git/git.git
12 years agohttp_init(): Fix config file parsing
Junio C Hamano [Tue, 10 Mar 2009 02:00:30 +0000 (19:00 -0700)]
http_init(): Fix config file parsing

We honor the command line options, environment variables, variables in
repository configuration file, variables in user's global configuration
file, variables in the system configuration file, and then finally use
built-in default.  To implement this semantics, the code should:

 - start from built-in default values;

 - call git_config() with the configuration parser callback, which
   implements "later definition overrides earlier ones" logic
   (git_config() reads the system's, user's and then repository's
   configuration file in this order);

 - override the result from the above with environment variables if set;

 - override the result from the above with command line options.

The initialization code http_init() for http transfer got this wrong, and
implemented a "first one wins, ignoring the later ones" in http_options(),
to compensate this mistake, read environment variables before calling
git_config().  This is all wrong.

As a second class citizen, the http codepath hasn't been audited as
closely as other parts of the system, but we should try to bring sanity to
it, before inviting contributors to improve on it.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agohttp.c: style cleanups
Junio C Hamano [Tue, 10 Mar 2009 01:47:29 +0000 (18:47 -0700)]
http.c: style cleanups

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge git://git.bogomips.org/git-svn
Junio C Hamano [Tue, 10 Mar 2009 01:38:18 +0000 (18:38 -0700)]
Merge git://git.bogomips.org/git-svn

* git://git.bogomips.org/git-svn:
  git-svn: support intermediate paths when matching tags/branches

12 years agoparseopt: prevent KEEP_UNKNOWN and STOP_AT_NON_OPTION from being used together
René Scharfe [Mon, 9 Mar 2009 20:57:38 +0000 (21:57 +0100)]
parseopt: prevent KEEP_UNKNOWN and STOP_AT_NON_OPTION from being used together

As suggested by Junio, disallow the flags PARSE_OPT_KEEP_UNKNOWN and
PARSE_OPT_STOP_AT_NON_OPTION to be turned on at the same time, as a
value of an unknown option could be mistakenly classified as a
non-option, stopping the parser early.  E.g.:

git cmd --known --unknown value arg0 arg1

The parser should have stopped at "arg0", but it already stops at
"value".

This patch makes parse_options() die if the two flags are used in
combination.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoparseopt: document KEEP_ARGV0, KEEP_UNKNOWN, NO_INTERNAL_HELP
René Scharfe [Mon, 9 Mar 2009 20:26:56 +0000 (21:26 +0100)]
parseopt: document KEEP_ARGV0, KEEP_UNKNOWN, NO_INTERNAL_HELP

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-svn: support intermediate paths when matching tags/branches
Michael Lai [Mon, 9 Mar 2009 18:45:47 +0000 (11:45 -0700)]
git-svn: support intermediate paths when matching tags/branches

For repositories laid out like the following:

[svn-remote "svn"]
      url = http://foo.com/svn/repos/bar
      fetch = myproject/trunk:refs/remotes/trunk
      branches = bar/myproject/branches/*:refs/remotes/*
      tags = bar/myproject/tags/*:refs/remotes/tags/*

The "bar" component above is considered the intermediate path
and was not handled correctly.

Signed-off-by: Michael Lai <myllai@gmail.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
12 years agoarchive: use parseopt for local-only options
René Scharfe [Sun, 8 Mar 2009 18:21:53 +0000 (19:21 +0100)]
archive: use parseopt for local-only options

Replace the hand-rolled parsers that find and remove --remote and --exec
by a parseopt parser that also handles --output.

All three options only have a meaning if no remote server is used or on
the local side.  They must be rejected by upload-archive and should not
be sent to the server by archive.

We can't use a single parser for both remote and local side because the
remote end possibly understands a different set of options than the
local side.  A local parser would then wrongly accuse options valid on
the other side as being incorrect.

This patch implements a very forgiving parser that understands only the
three options mentioned above.  All others are passed to the normal,
complete parser in archive.c (running either locally in archive, or
remotely in upload-archive).  This normal parser definition contains
dummy entries for the three options, in order for them to appear in the
help screen.

The parseopt parser allows multiple occurrences of --remote and --exec
unlike the previous one; the one specified last wins.  This looseness
is acceptable, I think.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoparseopt: make usage optional
René Scharfe [Sun, 8 Mar 2009 18:16:58 +0000 (19:16 +0100)]
parseopt: make usage optional

Allow usagestr to be NULL and don't display any help screen in
this case.  This is useful to implement incremental parsers.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoparseopt: add PARSE_OPT_NO_INTERNAL_HELP
René Scharfe [Sun, 8 Mar 2009 18:15:08 +0000 (19:15 +0100)]
parseopt: add PARSE_OPT_NO_INTERNAL_HELP

Add a parseopt flag, PARSE_OPT_NO_INTERNAL_HELP, that turns off internal
handling of -h, --help and --help-all.  This allows the implementation
of custom help option handlers or incremental parsers.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoparseopt: add PARSE_OPT_KEEP_UNKNOWN
René Scharfe [Sun, 8 Mar 2009 18:12:47 +0000 (19:12 +0100)]
parseopt: add PARSE_OPT_KEEP_UNKNOWN

Add a parseopt flag, PARSE_OPT_KEEP_UNKNOWN, that can be used to keep
unknown options in argv, similar to the existing KEEP flags.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'maint'
Junio C Hamano [Sun, 8 Mar 2009 06:34:13 +0000 (22:34 -0800)]
Merge branch 'maint'

* maint:
  builtin-revert.c: release index lock when cherry-picking an empty commit
  document config --bool-or-int
  t1300: use test_must_fail as appropriate
  cleanup: add isascii()
  Documentation: fix badly indented paragraphs in "--bisect-all" description

12 years agoMerge branch 'maint-1.6.1' into maint
Junio C Hamano [Sun, 8 Mar 2009 05:00:27 +0000 (21:00 -0800)]
Merge branch 'maint-1.6.1' into maint

* maint-1.6.1:
  builtin-revert.c: release index lock when cherry-picking an empty commit

12 years agohttp.c: use strbuf API in quote_ref_url
Tay Ray Chuan [Sat, 7 Mar 2009 16:47:21 +0000 (00:47 +0800)]
http.c: use strbuf API in quote_ref_url

In addition, ''quote_ref_url'' inserts a slash between the base URL and
remote ref path only if needed. Previously, this insertion wasn't
contingent on the lack of a separating slash.

Signed-off-by: Tay Ray Chuan <rctay89@gmail.com>
Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMove local variables to narrower scopes
Benjamin Kramer [Sat, 7 Mar 2009 20:02:26 +0000 (21:02 +0100)]
Move local variables to narrower scopes

These weren't used outside and can be safely moved

Signed-off-by: Benjamin Kramer <benny.kra@googlemail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoRemove unused function scope local variables
Benjamin Kramer [Sat, 7 Mar 2009 20:02:10 +0000 (21:02 +0100)]
Remove unused function scope local variables

These variables were unused and can be removed safely:

  builtin-clone.c::cmd_clone(): use_local_hardlinks, use_separate_remote
  builtin-fetch-pack.c::find_common(): len
  builtin-remote.c::mv(): symref
  diff.c::show_stats():show_stats(): total
  diffcore-break.c::should_break(): base_size
  fast-import.c::validate_raw_date(): date, sign
  fsck.c::fsck_tree(): o_sha1, sha1
  xdiff-interface.c::parse_num(): read_some

Signed-off-by: Benjamin Kramer <benny.kra@googlemail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMinGW: fix diff --no-index /dev/null ...
Johannes Schindelin [Sat, 7 Mar 2009 15:51:33 +0000 (16:51 +0100)]
MinGW: fix diff --no-index /dev/null ...

When launching "diff --no-index" with a parameter "/dev/null", the MSys
bash converts the "/dev/null" to a "nul", which usually makes sense.  But

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-p4: remove tabs from usermap file
Pete Wyckoff [Fri, 27 Feb 2009 18:53:59 +0000 (10:53 -0800)]
git-p4: remove tabs from usermap file

Some users have tabs in their names, oddly enough.  This
causes problems when loading the usercache from disk,
as split separates the fields on the wrong tabs.  When
fast-import's parse_ident() tries to parse the committer
field, it is unhappy about the unbalanced <..> angle brackets.

It is easy enough to convert the tabs to single spaces.

Signed-off-by: Pete Wyckoff <pw@padd.com>
Acked-by: Simon Hausmann <simon@lst.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoAdd an (optional, since expensive) test for >2gb clones
Johannes Schindelin [Fri, 6 Mar 2009 09:48:53 +0000 (10:48 +0100)]
Add an (optional, since expensive) test for >2gb clones

Define GIT_TEST_CLONE_2GB=t if you want the test not to be skipped.

The test works by constructing a repository larger than 2gb, and then
cloning it.

The repository is forced larger than 2gb by setting compression and
delta depth to zero, and then adding just enough unique objects of
a given size.

The objects consist of a running decimal number in ASCII, padded by
spaces.  Should that break in the future, e.g. when pack v4 becomes
default, there is a commented-out call to test-genrandom which can be
substituted, but that uses more cycles than the current method.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agobash completion: teach fetch, pull, and push to complete their options
Jay Soffian [Fri, 6 Mar 2009 04:39:33 +0000 (23:39 -0500)]
bash completion: teach fetch, pull, and push to complete their options

fetch, pull, and push didn't know their options. They do now. merge's
options are factored into a variable so they can be shared between
_git_merge and _git_pull

Signed-off-by: Jay Soffian <jaysoffian@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agobash completion: refactor --strategy completion
Jay Soffian [Fri, 6 Mar 2009 16:30:44 +0000 (11:30 -0500)]
bash completion: refactor --strategy completion

The code to complete --strategy was duplicated between _git_rebase and
_git_merge, and is about to gain a third caller (_git_pull). This patch
factors it into its own function.

Signed-off-by: Jay Soffian <jaysoffian@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agobash completion: fix completion issues with fetch, pull, and push
Jay Soffian [Fri, 6 Mar 2009 04:39:31 +0000 (23:39 -0500)]
bash completion: fix completion issues with fetch, pull, and push

Sverre Rabbelier noticed a completion issue with push:

 $ git push ori<tab>
 git push origin

 $ git push -f ori<tab>
 git push -f origin/

Markus Heidelberg pointed out that the issue extends to fetch and pull.

The reason is that the current code naively assumes that if
COMP_CWORD=2, it should complete a remote name, otherwise it should
complete a refspec. This assumption fails if there are any --options.

This patch fixes that issue by instead scanning COMP_CWORDS to see if
the remote has been completed yet (we now assume the first non-dashed
argument is the remote). The new logic is factored into a function,
shared by fetch, pull, and push.

The new function also properly handles '.' as the remote.

Signed-off-by: Jay Soffian <jaysoffian@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoBrown paper bag fix for MinGW 64-bit stat
Johannes Schindelin [Sat, 7 Mar 2009 14:37:18 +0000 (15:37 +0100)]
Brown paper bag fix for MinGW 64-bit stat

When overriding the identifier "stat" so that "struct stat" will be
substituted with "struct _stati64" everywhere, I tried to fix the calls
to the _function_ stat(), too, but I forgot to change the earlier
attempt "stat64" to "_stati64" there.

So, the stat() calls were overridden by calls to _stati64() instead.

Unfortunately, there is a function _stati64() so that I missed that
calls to stat() were not actually overridden by calls to mingw_lstat(),
but t4200-rerere.sh showed the error.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agobuiltin-revert.c: release index lock when cherry-picking an empty commit
Chris Johnsen [Sat, 7 Mar 2009 09:30:51 +0000 (03:30 -0600)]
builtin-revert.c: release index lock when cherry-picking an empty commit

When a cherry-pick of an empty commit is done, release the lock
held on the index.

The fix is the same as was applied to similar code in 4271666046.

Signed-off-by: Chris Johnsen <chris_johnsen@pobox.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agodocument config --bool-or-int
Jeff King [Sat, 7 Mar 2009 17:14:06 +0000 (12:14 -0500)]
document config --bool-or-int

The documentation is just a pointer to the --bool and --int
options, but it makes sense to at least mention that it
exists.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agot1300: use test_must_fail as appropriate
Jeff King [Sat, 7 Mar 2009 17:14:04 +0000 (12:14 -0500)]
t1300: use test_must_fail as appropriate

Some of the tests checked the exit code manually, even going
so far as to run git outside of the test_expect harness.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocleanup: add isascii()
René Scharfe [Sat, 7 Mar 2009 13:06:49 +0000 (14:06 +0100)]
cleanup: add isascii()

Add a standard definition of isascii() and use it to replace an open
coded high-bit test in pretty.c.  While we're there, write the ESC
char as the more commonly used '\033' instead of as 0x1b to enhance
its grepability.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation: fix badly indented paragraphs in "--bisect-all" description
Christian Couder [Sat, 7 Mar 2009 12:37:24 +0000 (13:37 +0100)]
Documentation: fix badly indented paragraphs in "--bisect-all" description

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoUpdate draft release notes to 1.6.3
Junio C Hamano [Fri, 6 Mar 2009 09:37:22 +0000 (01:37 -0800)]
Update draft release notes to 1.6.3

12 years agoMerge branch 'en/maint-hash-object'
Junio C Hamano [Thu, 5 Mar 2009 23:58:42 +0000 (15:58 -0800)]
Merge branch 'en/maint-hash-object'

* en/maint-hash-object:
  Ensure proper setup of git_dir for git-hash-object

Conflicts:
hash-object.c

12 years agoMerge branch 'jc/blame'
Junio C Hamano [Thu, 5 Mar 2009 23:41:43 +0000 (15:41 -0800)]
Merge branch 'jc/blame'

* jc/blame:
  blame: show "previous" information in --porcelain/--incremental format
  git-blame: refactor code to emit "porcelain format" output

12 years agoMerge branch 'ns/pretty-format'
Junio C Hamano [Thu, 5 Mar 2009 23:41:43 +0000 (15:41 -0800)]
Merge branch 'ns/pretty-format'

* ns/pretty-format:
  bash completion: add --format= and --oneline options for "git log"
  Add tests for git log --pretty, --format and --oneline.
  Add --oneline that is a synonym to "--pretty=oneline --abbrev-commit"
  Give short-hands to --pretty=tformat:%formatstring
  Add --format that is a synonym to --pretty

12 years agoMerge branch 'jk/sane-relative-time'
Junio C Hamano [Thu, 5 Mar 2009 23:41:43 +0000 (15:41 -0800)]
Merge branch 'jk/sane-relative-time'

* jk/sane-relative-time:
  never fallback relative times to absolute

12 years agoMerge branch 'fc/config-editor'
Junio C Hamano [Thu, 5 Mar 2009 23:41:42 +0000 (15:41 -0800)]
Merge branch 'fc/config-editor'

* fc/config-editor:
  git config: trivial cleanup for editor action
  git config: codestyle cleanups
  config: Add new option to open an editor.

12 years agoMerge branch 'js/send-email'
Junio C Hamano [Thu, 5 Mar 2009 23:41:42 +0000 (15:41 -0800)]
Merge branch 'js/send-email'

* js/send-email:
  send-email: add --confirm option and configuration setting
  send-email: don't create temporary compose file until it is needed
  send-email: --suppress-cc improvements
  send-email: handle multiple Cc addresses when reading mbox message
  send-email: allow send-email to run outside a repo

12 years agoMerge branch 'sg/rerere-cleanup'
Junio C Hamano [Thu, 5 Mar 2009 23:41:42 +0000 (15:41 -0800)]
Merge branch 'sg/rerere-cleanup'

* sg/rerere-cleanup:
  rerere: remove duplicated functions

12 years agoMerge branch 'jc/add-p-unquote'
Junio C Hamano [Thu, 5 Mar 2009 23:41:41 +0000 (15:41 -0800)]
Merge branch 'jc/add-p-unquote'

* jc/add-p-unquote:
  git-add -i/-p: learn to unwrap C-quoted paths

12 years agoMerge branch 'jw/imap-preformatted-html'
Junio C Hamano [Thu, 5 Mar 2009 23:41:41 +0000 (15:41 -0800)]
Merge branch 'jw/imap-preformatted-html'

* jw/imap-preformatted-html:
  imap.preformattedHTML to tell Thunderbird to send non-flowed text

12 years agoMerge branch 'jw/format-patch-attach'
Junio C Hamano [Thu, 5 Mar 2009 23:41:41 +0000 (15:41 -0800)]
Merge branch 'jw/format-patch-attach'

* jw/format-patch-attach:
  Enable setting attach as the default in .gitconfig for git-format-patch.

12 years agoMerge branch 'sr/force-rebase'
Junio C Hamano [Thu, 5 Mar 2009 23:41:40 +0000 (15:41 -0800)]
Merge branch 'sr/force-rebase'

* sr/force-rebase:
  Teach rebase to rebase even if upstream is up to date

12 years agoMerge branch 'fg/exclude-bq'
Junio C Hamano [Thu, 5 Mar 2009 23:41:39 +0000 (15:41 -0800)]
Merge branch 'fg/exclude-bq'

* fg/exclude-bq:
  Support "\" in non-wildcard exclusion entries

12 years agoMerge branch 'dm/add-i-edit-abort'
Junio C Hamano [Thu, 5 Mar 2009 23:41:39 +0000 (15:41 -0800)]
Merge branch 'dm/add-i-edit-abort'

* dm/add-i-edit-abort:
  add -i: revisit hunk on editor failure

12 years agoMerge branch 'tp/completion'
Junio C Hamano [Thu, 5 Mar 2009 23:41:39 +0000 (15:41 -0800)]
Merge branch 'tp/completion'

* tp/completion:
  Fixup: Add bare repository indicator for __git_ps1
  Add bare repository indicator for __git_ps1
  completion: More fixes to prevent unbound variable errors
  completion: Better __git_ps1 support when not in working directory
  completion: Use consistent if [...] convention, not "test"
  completion: For consistency, change "git rev-parse" to __gitdir calls

12 years agoMerge branch 'js/branch-symref'
Junio C Hamano [Thu, 5 Mar 2009 23:41:35 +0000 (15:41 -0800)]
Merge branch 'js/branch-symref'

* js/branch-symref:
  add basic branch display tests
  branch: clean up repeated strlen
  Avoid segfault with 'git branch' when the HEAD is detached
  builtin-branch: improve output when displaying remote branches

Conflicts:
builtin-branch.c

12 years agoMerge branch 'al/ansi-color'
Junio C Hamano [Thu, 5 Mar 2009 23:41:19 +0000 (15:41 -0800)]
Merge branch 'al/ansi-color'

* al/ansi-color:
  builtin-branch.c: Rename branch category color names
  Clean up use of ANSI color sequences

12 years agoMerge branch 'js/valgrind'
Junio C Hamano [Thu, 5 Mar 2009 23:41:18 +0000 (15:41 -0800)]
Merge branch 'js/valgrind'

* js/valgrind:
  valgrind: do not require valgrind 3.4.0 or newer
  test-lib: avoid assuming that templates/ are in the GIT_EXEC_PATH
  Tests: let --valgrind imply --verbose and --tee
  Add a script to coalesce the valgrind outputs
  t/Makefile: provide a 'valgrind' target
  test-lib.sh: optionally output to test-results/$TEST.out, too
  Valgrind support: check for more than just programming errors
  valgrind: ignore ldso and more libz errors
  Add valgrind support in test scripts

12 years agoMinGW: 64-bit file offsets
Johannes Schindelin [Thu, 5 Mar 2009 16:05:12 +0000 (17:05 +0100)]
MinGW: 64-bit file offsets

The type 'off_t' should be used everywhere so that the bit-depth of that
type can be adjusted in the standard C library, and you just need to
recompile your program to benefit from the extended precision.

Only that it was not done that way in the MS runtime library.

This patch reroutes off_t to off64_t and provides the other necessary
changes so that finally, clones larger than 2 gigabyte work on Windows
(provided you are on a file system that allows files larger than 2gb).

Initial patch by Sickboy <sb@dev-heaven.net>.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation - More examples for git bisect
John Tapsell [Thu, 5 Mar 2009 12:36:14 +0000 (12:36 +0000)]
Documentation - More examples for git bisect

Including passing parameters to the programs, and running more
complicated checks without requiring a seperate shell script.

Signed-off-by: John Tapsell <johnflux@gmail.com>
Acked-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDraft release notes: Carry forward the warning for behaviour changes
Junio C Hamano [Thu, 5 Mar 2009 22:50:00 +0000 (14:50 -0800)]
Draft release notes: Carry forward the warning for behaviour changes

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'maint'
Junio C Hamano [Thu, 5 Mar 2009 07:43:23 +0000 (23:43 -0800)]
Merge branch 'maint'

* maint:
  Make the 'lock file' exists error more informative

12 years agoimprove missing repository error message
Jeff King [Wed, 4 Mar 2009 08:32:29 +0000 (03:32 -0500)]
improve missing repository error message

Certain remote commands, when asked to do something in a
particular directory that was not actually a git repository,
would say "unable to chdir or not a git archive". The
"chdir" bit is an unnecessary detail, and the term "git
archive" is much less common these days than "git repository".

So let's switch them all to:

  fatal: '%s' does not appear to be a git repository

Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMake the 'lock file' exists error more informative
John Tapsell [Wed, 4 Mar 2009 15:00:44 +0000 (15:00 +0000)]
Make the 'lock file' exists error more informative

It looks like someone did 90% of the work, then forgot to actually use
the function in one place.

Also the helper function did not use the correct variable.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-archive: add --output=<file> to send output to a file
Carlos Manuel Duclos Vergara [Mon, 16 Feb 2009 17:20:25 +0000 (18:20 +0100)]
git-archive: add --output=<file> to send output to a file

When archiving a repository there is no way to specify a file as output.
This patch adds a new option "--output" that redirects the output to a
file instead of stdout.

Signed-off-by: Carlos Manuel Duclos Vergara <carlos.duclos@nokia.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoImprove error message for git-filter-branch
John Tapsell [Thu, 19 Feb 2009 07:36:35 +0000 (07:36 +0000)]
Improve error message for git-filter-branch

Tell the user that a backup (original) already exists, and how to solve
this problem (with -f option)

Signed-off-by: John Tapsell <johnflux@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoGoogle has renamed the imap folder
John Tapsell [Thu, 19 Feb 2009 07:36:11 +0000 (07:36 +0000)]
Google has renamed the imap folder

Also add a comment that the web interface wraps the lines

Signed-off-by: John Tapsell <johnflux@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoModify description file to say what this file is
John Tapsell [Thu, 19 Feb 2009 07:36:00 +0000 (07:36 +0000)]
Modify description file to say what this file is

A lot of people see this message for the first time on the gitweb
interface, where there is no clue as to what 'this file' means.

Signed-off-by: John Tapsell <johnflux@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agorev-list: estimate number of bisection step left
Christian Couder [Sat, 21 Feb 2009 08:26:01 +0000 (09:26 +0100)]
rev-list: estimate number of bisection step left

This patch teaches "git rev-list --bisect-vars" to output an estimate
of the number of bisection step left _after the current one_ along with
the other variables it already outputs.

This patch also makes "git-bisect.sh" display this number of steps left
_after the current one_, along with the estimate of the number of
revisions left to test (after the current one).

Here is a table to help analyse what should be the best estimate for
the number of bisect steps left.

N : linear case                    --> probabilities --> best
-------------------------------------------------------------
1 : G-B                            --> 0             --> 0
2 : G-U1-B                         --> 0             --> 0
3 : G-U1-U2-B                      --> 0(1/3) 1(2/3) --> 1
4 : G-U1-U2-U3-B                   --> 1             --> 1
5 : G-U1-U2-U3-U4-B                --> 1(3/5) 2(2/5) --> 1
6 : G-U1-U2-U3-U4-U5-B             --> 1(2/6) 2(4/6) --> 2
7 : G-U1-U2-U3-U4-U5-U6-B          --> 1(1/7) 2(6/7) --> 2
8 : G-U1-U2-U3-U4-U5-U6-U7-B       --> 2             --> 2
9 : G-U1-U2-U3-U4-U5-U6-U7-U8-B    --> 2(7/9) 3(2/9) --> 2
10: G-U1-U2-U3-U4-U5-U6-U7-U8-U9-B --> 2(6/10)3(4/10)--> 2

In the column "N", there is the number of revisions that could _now_
be the first bad commit we are looking for.

The "linear case" column describes the linear history corresponding to
the number in column N. G means good, B means bad, and Ux means
unknown. Note that the first bad revision we are looking for can be
any Ux or B.

In the "probabilities" column, there are the different outcomes in
number of steps with the odds of each outcome in parenthesis
corresponding to the linear case.

The "best" column gives the most accurate estimate among the different
outcomes in the "probabilities" column.

We have the following:

best(2^n) == n - 1

and for any x between 0 included and 2^n excluded, the probability for
n - 1 steps left looks like:

P(2^n + x) == (2^n - x) / (2^n + x)

and P(2^n + x) < 0.5 means 2^n < 3x

So the algorithm used in this patch calculates 2^n and x, and then
choose between returning n - 1 and n.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoFix neglect of diff_setup()/diff_setup_done() symmetry.
Keith Cascio [Tue, 17 Feb 2009 02:59:00 +0000 (18:59 -0800)]
Fix neglect of diff_setup()/diff_setup_done() symmetry.

Code that calls diff_setup(), including via init_revisions(), should later call
diff_setup_done(), possibly via setup_revisions(). Failure to do so could cause
errors, especially in the future when we add responsibilities to
diff_setup_done(). This instance causes no known errors with the present code.
But it resulted in an error with an experimental patch.

Signed-off-by: Keith Cascio <keith@cs.ucla.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoUse DIFF_XDL_SET/DIFF_OPT_SET instead of raw bit-masking
Keith Cascio [Tue, 17 Feb 2009 03:26:49 +0000 (19:26 -0800)]
Use DIFF_XDL_SET/DIFF_OPT_SET instead of raw bit-masking

Signed-off-by: Keith Cascio <keith@cs.ucla.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoBeginning of 1.6.3 development track
Junio C Hamano [Wed, 4 Mar 2009 08:41:01 +0000 (00:41 -0800)]
Beginning of 1.6.3 development track

12 years agoBeginning of 1.6.2 maintenance track
Junio C Hamano [Wed, 4 Mar 2009 08:37:50 +0000 (00:37 -0800)]
Beginning of 1.6.2 maintenance track

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoGIT 1.6.2 v1.6.2
Junio C Hamano [Wed, 4 Mar 2009 07:02:16 +0000 (23:02 -0800)]
GIT 1.6.2

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation: Typo / spelling / formatting fixes
Mike Ralphson [Tue, 3 Mar 2009 19:29:22 +0000 (19:29 +0000)]
Documentation: Typo / spelling / formatting fixes

Signed-off-by: Mike Ralphson <mike@abacus.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation: Expand a couple of abbreviations
Mike Ralphson [Tue, 3 Mar 2009 19:29:21 +0000 (19:29 +0000)]
Documentation: Expand a couple of abbreviations

These may not be obvious to non-native English speakers

Signed-off-by: Mike Ralphson <mike@abacus.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation: Typos / spelling fixes in RelNotes
Mike Ralphson [Tue, 3 Mar 2009 19:29:20 +0000 (19:29 +0000)]
Documentation: Typos / spelling fixes in RelNotes

Signed-off-by: Mike Ralphson <mike@abacus.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation/git-archive.txt: Note attributes
Roy Lee [Tue, 3 Mar 2009 16:49:29 +0000 (00:49 +0800)]
Documentation/git-archive.txt: Note attributes

Signed-off-by: Roy Lee <roylee17@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agosend-email: add --confirm option and configuration setting
Jay Soffian [Tue, 3 Mar 2009 04:52:18 +0000 (23:52 -0500)]
send-email: add --confirm option and configuration setting

send-email violates the principle of least surprise by automatically
cc'ing additional recipients without confirming this with the user.

This patch teaches send-email a --confirm option. It takes the
following values:

 --confirm=always   always confirm before sending
 --confirm=never    never confirm before sending
 --confirm=cc       confirm before sending when send-email has
                    automatically added addresses from the patch to
                    the Cc list
 --confirm=compose  confirm before sending the first message when
                    using --compose. (Needed to maintain backwards
                    compatibility with existing behavior.)
 --confirm=auto     'cc' + 'compose'

If sendemail.confirm is unconfigured, the option defaults to 'compose'
if any suppress-Cc related options have been used, otherwise it defaults
to 'auto'.

Unfortunately, it is impossible to introduce this patch such that it
helps new users without potentially annoying some existing users. We
attempt to mitigate the latter by:

 * Allowing the user to set 'git config sendemail.confirm never'
 * Allowing the user to say 'all' after the first prompt to not be
   prompted on remaining emails during the same invocation.
 * Telling the user about the 'sendemail.confirm' setting if it is
   unconfigured whenever we prompt due to Cc before sending.
 * Only prompting if no --suppress related options have been passed, as
   using such an option is likely to indicate an experienced send-email
   user.

There is a slight fib in message informing the user of the
sendemail.confirm setting and this is intentional. Setting 'auto'
differs from leaving sendemail.confirm unset in two ways: 1) 'auto'
obviously squelches the informational message; 2) 'auto' prompts when
the Cc list has been expanded even in the presence of a --suppress
related option, where leaving sendemail.confirm unset does not. This is
intentional to keep the message simple, and to avoid adding another
sendemail.confirm value ('auto-except-suppress'?).

Signed-off-by: Jay Soffian <jaysoffian@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation: minor grammatical fixes.
David J. Mellor [Mon, 2 Mar 2009 06:42:44 +0000 (22:42 -0800)]
Documentation: minor grammatical fixes.

The final hunk in this patch corrects what appears to be a typo:

of --> or

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation: minor grammatical fixes.
David J. Mellor [Mon, 2 Mar 2009 06:38:36 +0000 (22:38 -0800)]
Documentation: minor grammatical fixes.

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation: minor grammatical fixes.
David J. Mellor [Mon, 2 Mar 2009 06:37:41 +0000 (22:37 -0800)]
Documentation: minor grammatical fixes.

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agot3400-rebase: Move detached HEAD check earlier
Johannes Sixt [Sun, 1 Mar 2009 10:20:03 +0000 (11:20 +0100)]
t3400-rebase: Move detached HEAD check earlier

Short story: There is a section in t3400 that tests fundamental rebase
properties.  3ec7371f (Add two extra tests for git rebase, 2009-02-09)
added a check that rebase works on a detached HEAD, but the test was put
near the end of the file.  This moves it to a more suitable place.

Long story: The test that preceded the one in question tests that a
rebased commit degrades from a content change with mode change to a
mere mode change.  But on Windows, where we have core.filemode=false,
the original commit did not record the mode change, and so the rebase
operation did not rebase anything.  This caused the subsequent detached
HEAD test to fail.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge git://git.kernel.org/pub/scm/gitk/gitk
Junio C Hamano [Mon, 2 Mar 2009 06:20:52 +0000 (22:20 -0800)]
Merge git://git.kernel.org/pub/scm/gitk/gitk

* git://git.kernel.org/pub/scm/gitk/gitk:
  gitk: Fix possible infinite loop and display corruption

12 years agogitk: Fix possible infinite loop and display corruption
Paul Mackerras [Sun, 1 Mar 2009 22:38:17 +0000 (09:38 +1100)]
gitk: Fix possible infinite loop and display corruption

This fixes an issue reported by Johannes Sixt on the git mailing list:

> This recipe sends gitk into an endless loop. In git.git do:
>
> cd t
> # remove chmod a+x A near the end of the file
> sed -i 's/chmod/: chmod/' t3400-rebase.sh
> sh t3400-rebase.sh --debug
> cd trash\ directory.t3400-rebase/
> gitk master modechange modechange@{1}
>
>
> I briefly see the history chart, but the dot that should be modechange@{1}
> is missing. One automatically selected commit is shown in the diff section
> below. But then the commit list is cleared and gitk goes into an infinite
> loop.
>
> Things work alright if either modechange@{1} is dropped, or the 'chmod'
> line is left unchanged, which is a bit strange.
>
> This is with git version 1.6.1.2.390.gba743

There were actually two problems.  This recipe created a situation where
git log would output a child commit after its parent.  This meant that
we called fix_reversal which called splitvarc, which should call modify_arc
to note the fact that it has modified the arc that it has just split.  It
wasn't, which meant that displayorder and other variables got into an
inconsistent state (a commit appearing twice in displayorder).

This then meant that the targetrow/targetid logic in drawvisible thought
it need to redraw each time.  That, together with the fact that drawvisible
called drawcommits which called drawvisible if a redraw was needed, led
to the infinite loop.

In fact drawvisible is now the only caller of drawcommits.  Thus, the
start and end row arguments to drawcommits always encompass the whole
visible area, so drawcommits doesn't need to call drawvisible to redraw;
it just needs to clear the screen and draw what it's been asked to.

This fixes these two problems by adding a call to modify_arc in
splitvarc and by taking out the call to drawvisible in drawcommits.
It also removes an unrelated left-over debugging puts in external_blame.

Signed-off-by: Paul Mackerras <paulus@samba.org>
12 years agoMerge branch 'jc/maint-1.6.0-pack-directory'
Junio C Hamano [Sun, 1 Mar 2009 00:31:02 +0000 (16:31 -0800)]
Merge branch 'jc/maint-1.6.0-pack-directory'

* jc/maint-1.6.0-pack-directory:
  Fix odb_mkstemp() on AIX

12 years agoMerge branch 'maint'
Junio C Hamano [Sat, 28 Feb 2009 22:39:56 +0000 (14:39 -0800)]
Merge branch 'maint'

* maint:
  Documentation: minor grammatical fixes.
  added missing backtick in git-apply.txt

12 years agoDocumentation: minor grammatical fixes.
David J. Mellor [Sat, 28 Feb 2009 21:12:59 +0000 (13:12 -0800)]
Documentation: minor grammatical fixes.

Signed-off-by: David J. Mellor <dmellor@whistlingcat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoEnsure proper setup of git_dir for git-hash-object
Elijah Newren [Sat, 28 Feb 2009 19:56:49 +0000 (12:56 -0700)]
Ensure proper setup of git_dir for git-hash-object

Call setup_git_directory() before git_config() to make sure git_dir is set
to the proper value.

Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'maint-1.6.0' into maint
Junio C Hamano [Sat, 28 Feb 2009 22:05:09 +0000 (14:05 -0800)]
Merge branch 'maint-1.6.0' into maint

* maint-1.6.0:
  added missing backtick in git-apply.txt

12 years agoadded missing backtick in git-apply.txt
Danijel Tasov [Sat, 28 Feb 2009 20:03:54 +0000 (21:03 +0100)]
added missing backtick in git-apply.txt

Signed-off-by: Danijel Tasov <dt@korn.shell.la>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-rebase: Update --whitespace documentation
Todd Zullinger [Sat, 28 Feb 2009 18:42:02 +0000 (13:42 -0500)]
git-rebase: Update --whitespace documentation

The parameters accepted by the --whitespace option of "git apply" have
changed over time, and the documentation for "git rebase" was out of
sync.  Remove the specific parameter list from the "git rebase"
documentation and simply point to the "git apply" documentation for
details, as is already done in the "git am" documentation.

Signed-off-by: Todd Zullinger <tmz@pobox.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-svn - return original format_svn_date semantics
Ben Walton [Sat, 28 Feb 2009 03:11:45 +0000 (22:11 -0500)]
git-svn - return original format_svn_date semantics

When %z was removed from the strftime call and subsituted with a
local gmt offset calculation, time() was no longer the default for
all time functions as it was with the previous localtime(shift).
This is now corrected so that format_svn_time behaves as it used to.

Signed-off-by: Ben Walton <bwalton@artsci.utoronto.ca>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-svn: disable broken symlink workaround by default
Eric Wong [Sat, 28 Feb 2009 03:40:16 +0000 (19:40 -0800)]
git-svn: disable broken symlink workaround by default

Even though this will break things for some extremely rare repositories
used by broken Windows clients, it's probably not worth enabling this by
default as it has negatively affected many more users than it has helped
from what we've seen so far.

The extremely rare repositories that have broken symlinks in them will be
silently corrupted in import; but users can still reenable this option and
restart the import.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'cc/maint-1.6.0-bisect-fix'
Junio C Hamano [Sat, 28 Feb 2009 00:00:33 +0000 (16:00 -0800)]
Merge branch 'cc/maint-1.6.0-bisect-fix'

* cc/maint-1.6.0-bisect-fix:
  bisect: fix another instance of eval'ed string

Conflicts:
git-bisect.sh

12 years agobisect: fix another instance of eval'ed string
Christian Couder [Fri, 27 Feb 2009 06:31:22 +0000 (07:31 +0100)]
bisect: fix another instance of eval'ed string

When there is nothing to be skipped, the output from
rev-list --bisect-vars was eval'ed without first being
strung together with &&; this is probably not a problem
as it is much less likely to be a bad input than the list
handcrafted by the filter_skip function, but it still is
a good discipline.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agobash completion: add --format= and --oneline options for "git log"
Teemu Likonen [Tue, 24 Feb 2009 13:33:29 +0000 (15:33 +0200)]
bash completion: add --format= and --oneline options for "git log"

We also add --format= completion for "git show".

Signed-off-by: Teemu Likonen <tlikonen@iki.fi>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'cc/maint-1.6.0-bisect-fix'
Junio C Hamano [Fri, 27 Feb 2009 09:03:21 +0000 (01:03 -0800)]
Merge branch 'cc/maint-1.6.0-bisect-fix'

* cc/maint-1.6.0-bisect-fix:
  bisect: fix quoting TRIED revs when "bad" commit is also "skip"ped

Conflicts:
git-bisect.sh

12 years agobisect: fix quoting TRIED revs when "bad" commit is also "skip"ped
Christian Couder [Fri, 27 Feb 2009 06:31:22 +0000 (07:31 +0100)]
bisect: fix quoting TRIED revs when "bad" commit is also "skip"ped

When the "bad" commit was also "skip"ped and when more than one
commit was skipped, the "filter_skipped" function would have
printed something like:

    bisect_rev=<hash1>|<hash2>

(where <hash1> and <hash2> are hexadecimal sha1 hashes)

and this would have been evaled later as piping "bisect_rev=<hash1>"
into "<hash2>", which would have failed.

So this patch makes the "filter_skipped" function properly quote
what it outputs, so that it will print something like:

bisect_rev='<hash1>|<hash2>'

which will be properly evaled later.  The caller was not stopping
properly because the scriptlet this function returned to be evaled
was not strung together with && and because of this, an error in
an earlier part of the output was simply ignored.

A test case is added to the test suite.

And while at it, we also initialize the VARS, FOUND and TRIED
variables, so that we protect ourselves from environment variables
the user may have with these names.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-am: make --abort less dangerous
Junio C Hamano [Thu, 26 Feb 2009 19:24:29 +0000 (11:24 -0800)]
git-am: make --abort less dangerous

When you are in the middle of "git rebase", "git am --abort" by mistake
would have referred to nonexistent ORIG_HEAD and barfed, or worse yet, used
a stale ORIG_HEAD and taken you to an unexpected commit.

Also the option parsing did not reject "git am --abort --skip".

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-am: Keep index in case of abort with dirty index
Michael J Gruber [Thu, 26 Feb 2009 09:52:53 +0000 (10:52 +0100)]
git-am: Keep index in case of abort with dirty index

git am --abort resets the index unconditionally. But in case a previous
git am exited due to a dirty index it is preferable to keep that index.
Make it so.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agot5540-http-push.sh: avoid non-portable grep -P
Jay Soffian [Thu, 26 Feb 2009 23:44:40 +0000 (18:44 -0500)]
t5540-http-push.sh: avoid non-portable grep -P

OS X's GNU grep does not support -P/--perl-regexp.

We use a basic RE instead, and simplify the pattern slightly by
replacing '+' with '*' so it can be more easily expressed using a basic
RE.  The important part of pattern, checking for a SHA-1 has suffix in
the successful PUT/MOVE operations, remains the same.  Also, a-z instead
of a-f was an obvious mistake in the original RE. Here are samples of
what we want to match:

127.0.0.1 - - [26/Feb/2009:22:38:13 +0000] "PUT /test_repo.git/objects/3e/a4fbb9e18a401a6463c595d08118fcb9fb7426_fab55116904c665a95438bcc78521444a7db6096 HTTP/1.1" 201 277
127.0.0.1 - - [26/Feb/2009:22:38:13 +0000] "MOVE /test_repo.git/objects/3e/a4fbb9e18a401a6463c595d08118fcb9fb7426_fab55116904c665a95438bcc78521444a7db6096 HTTP/1.1" 201 277

Signed-off-by: Jay Soffian <jaysoffian@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoFix odb_mkstemp() on AIX
Mike Ralphson [Thu, 26 Feb 2009 15:31:52 +0000 (16:31 +0100)]
Fix odb_mkstemp() on AIX

The AIX mkstemp() modifies its template parameter to an empty string if
the call fails.  The existing code had already recomputed the template,
but too late to be good.

See also 6ff6af62, which fixed this problem in a different spot.

Signed-off-by: Mike Ralphson <mike@abacus.co.uk>
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoGIT 1.6.2-rc2 v1.6.2-rc2
Junio C Hamano [Wed, 25 Feb 2009 22:50:52 +0000 (14:50 -0800)]
GIT 1.6.2-rc2

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'jc/maint-1.6.0-pack-directory'
Junio C Hamano [Wed, 25 Feb 2009 22:48:30 +0000 (14:48 -0800)]
Merge branch 'jc/maint-1.6.0-pack-directory'

* jc/maint-1.6.0-pack-directory:
  Make sure objects/pack exists before creating a new pack

12 years agoMake sure objects/pack exists before creating a new pack
Junio C Hamano [Wed, 25 Feb 2009 07:11:29 +0000 (23:11 -0800)]
Make sure objects/pack exists before creating a new pack

In a repository created with git older than f49fb35 (git-init-db: create
"pack" subdirectory under objects, 2005-06-27), objects/pack/ directory is
not created upon initialization.  It was Ok because subdirectories are
created as needed inside directories init-db creates, and back then,
packfiles were recent invention.

After the said commit, new codepaths started relying on the presense of
objects/pack/ directory in the repository.  This was exacerbated with
8b4eb6b (Do not perform cross-directory renames when creating packs,
2008-09-22) that moved the location temporary pack files are created from
objects/ directory to objects/pack/ directory, because moving temporary to
the final location was done carefully with lazy leading directory creation.

Many packfile related operations in such an old repository can fail
mysteriously because of this.

This commit introduces two helper functions to make things work better.

 - odb_mkstemp() is a specialized version of mkstemp() to refactor the
   code and teach it to create leading directories as needed;

 - odb_pack_keep() refactors the code to create a ".keep" file while
   create leading directories as needed.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoAllow HTTP tests to run on Darwin
Jay Soffian [Wed, 25 Feb 2009 08:28:15 +0000 (03:28 -0500)]
Allow HTTP tests to run on Darwin

This patch allows the HTTP tests to run on OS X 10.5. It is not
sufficient to be able to pass in LIB_HTTPD_PATH and
LIB_HTTPD_MODULE_PATH alone, as the apache.conf also needs a couple
tweaks.

These changes are put into an <IfDefine> to keep them Darwin specific,
but this means lib-httpd.sh needs to be modified to pass -DDarwin to
apache when running on Darwin. As long as we're making this change to
lib-httpd.sh, we may as well set LIB_HTTPD_PATH and
LIB_HTTPD_MODULE_PATH to appropriate default values for the platform.

Note that we now pass HTTPD_PARA to apache at shutdown as well.
Otherwise apache will emit a harmless, but noisy warning that LogFormat
is an unknown directive.

Signed-off-by: Jay Soffian <jaysoffian@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoFix typo in contrib/examples/git-svnimport.txt
Michael J Gruber [Wed, 25 Feb 2009 14:05:17 +0000 (15:05 +0100)]
Fix typo in contrib/examples/git-svnimport.txt

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogitattributes.txt: Path matching rules are explained in gitignore.txt
Johannes Sixt [Wed, 25 Feb 2009 07:54:42 +0000 (08:54 +0100)]
gitattributes.txt: Path matching rules are explained in gitignore.txt

The rules how the patterns are matched against path names are the same
for .gitattribute and .gitignore files.

This also replace the notion "glob pattern" by "pattern" because
gitignore.txt talks about "glob" only in some contexts where the pattern
is mentioned.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agosha1_file.c: fix typo
Felipe Contreras [Tue, 24 Feb 2009 21:59:05 +0000 (23:59 +0200)]
sha1_file.c: fix typo

it's != its

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit add: trivial codestyle cleanup
Felipe Contreras [Tue, 24 Feb 2009 21:59:03 +0000 (23:59 +0200)]
git add: trivial codestyle cleanup

Global static variables don't need to be initialized to 0/NULL.

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoREADME: fix path to "gitcvs-migration.txt" and be more consistent
Christian Couder [Tue, 24 Feb 2009 20:16:37 +0000 (21:16 +0100)]
README: fix path to "gitcvs-migration.txt" and be more consistent

README suggested to look at "Documentation/gittutorial.txt" for the
tutorial and to use "man git-commandname" for documentation of each
command.

This was not consistent because the tutorial can also be available with
"man gittutorial" once git is installed, and the documentation for each
command can be available at "Documentation/git-commandname.txt" before
installing git.

This patch tries to make the description more consistent.  It also fixes
the path to the cvs-migration documentation that changed from
"Documentation/cvs-migration.txt" to "Documentation/gitcvs-migration.txt".

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agotrace: Fixed a minor typo in an error message.
Allan Caffee [Wed, 25 Feb 2009 01:22:20 +0000 (20:22 -0500)]
trace: Fixed a minor typo in an error message.

Signed-off-by: Allan Caffee <allan.caffee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>