when showing objects involving submodules

17 years agoMerge branch 'jc/maint-add-sync-stat' into maint
Junio C Hamano [Sun, 25 Nov 2007 02:03:04 +0000 (18:03 -0800)]
Merge branch 'jc/maint-add-sync-stat' into maint

* jc/maint-add-sync-stat:
  t2200: test more cases of "add -u"
  git-add: make the entry stat-clean after re-adding the same contents
  ce_match_stat, run_diff_files: use symbolic constants for readability

17 years agoMerge branch 'jc/maint-format-patch-encoding' into maint
Junio C Hamano [Sun, 25 Nov 2007 02:02:30 +0000 (18:02 -0800)]
Merge branch 'jc/maint-format-patch-encoding' into maint

* jc/maint-format-patch-encoding:
  test format-patch -s: make sure MIME content type is shown as needed
  format-patch -s: add MIME encoding header if signer's name requires so

17 years agoMerge branch 'bs/maint-t7005' into maint
Junio C Hamano [Sun, 25 Nov 2007 02:01:46 +0000 (18:01 -0800)]
Merge branch 'bs/maint-t7005' into maint

* bs/maint-t7005:
  t7005-editor.sh: Don't invoke real vi when it is in GIT_EXEC_PATH

17 years agoMerge branch 'bs/maint-commit-options' into maint
Junio C Hamano [Sun, 25 Nov 2007 01:54:08 +0000 (17:54 -0800)]
Merge branch 'bs/maint-commit-options' into maint

* bs/maint-commit-options:
  git-commit: Add tests for invalid usage of -a/--interactive with paths
  git-commit.sh: Fix usage checks regarding paths given when they do not make sense

17 years agoDeprecate peek-remote
Junio C Hamano [Sun, 25 Nov 2007 01:41:18 +0000 (17:41 -0800)]
Deprecate peek-remote

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agot4119: correct overeager war-on-whitespace
Junio C Hamano [Sat, 24 Nov 2007 04:13:41 +0000 (20:13 -0800)]
t4119: correct overeager war-on-whitespace

Earlier a6080a0a44d5ead84db3dabbbc80e82df838533d (War on whitespace)
dropped a necessary trailing whitespace from the test vector.

17 years agoMerge branch 'jk/send-pack'
Junio C Hamano [Sun, 25 Nov 2007 00:45:37 +0000 (16:45 -0800)]
Merge branch 'jk/send-pack'

* jk/send-pack: (24 commits)
  send-pack: cluster ref status reporting
  send-pack: fix "everything up-to-date" message
  send-pack: tighten remote error reporting
  make "find_ref_by_name" a public function
  Fix warning about bitfield in struct ref
  send-pack: assign remote errors to each ref
  send-pack: check ref->status before updating tracking refs
  send-pack: track errors for each ref
  git-push: add documentation for the newly added --mirror mode
  Add tests for git push'es mirror mode
  Update the tracking references only if they were succesfully updated on remote
  Add a test checking if send-pack updated local tracking branches correctly
  git-push: plumb in --mirror mode
  Teach send-pack a mirror mode
  send-pack: segfault fix on forced push
  Reteach builtin-ls-remote to understand remotes
  send-pack: require --verbose to show update of tracking refs
  receive-pack: don't mention successful updates
  more terse push output
  Build in ls-remote
  ...

17 years agoMerge branch 'sb/clean'
Junio C Hamano [Sun, 25 Nov 2007 00:31:34 +0000 (16:31 -0800)]
Merge branch 'sb/clean'

* sb/clean:
  Teach git clean to use setup_standard_excludes()
  git-clean: Fix error message if clean.requireForce is not set.
  Make git-clean a builtin

17 years agoMerge branch 'js/mingw-fallouts'
Junio C Hamano [Sun, 25 Nov 2007 00:31:25 +0000 (16:31 -0800)]
Merge branch 'js/mingw-fallouts'

* js/mingw-fallouts:
  fetch-pack: Prepare for a side-band demultiplexer in a thread.
  rehabilitate some t5302 tests on 32-bit off_t machines
  Allow ETC_GITCONFIG to be a relative path.
  Introduce git_etc_gitconfig() that encapsulates access of ETC_GITCONFIG.
  Allow a relative builtin template directory.
  Close files opened by lock_file() before unlinking.
  builtin run_command: do not exit with -1.
  Move #include <sys/select.h> and <sys/ioctl.h> to git-compat-util.h.
  Use is_absolute_path() in sha1_file.c.
  Skip t3902-quoted.sh if the file system does not support funny names.
  t5302-pack-index: Skip tests of 64-bit offsets if necessary.
  t7501-commit.sh: Not all seds understand option -i
  t5300-pack-object.sh: Split the big verify-pack test into smaller parts.

17 years agoMerge branch 'mh/rebase-skip-hard'
Junio C Hamano [Sun, 25 Nov 2007 00:31:08 +0000 (16:31 -0800)]
Merge branch 'mh/rebase-skip-hard'

* mh/rebase-skip-hard:
  Do git reset --hard HEAD when using git rebase --skip

17 years agoMerge branch 'cc/bisect'
Junio C Hamano [Sun, 25 Nov 2007 00:31:02 +0000 (16:31 -0800)]
Merge branch 'cc/bisect'

* cc/bisect:
  Bisect reset: do nothing when not bisecting.
  Bisect: use "$GIT_DIR/BISECT_NAMES" to check if we are bisecting.
  Bisect visualize: use "for-each-ref" to list all good refs.
  git-bisect: modernize branch shuffling hack
  git-bisect: use update-ref to mark good/bad commits
  git-bisect: war on "sed"
  Bisect reset: remove bisect refs that may have been packed.

17 years agoMerge branch 'maint'
Junio C Hamano [Fri, 23 Nov 2007 07:25:42 +0000 (23:25 -0800)]
Merge branch 'maint'

* maint:
  Fix sample pre-commit hook

17 years agoFix sample pre-commit hook
Junio C Hamano [Fri, 23 Nov 2007 07:25:19 +0000 (23:25 -0800)]
Fix sample pre-commit hook

If the worktree happened to have a file called HEAD, "diff-index --cached HEAD"
would complain about the ambiguity between revision and path.  Avoid it by
using an explicit "--" for disambiguation.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoAddendum to "MaintNotes"
Junio C Hamano [Thu, 22 Nov 2007 11:08:39 +0000 (06:08 -0500)]
Addendum to "MaintNotes"

Add "how to maintain git" document. Foreward by Jeff King.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMake test scripts executable.
Junio C Hamano [Fri, 23 Nov 2007 00:52:17 +0000 (16:52 -0800)]
Make test scripts executable.

17 years agoMerge branch 'maint'
Junio C Hamano [Fri, 23 Nov 2007 00:51:18 +0000 (16:51 -0800)]
Merge branch 'maint'

* maint:
  Make test scripts executable.
  bundle create: keep symbolic refs' names instead of resolving them

17 years agoMake test scripts executable.
Junio C Hamano [Fri, 23 Nov 2007 00:48:55 +0000 (16:48 -0800)]
Make test scripts executable.

17 years agorebase -i: move help to end of todo file
Johannes Schindelin [Thu, 22 Nov 2007 12:30:10 +0000 (12:30 +0000)]
rebase -i: move help to end of todo file

[PATCH] rebase -i: move help to end of todo file

Many editors start in the first line, so the 9-line help text was an
annoyance.  So move it to the end.

Requested by Junio.

While at it, add a hint how to abort the rebase.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agobundle create: keep symbolic refs' names instead of resolving them
Johannes Schindelin [Thu, 22 Nov 2007 12:24:59 +0000 (12:24 +0000)]
bundle create: keep symbolic refs' names instead of resolving them

When creating a bundle, symbolic refs used to be resolved to the
non-symbolic refs they point to before being written to the list
of contained refs.  I.e. "git bundle create a1.bundle HEAD master"
would show something like

388afe7881b33102fada216dd07806728773c011        refs/heads/master
388afe7881b33102fada216dd07806728773c011        refs/heads/master

instead of

388afe7881b33102fada216dd07806728773c011        HEAD
388afe7881b33102fada216dd07806728773c011        refs/heads/master

Introduce a special handling so that the symbolic refs are listed
with the names passed on the command line.

Noticed by Santi Béjar.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogit-merge-ours: make it a builtin.
Thomas Harning [Thu, 22 Nov 2007 20:19:40 +0000 (15:19 -0500)]
git-merge-ours: make it a builtin.

Except that this fixes a longstanding corner case bug by
tightening the way underlying diff-index command is run, it is
functionally equivalent to the scripted version.

Signed-off-by: Thomas Harning Jr <harningt@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoRefactor patch_update_cmd
Wincent Colaiuta [Wed, 21 Nov 2007 12:36:38 +0000 (13:36 +0100)]
Refactor patch_update_cmd

Split patch_update_cmd into two functions, one to prompt the user for
a path to patch and another to do the actual work given that file path.
This lays the groundwork for a future commit which will teach
git-add--interactive to accept a path parameter and jump directly to
the patch subcommand for that path, bypassing the interactive prompt.

Signed-off-by: Wincent Colaiuta <win@wincent.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoAuthenticate only once in git-send-email
Wincent Colaiuta [Wed, 21 Nov 2007 12:35:05 +0000 (13:35 +0100)]
Authenticate only once in git-send-email

When using git-send-email with SMTP authentication sending a patch series
would redundantly authenticate multiple times, once for each patch. In
the worst case, this would actually prevent the series from being sent
because the server would reply with a "5.5.0 Already Authenticated"
status code which would derail the process.

This commit teaches git-send-email to authenticate once and only once at
the beginning of the series.

Signed-off-by: Wincent Colaiuta <win@wincent.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMerge git://git.bogomips.org/git-svn
Junio C Hamano [Thu, 22 Nov 2007 08:34:49 +0000 (00:34 -0800)]
Merge git://git.bogomips.org/git-svn

* git://git.bogomips.org/git-svn:
  git-svn: allow `info' command to work offline
  git-svn: info --url [path]
  git-svn info: implement info command
  git-svn: extract reusable code into utility functions
  t9106: fix a race condition that caused svn to miss modifications

17 years agoFix "quote" misconversion for rewrite diff output.
Junio C Hamano [Thu, 22 Nov 2007 07:06:44 +0000 (23:06 -0800)]
Fix "quote" misconversion for rewrite diff output.

663af3422a648e87945e4d8c0cc3e13671f2bbde (Full rework of
quote_c_style and write_name_quoted.) mistakenly used puts()
when writing out a fixed string when it did not want to add a
terminating LF.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogit-svn: allow `info' command to work offline
Eric Wong [Thu, 22 Nov 2007 02:20:57 +0000 (18:20 -0800)]
git-svn: allow `info' command to work offline

Cache the repository root whenever we connect to the repository.
This will allow us to notice URL changes if the user changes the
URL in .git/config, too.

If the repository is no longer accessible, or if `git svn info'
is the first and only command run; then '(offline)' will be
displayed for "Repository Root:" in the output.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
17 years agogit-svn: info --url [path]
David D. Kilzer [Wed, 21 Nov 2007 19:57:19 +0000 (11:57 -0800)]
git-svn: info --url [path]

Return the svn URL for the given path, or return the svn
repository URL if no path is given.

Added 18 tests to t/t9119-git-svn-info.sh.

Signed-off-by: David D. Kilzer <ddkilzer@kilzer.net>
Acked-by: Eric Wong <normalperson@yhbt.net>
17 years agogit-svn info: implement info command
David D. Kilzer [Wed, 21 Nov 2007 19:57:18 +0000 (11:57 -0800)]
git-svn info: implement info command

Implement "git-svn info" for files and directories based on the
"svn info" command.  Note that the -r/--revision argument is not
supported yet.

Added 18 tests in t/t9119-git-svn-info.sh.

[ew: small fix to work without arguments on all working directories]

Signed-off-by: David D. Kilzer <ddkilzer@kilzer.net>
Acked-by: Eric Wong <normalperson@yhbt.net>
17 years agogit-svn: extract reusable code into utility functions
David D. Kilzer [Wed, 21 Nov 2007 06:43:17 +0000 (22:43 -0800)]
git-svn: extract reusable code into utility functions

Extacted canonicalize_path() in the main package.

Created new Git::SVN::Util package with an md5sum() function.  A
new package was created so that Digest::MD5 did not have to be
loaded in the main package.  Replaced code in the SVN::Git::Editor
and SVN::Git::Fetcher packages with calls to md5sum().

Extracted the format_svn_date(), parse_git_date() and
set_local_timezone() functions within the Git::SVN::Log package.

Signed-off-by: David D. Kilzer <ddkilzer@kilzer.net>
Acked-by: Eric Wong <normalperson@yhbt.net>
17 years agot9106: fix a race condition that caused svn to miss modifications
Eric Wong [Wed, 21 Nov 2007 08:57:33 +0000 (00:57 -0800)]
t9106: fix a race condition that caused svn to miss modifications

carbonated beverage noticed this test was occasionally failing.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
17 years agoMerge git://repo.or.cz/git-gui
Junio C Hamano [Wed, 21 Nov 2007 08:00:56 +0000 (00:00 -0800)]
Merge git://repo.or.cz/git-gui

* git://repo.or.cz/git-gui: (96 commits)
  git-gui 0.9.0
  git-gui: Bind Meta-T for "Stage To Commit" menu action
  git-gui: Allow users to set font weights to bold
  git-gui: Update Japanese strings (part 2)
  git-gui: Update Japanese strings
  Updated russian translation of git-gui
  po2msg: actually output statistics
  po2msg: ignore untranslated messages
  po2msg: ignore entries marked with "fuzzy"
  git-gui: Protect against bad translation strings
  git-gui: Make sure we get errors from git-update-index
  More updates and corrections to the russian translation of git-gui
  Updated Russian translation.
  git-gui: Update German translation
  git-gui: Add more terms to glossary.
  git-gui: Paper bag fix the global config parsing
  git-gui: Honor a config.mak in git-gui's top level
  git-gui: Collapse $env(HOME) to ~/ in recent repositories on Windows
  git-gui: Support cloning Cygwin based work-dirs
  git-gui: Use proper Windows shortcuts instead of bat files
  ...

17 years agogit-p4: Fix typo in --detect-labels
Shun Kei Leung [Wed, 21 Nov 2007 03:01:19 +0000 (11:01 +0800)]
git-p4: Fix typo in --detect-labels

Signed-off-by: Kevin Leung <kevinlsk@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agosend-pack: cluster ref status reporting
Jeff King [Tue, 20 Nov 2007 11:18:01 +0000 (06:18 -0500)]
send-pack: cluster ref status reporting

Instead of intermingling success and failure, we now print:

  1. all uptodate refs (if args.verbose is enabled)
  2. successfully pushed refs
  3. failed refs

with the assumption that the user is most likely to see the
ones at the end, and therefore we order them from "least
interesting" to "most interesting."

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoavoid "defined but not used" warning for fetch_objs_via_walker
Jeff King [Sun, 18 Nov 2007 08:17:23 +0000 (03:17 -0500)]
avoid "defined but not used" warning for fetch_objs_via_walker

Because this function is static and used only by the
http-walker, when NO_CURL is defined, gcc emits a "defined
but not used" warning.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogit-gui 0.9.0 gitgui-0.9.0
Shawn O. Pearce [Wed, 21 Nov 2007 07:10:03 +0000 (02:10 -0500)]
git-gui 0.9.0

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
17 years agoMerge branch 'maint'
Junio C Hamano [Wed, 21 Nov 2007 07:00:07 +0000 (23:00 -0800)]
Merge branch 'maint'

* maint:
  send-email: add transfer encoding header with content-type
  Doc fix for git-reflog: mention @{...} syntax, and <ref> in synopsys.
  config: clarify compression defaults
  config: correct core.loosecompression documentation

17 years agosend-email: add transfer encoding header with content-type
Jeff King [Tue, 20 Nov 2007 12:54:04 +0000 (07:54 -0500)]
send-email: add transfer encoding header with content-type

We add the content-type header only when we have non-7bit
characters from the 'From' header, so we really need to
specify the encoding (in other cases, where the commit text
needed a content-type, git-format-patch will already have
added the encoding header).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogit-compat-util.h: auto-adjust to compiler support of FLEX_ARRAY a bit better
Junio C Hamano [Tue, 20 Nov 2007 20:08:06 +0000 (12:08 -0800)]
git-compat-util.h: auto-adjust to compiler support of FLEX_ARRAY a bit better

When declaring a structure with a flexible array member, instead
of defaulting to the c99 syntax for non-gnu compilers (which
burned people with older compilers), default to the traditional
and more portable "member[1]; /* more */" syntax.

At the same time, other c99 compilers should be able to take
advantage of the modern syntax to flexible array members without
being gcc.  Check __STDC_VERSION__ for that.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoFlush progress message buffer in display().
Johannes Sixt [Mon, 19 Nov 2007 19:48:58 +0000 (20:48 +0100)]
Flush progress message buffer in display().

This will make progress display from pack-objects (invoked via
upload-pack) more responsive on platforms with an implementation
of stdio whose stderr is line buffered.

The standard error stream is defined to be merely "not fully
buffered"; this is different from "unbuffered".  If the
implementation of the stdio library chooses to make it line
buffered, progress reports that end with CR but do not contain
LF will accumulate in the stdio buffer before written out.  A
fflush() after each progress display gives a nice continuous
progress.

Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoautoconf: Add tests for memmem, strtoumax and mkdtemp functions
Jakub Narebski [Mon, 19 Nov 2007 18:47:05 +0000 (19:47 +0100)]
autoconf: Add tests for memmem, strtoumax and mkdtemp functions

Update configure.ac (and config.mak.in) to keep up with git
development by adding tests for memmem (NO_MEMMEM), strtoumax
(NO_STRTOUMAX) and mkdtemp (NO_MKDTEMP) functions.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogitweb: Put project README in div.readme, fix its padding
Jakub Narebski [Mon, 19 Nov 2007 13:16:12 +0000 (14:16 +0100)]
gitweb: Put project README in div.readme, fix its padding

Put (optional) projects README on "summary" page in <div> element
using "readme" class.  This allow to style it using CSS.

Add padding to project's README to make it line out with the rest
of the page.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
17 years agogitweb: Style all tables using CSS
Jakub Narebski [Mon, 19 Nov 2007 13:16:11 +0000 (14:16 +0100)]
gitweb: Style all tables using CSS

Remove all cellspacing="0" attributes from tables in gitweb,
replacing it by CSS rule.  Add CSS classes for all tables.

While at it, change class(es) of table for commit message and commit
authorship search from "grep" to "commit_search"; similarly,
"grep_search" class is now used for table with results of grep (files)
search.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
17 years agosend-email: Don't add To: recipients to the Cc: header
Ask Bjørn Hansen [Mon, 19 Nov 2007 11:00:26 +0000 (03:00 -0800)]
send-email: Don't add To: recipients to the Cc: header

Signed-off-by: Ask Bjørn Hansen <ask@develooper.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoBisect reset: do nothing when not bisecting.
Christian Couder [Tue, 20 Nov 2007 05:39:53 +0000 (06:39 +0100)]
Bisect reset: do nothing when not bisecting.

Before this patch, using "git bisect reset" when not bisecting
did a "git checkout master" for no good reason.

This also happened using "git bisect replay" when not bisecting
because "bisect_replay" starts by calling "bisect_reset".

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoDoc fix for git-reflog: mention @{...} syntax, and <ref> in synopsys.
Matthieu Moy [Mon, 19 Nov 2007 18:25:11 +0000 (19:25 +0100)]
Doc fix for git-reflog: mention @{...} syntax, and <ref> in synopsys.

The HEAD@{...} syntax was documented in git-rev-parse manpage, which
is hard to find by someone looking for the documentation of porcelain.
git-reflog is probably the place where one expects to find this.

While I'm there, "git revlog show whatever" was also undocumented.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoconfig: clarify compression defaults
Brian Downing [Mon, 19 Nov 2007 16:58:51 +0000 (10:58 -0600)]
config: clarify compression defaults

* Clarify that core.compression provides a system-wide default to
  other compression parameters.

* Explain that the default for pack.compression, -1, is "a default
  compromise between speed and compression (currently equivalent
  to level 6)" according to zlib.h.

Signed-off-by: Brian Downing <bdowning@lavos.net>
Acked-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoconfig: correct core.loosecompression documentation
Brian Downing [Mon, 19 Nov 2007 16:58:50 +0000 (10:58 -0600)]
config: correct core.loosecompression documentation

* core.loosecompression stated that the default was "0 (best speed)",
  when in fact 0 is "no compression", and the default is Z_BEST_SPEED,
  which is 1.

Signed-off-by: Brian Downing <bdowning@lavos.net>
Acked-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogitview: import only one of gtksourceview and gtksourceview2
Anton Gyllenberg [Mon, 19 Nov 2007 10:37:16 +0000 (12:37 +0200)]
gitview: import only one of gtksourceview and gtksourceview2

Importing both gtksourceview and gtksourceview2 will make python segfault
on my system (ubuntu 7.10). Change so that gtksourceview is only imported
if importing gtksourceview2 fails. This should be safe as gtksourceview
is only used if gtksourceview2 is not available.

Signed-off-by: Anton Gyllenberg <anton@iki.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogit-send-email: show all headers when sending mail
David D. Kilzer [Mon, 19 Nov 2007 04:14:55 +0000 (20:14 -0800)]
git-send-email: show all headers when sending mail

As a git newbie, it was confusing to set an In-Reply-To header but then
not see it printed when the git-send-email command was run.

This patch prints all headers that would be sent to sendmail or an SMTP
server instead of only printing From, Subject, Cc, To.  It also removes
the now-extraneous Date header after the "Log says" line.

Added test to t/t9001-send-email.sh.

Signed-off-by: David D. Kilzer <ddkilzer@kilzer.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMerge branch 'maint'
Junio C Hamano [Mon, 19 Nov 2007 08:19:15 +0000 (00:19 -0800)]
Merge branch 'maint'

* maint:
  Documentation: Fix references to deprecated commands
  user-manual: mention "..." in "Generating diffs", etc.
  user-manual: Add section "Why bisecting merge commits can be harder ..."
  git-remote.txt: fix example url

17 years agoFurther clarify clean.requireForce changes
Wincent Colaiuta [Mon, 19 Nov 2007 08:06:31 +0000 (09:06 +0100)]
Further clarify clean.requireForce changes

Mention the -f switch in the release notes for clean.requireForce to avoid
possible misunderstandings.

Signed-off-by: Wincent Colaiuta <win@wincent.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMerge branch 'maint' of git://linux-nfs.org/~bfields/git into maint
Junio C Hamano [Mon, 19 Nov 2007 07:56:01 +0000 (23:56 -0800)]
Merge branch 'maint' of git://linux-nfs.org/~bfields/git into maint

* 'maint' of git://linux-nfs.org/~bfields/git:
  Documentation: Fix references to deprecated commands
  user-manual: mention "..." in "Generating diffs", etc.
  user-manual: Add section "Why bisecting merge commits can be harder ..."
  git-remote.txt: fix example url

17 years agoTeach git clean to use setup_standard_excludes()
Shawn Bohrer [Thu, 15 Nov 2007 05:00:54 +0000 (23:00 -0600)]
Teach git clean to use setup_standard_excludes()

Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agogit-clean: Fix error message if clean.requireForce is not set.
Junio C Hamano [Tue, 13 Nov 2007 05:13:05 +0000 (21:13 -0800)]
git-clean: Fix error message if clean.requireForce is not set.

It was distracting to see this error message:

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

even though clean.requireForce was not set at all. This patch distinguishes
the cases and gives a different message depending on whether the
configuration variable is not set or set to true.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMake git-clean a builtin
Shawn Bohrer [Mon, 12 Nov 2007 01:48:47 +0000 (19:48 -0600)]
Make git-clean a builtin

This replaces git-clean.sh with builtin-clean.c, and moves
git-clean.sh to the examples.

This also introduces a change in behavior when removing directories
explicitly specified as a path.  For example currently:

1. When dir has only untracked files, these two behave differently:

    $ git clean -n dir
    $ git clean -n dir/

the former says "Would not remove dir/", while the latter would say
"Would remove dir/untracked" for all paths under it, but not the
directory itself.

With -d, the former would stop refusing, however since the user
explicitly asked to remove the directory the -d is no longer required.

2. When there are more parameters:

    $ git clean -n dir foo
    $ git clean -n dir/ foo

both cases refuse to remove dir/ unless -d is specified.  Once again
since both cases requested to remove dir the -d is no longer required.

Thanks to Johannes Schindelin for the conversion to using the
parse-options API.

Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoDocumentation: Fix references to deprecated commands
J. Bruce Fields [Mon, 19 Nov 2007 01:50:57 +0000 (20:50 -0500)]
Documentation: Fix references to deprecated commands

... by changing git-tar-tree reference to git-archive and removing
seemingly unrelevant footnote about git-ssh-{fetch,upload}.

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
17 years agoBisect: use "$GIT_DIR/BISECT_NAMES" to check if we are bisecting.
Christian Couder [Sun, 18 Nov 2007 15:34:03 +0000 (16:34 +0100)]
Bisect: use "$GIT_DIR/BISECT_NAMES" to check if we are bisecting.

Previously we tested if the "$GIT_DIR/refs/bisect" directory
existed, to check if we were bisecting.

Now with packed refs, it is simpler to check if the file
"$GIT_DIR/BISECT_NAMES" exists, as it is already created when
starting bisection and removed when reseting bisection.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoUpdate draft release notes for 1.5.4
Junio C Hamano [Mon, 19 Nov 2007 00:24:14 +0000 (16:24 -0800)]
Update draft release notes for 1.5.4

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agouser-manual: mention "..." in "Generating diffs", etc.
J. Bruce Fields [Mon, 19 Nov 2007 00:18:27 +0000 (19:18 -0500)]
user-manual: mention "..." in "Generating diffs", etc.

We should mention the use of the "..." syntax for git-diff here.  The
note about the difference between diff and the combined output of
git-format-patch then no longer fits so well, so remove it.  Add a
reference to the git-format-patch[1] manpage.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
17 years agoMerge branch 'lt/rev-list-gitlink'
Junio C Hamano [Mon, 19 Nov 2007 00:16:37 +0000 (16:16 -0800)]
Merge branch 'lt/rev-list-gitlink'

* lt/rev-list-gitlink:
  Fix rev-list when showing objects involving submodules

17 years agoMerge branch 'ds/checkout-upper'
Junio C Hamano [Mon, 19 Nov 2007 00:04:17 +0000 (16:04 -0800)]
Merge branch 'ds/checkout-upper'

* ds/checkout-upper:
  git-checkout: Test for relative path use.
  git-checkout: Support relative paths containing "..".

17 years agoMerge branch 'sh/p4'
Junio C Hamano [Mon, 19 Nov 2007 00:03:58 +0000 (16:03 -0800)]
Merge branch 'sh/p4'

* sh/p4:
  git-p4: Fix direct import from perforce after fetching changes through git from origin

17 years agoMerge branch 'lt/rev-list-interactive'
Junio C Hamano [Mon, 19 Nov 2007 00:03:24 +0000 (16:03 -0800)]
Merge branch 'lt/rev-list-interactive'

* lt/rev-list-interactive:
  Fix parent rewriting in --early-output
  revision walker: mini clean-up
  Enhance --early-output format
  Add "--early-output" log flag for interactive GUI use
  Simplify topo-sort logic

17 years agoMerge branch 'ph/diffopts'
Junio C Hamano [Sun, 18 Nov 2007 23:50:16 +0000 (15:50 -0800)]
Merge branch 'ph/diffopts'

* ph/diffopts:
  Reorder diff_opt_parse options more logically per topics.
  Make the diff_options bitfields be an unsigned with explicit masks.
  Use OPT_BIT in builtin-pack-refs
  Use OPT_BIT in builtin-for-each-ref
  Use OPT_SET_INT and OPT_BIT in builtin-branch
  parse-options new features.

17 years agoDraft release notes: fix clean.requireForce description
Junio C Hamano [Sun, 18 Nov 2007 23:22:51 +0000 (15:22 -0800)]
Draft release notes: fix clean.requireForce description

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMerge branch 'maint' to synchronize with 1.5.3.6
Junio C Hamano [Sun, 18 Nov 2007 23:15:47 +0000 (15:15 -0800)]
Merge branch 'maint' to synchronize with 1.5.3.6

* maint:
  GIT 1.5.3.6
  grep -An -Bm: fix invocation of external grep command

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agouser-manual: Add section "Why bisecting merge commits can be harder ..."
Steffen Prohaska [Sat, 10 Nov 2007 13:49:54 +0000 (14:49 +0100)]
user-manual: Add section "Why bisecting merge commits can be harder ..."

This commit adds a discussion of the challenge of bisecting
merge commits to the user manual.  The original author is
Junio C Hamano <gitster@pobox.com>, who posted the text to
the mailing list <http://marc.info/?l=git&m=119403257315527&w=2>.
His email was adapted for the manual.

The discussion is added to "Rewriting history and maintainig
patch series".  The text added requires good understanding of
merging and rebasing.  Therefore it should not be placed too
early in the manual.  Right after the section on "Problems with
rewriting history", the discussion of bisect gives another reason
for linearizing as much of the history as possible.

The text includes suggestions and fixes by
Ralf Wildenhues <Ralf.Wildenhues@gmx.de> and
Benoit Sigoure <tsuna@lrde.epita.fr>.

Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
17 years agoGIT 1.5.3.6 v1.5.3.6
Junio C Hamano [Sun, 18 Nov 2007 22:00:38 +0000 (14:00 -0800)]
GIT 1.5.3.6

17 years agoUse compat mkdtemp() on Solaris boxes
Guido Ostkamp [Fri, 16 Nov 2007 18:59:58 +0000 (19:59 +0100)]
Use compat mkdtemp() on Solaris boxes

Define NO_MKDTEMP for all variants of SunOS; Solaris 10 does not
have mkdtemp() and all the other versions our Makefile knows
about don't have it either.

NO_{SETENV,UNSETENV,C99_FORMAT,STRTOUMAX} definitions cannot be
unified across versions.  Beginning with Solaris 10, the C-library
provides unsetenv(), setenv() and strtoumax().  Also 'z'/'t' formats
are supported.  However, older versions of Solaris do not support
these.

Signed-off-by: Guido Ostkamp <git@ostkamp.fastmail.fm>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agosend-pack: fix "everything up-to-date" message
Jeff King [Sun, 18 Nov 2007 08:08:22 +0000 (03:08 -0500)]
send-pack: fix "everything up-to-date" message

This has always been slightly inaccurate, since it used the
new_refs counter, which really meant "did we send any
objects," so deletions were not counted.

It has gotten even worse with recent patches, since we no
longer look at the 'ret' value, meaning we would say "up to
date" if non-ff pushes were rejected.

Instead, we now claim up to date iff every ref is either
unmatched or up to date. Any other case should already have
generated a status line.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agosend-pack: tighten remote error reporting
Jeff King [Sun, 18 Nov 2007 07:16:52 +0000 (02:16 -0500)]
send-pack: tighten remote error reporting

Previously, we set all ref pushes to 'OK', and then marked
them as errors if the remote reported so. This has the
problem that if the remote dies or fails to report a ref, we
just assume it was OK.

Instead, we use a new non-OK state to indicate that we are
expecting status (if the remote doesn't support the
report-status feature, we fall back on the old behavior).
Thus we can flag refs for which we expected a status, but
got none (conversely, we now also print a warning for refs
for which we get a status, but weren't expecting one).

This also allows us to simplify the receive_status exit
code, since each ref is individually marked with failure
until we get a success response. We can just print the usual
status table, so the user still gets a sense of what we were
trying to do when the failure happened.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agomake "find_ref_by_name" a public function
Jeff King [Sun, 18 Nov 2007 07:13:10 +0000 (02:13 -0500)]
make "find_ref_by_name" a public function

This was a static in remote.c, but is generally useful.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoFix warning about bitfield in struct ref
Shawn O. Pearce [Sun, 18 Nov 2007 09:31:37 +0000 (04:31 -0500)]
Fix warning about bitfield in struct ref

cache.h:503: warning: type of bit-field 'force' is a GCC extension
cache.h:504: warning: type of bit-field 'merge' is a GCC extension
cache.h:505: warning: type of bit-field 'nonfastforward' is a GCC extension
cache.h:506: warning: type of bit-field 'deletion' is a GCC extension

So we change it to an 'unsigned int' which is not a GCC extension.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agoMerge branch 'ph/parseopt-sh'
Junio C Hamano [Sun, 18 Nov 2007 05:39:37 +0000 (21:39 -0800)]
Merge branch 'ph/parseopt-sh'

* ph/parseopt-sh:
  git-quiltimport.sh fix --patches handling
  git-am: -i does not take a string parameter.
  sh-setup: don't let eval output to be shell-expanded.
  git-sh-setup: fix parseopt `eval` string underquoting
  Give git-am back the ability to add Signed-off-by lines.
  git-rev-parse --parseopt
  scripts: Add placeholders for OPTIONS_SPEC
  Migrate git-repack.sh to use git-rev-parse --parseopt
  Migrate git-quiltimport.sh to use git-rev-parse --parseopt
  Migrate git-checkout.sh to use git-rev-parse --parseopt --keep-dashdash
  Migrate git-instaweb.sh to use git-rev-parse --parseopt
  Migrate git-merge.sh to use git-rev-parse --parseopt
  Migrate git-am.sh to use git-rev-parse --parseopt
  Migrate git-clone to use git-rev-parse --parseopt
  Migrate git-clean.sh to use git-rev-parse --parseopt.
  Update git-sh-setup(1) to allow transparent use of git-rev-parse --parseopt
  Add a parseopt mode to git-rev-parse to bring parse-options to shell scripts.

17 years agogrep -An -Bm: fix invocation of external grep command
Junio C Hamano [Sun, 18 Nov 2007 05:18:14 +0000 (21:18 -0800)]
grep -An -Bm: fix invocation of external grep command

When building command line to invoke external grep, the
arguments to -A/-B/-C options were placd in randarg[] buffer,
but the code forgot that snprintf() does not count terminating
NUL in its return value.  This caused "git grep -A1 -B2" to
invoke external grep with "-B21 -A1".

Signed-off-by: Junio C Hamano <gitster@pobox.com>
17 years agofetch-pack: Prepare for a side-band demultiplexer in a thread.
Johannes Sixt [Sat, 17 Nov 2007 22:09:28 +0000 (23:09 +0100)]
fetch-pack: Prepare for a side-band demultiplexer in a thread.

get_pack() receives a pair of file descriptors that communicate with
upload-pack at the remote end. In order to support the case where the
side-band demultiplexer runs in a thread, and, hence, in the same process
as the main routine, we must not close the readable file descriptor early.

The handling of the readable fd is changed in the case where upload-pack
supports side-band communication: The old code closed the fd after it was
inherited to the side-band demultiplexer process. Now we do not close it.
The caller (do_fetch_pack) will close it later anyway. The demultiplexer
is the only reader, it does not matter that the fd remains open in the
main process as well as in unpack-objects/index-pack, which inherits it.

The writable fd is not needed in get_pack(), hence, the old code closed
the fd. For symmetry with the readable fd, we now do not close it; the
caller (do_fetch_pack) will close it later anyway. Therefore, the new
behavior is that the channel now remains open during the entire
conversation, but this has no ill effects because upload-pack does not read
from it once it has begun to send the pack data. For the same reason it
does not matter that the writable fd is now inherited to the demultiplexer
and unpack-objects/index-pack processes.

Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>