git.git
11 years agorebase --preserve-merges: keep all merge commits including empty ones
Phil Hord [Sat, 12 Jan 2013 20:46:01 +0000 (15:46 -0500)]
rebase --preserve-merges: keep all merge commits including empty ones

Since 90e1818f9a  (git-rebase: add keep_empty flag, 2012-04-20)
'git rebase --preserve-merges' fails to preserve empty merge commits
unless --keep-empty is also specified.  Merge commits should be
preserved in order to preserve the structure of the rebased graph,
even if the merge commit does not introduce changes to the parent.

Teach rebase not to drop merge commits only because they are empty.

A special case which is not handled by this change is for a merge commit
whose parents are now the same commit because all the previous different
parents have been dropped as a result of this rebase or some previous
operation.

Signed-off-by: Phil Hord <hordp@cisco.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'nh/empty-rebase'
Junio C Hamano [Mon, 30 Apr 2012 21:58:00 +0000 (14:58 -0700)]
Merge branch 'nh/empty-rebase'

"git rebase" learned to optionally keep commits that do not introduce
any change in the original history.

By Neil Horman
* nh/empty-rebase:
  git-rebase: add keep_empty flag
  git-cherry-pick: Add test to validate new options
  git-cherry-pick: Add keep-redundant-commits option
  git-cherry-pick: add allow-empty option

12 years agoMerge branch 'jk/maint-config-bogus-section'
Junio C Hamano [Mon, 30 Apr 2012 21:46:46 +0000 (14:46 -0700)]
Merge branch 'jk/maint-config-bogus-section'

"git config --rename-section" to rename an existing section into a
bogus one did not check the new name.

By Jeff King
* jk/maint-config-bogus-section:
  config: reject bogus section names for --rename-section

12 years agoThe seventh batch of topics graduated to 'master'
Junio C Hamano [Mon, 30 Apr 2012 01:00:47 +0000 (18:00 -0700)]
The seventh batch of topics graduated to 'master'

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'js/daemon-test-race-fix'
Junio C Hamano [Mon, 30 Apr 2012 00:52:02 +0000 (17:52 -0700)]
Merge branch 'js/daemon-test-race-fix'

The shell construct to launch git-daemon and wait for it to start
serving during the test was faulty, and this fixes it.

By Johannes Sixt
* js/daemon-test-race-fix:
  t5570: fix forwarding of git-daemon messages via cat

12 years agoMerge branch 'kk/gitweb-omit-expensive'
Junio C Hamano [Mon, 30 Apr 2012 00:52:00 +0000 (17:52 -0700)]
Merge branch 'kk/gitweb-omit-expensive'

"gitweb" learned to optionally omit output of fields that are expensive
to generate.

By Kacper Kornet
* kk/gitweb-omit-expensive:
  gitweb: Option to not display information about owner
  gitweb: Option to omit column with time of the last change

12 years agoMerge branch 'kk/maint-gitweb-missing-owner'
Junio C Hamano [Mon, 30 Apr 2012 00:51:56 +0000 (17:51 -0700)]
Merge branch 'kk/maint-gitweb-missing-owner'

By Kacper Kornet
* kk/maint-gitweb-missing-owner:
  gitweb: Don't set owner if got empty value from projects.list

12 years agoMerge branch 'rs/commit-list-append'
Junio C Hamano [Mon, 30 Apr 2012 00:51:30 +0000 (17:51 -0700)]
Merge branch 'rs/commit-list-append'

There is no need for "commit_list_reverse()" function that only invites
inefficient code.

By René Scharfe
* rs/commit-list-append:
  commit: remove commit_list_reverse()
  revision: append to list instead of insert and reverse
  sequencer: export commit_list_append()

12 years agoMerge branch 'mm/include-userpath'
Junio C Hamano [Mon, 30 Apr 2012 00:51:27 +0000 (17:51 -0700)]
Merge branch 'mm/include-userpath'

The new "include.path" directive in the configuration files learned
to understand "~/path" and "~user/path".

By Jeff King
* mm/include-userpath:
  config: expand tildes in include.path variable

12 years agoMerge branch 'pw/t5800-import-race-fix'
Junio C Hamano [Mon, 30 Apr 2012 00:51:23 +0000 (17:51 -0700)]
Merge branch 'pw/t5800-import-race-fix'

Longstanding bug in a test scaffolding that occasionally made t5800
hang was fixed.

By Pete Wyckoff
* pw/t5800-import-race-fix:
  git-remote-testgit: fix race when spawning fast-import

12 years agoMerge branch 'jk/repack-no-explode-objects-from-old-pack'
Junio C Hamano [Mon, 30 Apr 2012 00:50:27 +0000 (17:50 -0700)]
Merge branch 'jk/repack-no-explode-objects-from-old-pack'

Avoid writing out unreachable objects as loose objects when repacking,
if such loose objects will immediately pruned due to its age anyway.

By Jeff King
* jk/repack-no-explode-objects-from-old-pack:
  gc: use argv-array for sub-commands
  argv-array: add a new "pushl" method
  argv-array: refactor empty_argv initialization
  gc: do not explode objects which will be immediately pruned

12 years agoconfig: expand tildes in include.path variable
Jeff King [Wed, 25 Apr 2012 12:00:36 +0000 (08:00 -0400)]
config: expand tildes in include.path variable

You can already use relative paths in include.path, which
means that including "foo" from your global "~/.gitconfig"
will look in your home directory. However, you might want to
do something clever like putting "~/.gitconfig-foo" in a
specific repository's config file.

Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge git://git.bogomips.org/git-svn
Junio C Hamano [Mon, 30 Apr 2012 00:00:42 +0000 (17:00 -0700)]
Merge git://git.bogomips.org/git-svn

By Matthijs Kooijman
via Eric Wong
* git://git.bogomips.org/git-svn:
  git-svn: use platform specific auth providers

12 years agoThe sixth batch of topics graduated to 'master'
Junio C Hamano [Fri, 27 Apr 2012 21:12:56 +0000 (14:12 -0700)]
The sixth batch of topics graduated to 'master'

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'jc/merge-reduce-parents-early'
Junio C Hamano [Fri, 27 Apr 2012 20:59:20 +0000 (13:59 -0700)]
Merge branch 'jc/merge-reduce-parents-early'

Octopus merge strategy did not reduce heads that are recorded in the
final commit correctly.

By Junio C Hamano (4) and Michał Kiedrowicz (1)
* jc/merge-reduce-parents-early:
  fmt-merge-msg: discard needless merge parents
  builtin/merge.c: reduce parents early
  builtin/merge.c: collect other parents early
  builtin/merge.c: remove "remoteheads" global variable
  merge tests: octopus with redundant parents

12 years agoMerge branch 'js/fast-import-test-9300'
Junio C Hamano [Fri, 27 Apr 2012 20:59:04 +0000 (13:59 -0700)]
Merge branch 'js/fast-import-test-9300'

By Johannes Sixt
* js/fast-import-test-9300:
  t9300-fast-import: avoid 'exit' in test_expect_success snippets

12 years agoMerge branch 'rt/cherry-revert-conflict-summary'
Junio C Hamano [Fri, 27 Apr 2012 20:58:53 +0000 (13:58 -0700)]
Merge branch 'rt/cherry-revert-conflict-summary'

In the older days, the header "Conflicts:" in "cherry-pick" and
"merge" was separated by a blank line from the list of paths that
follow for readability, but when "merge" was rewritten in C, we lost
it by mistake. Remove the newline from "cherry-pick" to make them
match again.

By Ralf Thielow
* rt/cherry-revert-conflict-summary:
  sequencer: remove additional blank line

12 years agoMerge branch 'cb/maint-report-mount-point-correctly-in-setup'
Junio C Hamano [Fri, 27 Apr 2012 20:58:43 +0000 (13:58 -0700)]
Merge branch 'cb/maint-report-mount-point-correctly-in-setup'

The filesystem boundary was not correctly reported when .git directory
discovery stopped at a mount point.

By Clemens Buchacher
* cb/maint-report-mount-point-correctly-in-setup:
  properly keep track of current working directory

12 years agoMerge branch 'cb/cherry-pick-rev-path-confusion'
Junio C Hamano [Fri, 27 Apr 2012 20:58:02 +0000 (13:58 -0700)]
Merge branch 'cb/cherry-pick-rev-path-confusion'

The command line parser choked "git cherry-pick $name" when $name can be
both revision name and a pathname, even though $name can never be a path
in the context of the command.

The issue the patch addresses is real, but the way it is implemented felt
unnecessarily invasive a bit.  It may be cleaner for this caller to add
the "--" to the end of the argv_array it passes to setup_revisions().

By Clemens Buchacher
* cb/cherry-pick-rev-path-confusion:
  cherry-pick: do not expect file arguments

12 years agoMerge branch 'cb/http-multi-curl-auth'
Junio C Hamano [Fri, 27 Apr 2012 20:57:49 +0000 (13:57 -0700)]
Merge branch 'cb/http-multi-curl-auth'

Fixes http authentication breakage when we keep multiple HTTP requests in
flight using curl-multi.

By Jeff King (3) and Clemens Buchacher (1)
* cb/http-multi-curl-auth:
  http: use newer curl options for setting credentials
  http: clean up leak in init_curl_http_auth
  fix http auth with multiple curl handles
  http auth fails with multiple curl handles

12 years agot5570: fix forwarding of git-daemon messages via cat
Johannes Sixt [Thu, 26 Apr 2012 21:00:39 +0000 (23:00 +0200)]
t5570: fix forwarding of git-daemon messages via cat

The shell function that starts git-daemon wants to read the first line of
the daemon's stderr to ensure that it started correctly. Subsequent daemon
errors should be redirected to fd 4 (which is the terminal in verbose mode
or /dev/null in quiet mode). To that end the shell script used 'read' to
get the first line of output, and then 'cat &' to forward everything else
in a background process.

The problem is, that 'cat >&4 &' does not produce any output because the
shell redirects a background process's stdin to /dev/null. To have this
command invocation do anything useful, we have to redirect its stdin
explicitly (which overrides the /dev/null redirection).

The shell function connects the daemon's stderr to its consumers via a
FIFO. We cannot just do this:

   read line <git_daemon_output
   cat <git_daemon_output >&4 &

because after the first redirection the pipe is closed and the daemon
could receive SIGPIPE if it writes at the wrong moment. Therefore, we open
the readable end of the FIFO only once on fd 7 in the shell and dup from
there to the stdin of the two consumers.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-svn: use platform specific auth providers
Matthijs Kooijman [Thu, 26 Apr 2012 19:34:02 +0000 (21:34 +0200)]
git-svn: use platform specific auth providers

On Linux, this makes authentication using passwords from gnome-keyring
and kwallet work (only the former was tested). On Mac OS X, this allows
using the OS X Keychain.

Signed-off-by: Matthijs Kooijman <matthijs@stdin.nl>
Acked-by: Eric Wong <normalperson@yhbt.net>
12 years agogitweb: Option to not display information about owner
Kacper Kornet [Thu, 26 Apr 2012 16:45:44 +0000 (18:45 +0200)]
gitweb: Option to not display information about owner

In some setups the repository owner is not a well defined concept
and administrator can prefer it to be not shown. This commit add
and an option that enable to reach this effect.

Signed-off-by: Kacper Kornet <draenog@pld-linux.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoUpdate draft release notes to 1.7.11
Junio C Hamano [Thu, 26 Apr 2012 18:02:36 +0000 (11:02 -0700)]
Update draft release notes to 1.7.11

A handful of topics have been merged to maintenance releases, and
the first half of 6th batch graduates to 'master'.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'rl/maint-stash-i18n-save-error'
Junio C Hamano [Thu, 26 Apr 2012 17:59:08 +0000 (10:59 -0700)]
Merge branch 'rl/maint-stash-i18n-save-error'

The i18n of error message "git stash save" was not properly done.

By Ross Lagerwall
* rl/maint-stash-i18n-save-error:
  stash: use eval_gettextln correctly

12 years agoMerge branch 'mb/fetch-call-a-non-branch-a-ref'
Junio C Hamano [Thu, 26 Apr 2012 17:57:43 +0000 (10:57 -0700)]
Merge branch 'mb/fetch-call-a-non-branch-a-ref'

The report from "git fetch" said "new branch" even for a non branch
ref.

By Marc Branchaud
* mb/fetch-call-a-non-branch-a-ref:
  fetch: describe new refs based on where it came from
  fetch: Give remote_ref to update_local_ref() as well

12 years agoMerge branch 'jm/maint-strncpy-diff-no-index'
Junio C Hamano [Thu, 26 Apr 2012 17:56:19 +0000 (10:56 -0700)]
Merge branch 'jm/maint-strncpy-diff-no-index'

The "diff --no-index" codepath used limited-length buffers,
risking pathnames getting truncated.  Update it to use the
strbuf API.

By Jim Meyering (1) and Junio C Hamano (1)
* jm/maint-strncpy-diff-no-index:
  diff --no-index: use strbuf for temporary pathnames
  diff: avoid stack-buffer-read-overrun for very long name

12 years agoMerge bundle error message fix in
Junio C Hamano [Thu, 26 Apr 2012 17:54:58 +0000 (10:54 -0700)]
Merge bundle error message fix in

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoStart preparing for 1.7.10.1
Junio C Hamano [Thu, 26 Apr 2012 17:40:08 +0000 (10:40 -0700)]
Start preparing for 1.7.10.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'jk/run-command-eacces' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:51:40 +0000 (10:51 -0700)]
Merge branch 'jk/run-command-eacces' into maint

When PATH contains an unreadable directory, alias expansion code did
not kick in, and failed with an error that said "git-subcmd" was not
found.

By Jeff King (1) and Ramsay Jones (1)
* jk/run-command-eacces:
  run-command: treat inaccessible directories as ENOENT
  compat/mingw.[ch]: Change return type of exec functions to int

12 years agoMerge branch 'jc/push-upstream-sanity' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:51:18 +0000 (10:51 -0700)]
Merge branch 'jc/push-upstream-sanity' into maint

The 'push to upstream' implementation was broken in some corner
cases. "git push $there" without refspec, when the current branch is
set to push to a remote different from $there, used to push to $there
using the upstream information to a remote unreleated to $there.

* jc/push-upstream-sanity:
  push: error out when the "upstream" semantics does not make sense

12 years agoMerge branch 'jc/maint-clean-nested-worktree-in-subdir' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:35:42 +0000 (10:35 -0700)]
Merge branch 'jc/maint-clean-nested-worktree-in-subdir' into maint

"git clean -d -f" (not "-d -f -f") is supposed to protect nested
working trees of independent git repositories that exist in the
current project working tree from getting removed, but the protection
applied only to such working trees that are at the top-level of the
current project by mistake.

* jc/maint-clean-nested-worktree-in-subdir:
  clean: preserve nested git worktree in subdirectories

12 years agoMerge branch 'jk/diff-no-rename-empty' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:35:33 +0000 (10:35 -0700)]
Merge branch 'jk/diff-no-rename-empty' into maint

Rename detection logic used to match two empty files as renames during
merge-recursive, leading unnatural mismerges.

By Jeff King
* jk/diff-no-rename-empty:
  merge-recursive: don't detect renames of empty files
  teach diffcore-rename to optionally ignore empty content
  make is_empty_blob_sha1 available everywhere
  drop casts from users EMPTY_TREE_SHA1_BIN

12 years agoMerge branch 'rs/combine-diff-zero-context-at-the-beginning' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:35:25 +0000 (10:35 -0700)]
Merge branch 'rs/combine-diff-zero-context-at-the-beginning' into maint

An age-old corner case bug in combine diff (only triggered with -U0
and the hunk at the beginning of the file needs to be shown) has been
fixed.

By René Scharfe
* rs/combine-diff-zero-context-at-the-beginning:
  combine-diff: fix loop index underflow

12 years agoMerge branch 'jc/commit-unedited-template' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:35:15 +0000 (10:35 -0700)]
Merge branch 'jc/commit-unedited-template' into maint

When "git commit --template F" errors out because the user did not
touch the message, it claimed that it aborts due to "empty message",
which was utterly wrong.

By Junio C Hamano (4) and Adam Monsen (1)
* jc/commit-unedited-template:
  Documentation/git-commit: rephrase the "initial-ness" of templates
  git-commit.txt: clarify -t requires editing message
  commit: rephrase the error when user did not touch templated log message
  commit: do not trigger bogus "has templated message edited" check
  t7501: test the right kind of breakage

12 years agoMerge branch 'jk/add-p-skip-conflicts' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:35:07 +0000 (10:35 -0700)]
Merge branch 'jk/add-p-skip-conflicts' into maint

"git add -p" is not designed to deal with unmerged paths but did
not exclude them and tried to apply funny patches only to fail.

By Jeff King
* jk/add-p-skip-conflicts:
  add--interactive: ignore unmerged entries in patch mode

12 years agoMerge branch 'jc/commit-hook-authorship' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:34:53 +0000 (10:34 -0700)]
Merge branch 'jc/commit-hook-authorship' into maint

"git commit --author=$name" did not tell the name that was being
recorded in the resulting commit to hooks, even though it does do so
when the end user overrode the authorship via the "GIT_AUTHOR_NAME"
environment variable.

* jc/commit-hook-authorship:
  commit: pass author/committer info to hooks
  t7503: does pre-commit-hook learn authorship?
  ident.c: add split_ident_line() to parse formatted ident line

12 years agoMerge branch 'tr/maint-word-diff-regex-sticky' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:34:43 +0000 (10:34 -0700)]
Merge branch 'tr/maint-word-diff-regex-sticky' into maint

The regexp configured with diff.wordregex was incorrectly reused
across files.

By Thomas Rast (2) and Johannes Sixt (1)
* tr/maint-word-diff-regex-sticky:
  diff: tweak a _copy_ of diff_options with word-diff
  diff: refactor the word-diff setup from builtin_diff_cmd
  t4034: diff.*.wordregex should not be "sticky" in --word-diff

12 years agoMerge branch 'jh/notes-merge-in-git-dir-worktree' into maint
Junio C Hamano [Thu, 26 Apr 2012 17:32:55 +0000 (10:32 -0700)]
Merge branch 'jh/notes-merge-in-git-dir-worktree' into maint

Running "notes merge --commit" failed to perform correctly when run
from any directory inside $GIT_DIR/.  When "notes merge" stops with
conflicts, $GIT_DIR/NOTES_MERGE_WORKTREE is the place a user edits
to resolve it.

By Johan Herland (3) and Junio C Hamano (1)
* jh/notes-merge-in-git-dir-worktree:
  notes-merge: Don't remove .git/NOTES_MERGE_WORKTREE; it may be the user's cwd
  notes-merge: use opendir/readdir instead of using read_directory()
  t3310: illustrate failure to "notes merge --commit" inside $GIT_DIR/
  remove_dir_recursively(): Add flag for skipping removal of toplevel dir

12 years agoSync with 1.7.9.7
Junio C Hamano [Thu, 26 Apr 2012 17:31:43 +0000 (10:31 -0700)]
Sync with 1.7.9.7

12 years agoGit 1.7.9.7 v1.7.9.7
Junio C Hamano [Thu, 26 Apr 2012 17:23:15 +0000 (10:23 -0700)]
Git 1.7.9.7

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoSync with 1.7.8.6
Junio C Hamano [Thu, 26 Apr 2012 17:22:20 +0000 (10:22 -0700)]
Sync with 1.7.8.6

12 years agoGit 1.7.8.6 v1.7.8.6
Junio C Hamano [Thu, 26 Apr 2012 16:59:22 +0000 (09:59 -0700)]
Git 1.7.8.6

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoSync with 1.7.7.7
Junio C Hamano [Thu, 26 Apr 2012 16:52:33 +0000 (09:52 -0700)]
Sync with 1.7.7.7

12 years agoGit 1.7.7.7 v1.7.7.7
Junio C Hamano [Thu, 26 Apr 2012 16:41:52 +0000 (09:41 -0700)]
Git 1.7.7.7

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agodiff --no-index: use strbuf for temporary pathnames
Junio C Hamano [Wed, 25 Apr 2012 19:37:38 +0000 (12:37 -0700)]
diff --no-index: use strbuf for temporary pathnames

Instead of using limited-length buffers and risking of pathname
truncation, we should be taking advantage of strbuf API nowadays.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agobundle: remove stray single-quote from error message
Jonathan Nieder [Thu, 26 Apr 2012 05:53:59 +0000 (00:53 -0500)]
bundle: remove stray single-quote from error message

After running rev-list --boundary to retrieve the list of boundary
commits, "git bundle create" runs its own revision walk.  If in this
stage git encounters an unfamiliar option, it writes a message with an
unbalanced quotation mark:

error: unrecognized argument: --foo'

Drop the stray quote to match the "unrecognized argument: %s" message
used elsewhere and save translators some work.

This is mostly a futureproofing measure: for now, the "rev-list
--boundary" command catches most strange arguments on its own and the
above message is not seen unless you try something esoteric like "git
bundle create test.bundle --header HEAD".

Reported-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoconfig: reject bogus section names for --rename-section
Jeff King [Thu, 26 Apr 2012 01:47:14 +0000 (21:47 -0400)]
config: reject bogus section names for --rename-section

You can feed junk to "git config --rename-section", which
will result in a config file that git will not even parse
(so you cannot fix it with git-config). We already have
syntactic sanity checks when setting a variable; let's do
the same for section names.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogitweb: Option to omit column with time of the last change
Kacper Kornet [Tue, 24 Apr 2012 17:39:15 +0000 (19:39 +0200)]
gitweb: Option to omit column with time of the last change

Generating information about last change for a large number of git
repositories can be very time consuming. This commit add an option to
omit 'Last Change' column when presenting the list of repositories.

Signed-off-by: Kacper Kornet <draenog@pld-linux.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogitweb: Don't set owner if got empty value from projects.list
Kacper Kornet [Tue, 24 Apr 2012 17:50:05 +0000 (19:50 +0200)]
gitweb: Don't set owner if got empty value from projects.list

Prevent setting owner to an empty value if it is not specified in
projects.list file. Otherwise it stops retrieving information about the
owner from other files.

Signed-off-by: Kacper Kornet <draenog@pld-linux.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocommit: remove commit_list_reverse()
René Scharfe [Wed, 25 Apr 2012 20:35:54 +0000 (22:35 +0200)]
commit: remove commit_list_reverse()

The function commit_list_reverse() is not used anymore; delete it.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agorevision: append to list instead of insert and reverse
René Scharfe [Wed, 25 Apr 2012 20:35:41 +0000 (22:35 +0200)]
revision: append to list instead of insert and reverse

By using commit_list_insert(), we added new items to the top of the
list and, since this is not the order we want, reversed it afterwards.
Simplify this process by adding new items at the bottom instead,
getting rid of the reversal step.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agosequencer: export commit_list_append()
René Scharfe [Wed, 25 Apr 2012 20:35:27 +0000 (22:35 +0200)]
sequencer: export commit_list_append()

This function can be used in other parts of git.  Give it a new home
in commit.c.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-rebase: add keep_empty flag
Neil Horman [Fri, 20 Apr 2012 14:36:17 +0000 (10:36 -0400)]
git-rebase: add keep_empty flag

Add a command line switch to git-rebase to allow a user the ability to specify
that they want to keep any commits in a series that are empty.

When git-rebase's type is am, then this option will automatically keep any
commit that has a tree object identical to its parent.

This patch changes the default behavior of interactive rebases as well.  With
this patch, git-rebase -i will produce a revision set passed to
git-revision-editor, in which empty commits are commented out.  Empty commits
may be kept manually by uncommenting them.  If the new --keep-empty option is
used in an interactive rebase the empty commits will automatically all be
uncommented in the editor.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-remote-testgit: fix race when spawning fast-import
Pete Wyckoff [Sun, 22 Apr 2012 20:30:58 +0000 (16:30 -0400)]
git-remote-testgit: fix race when spawning fast-import

Test "pushing to local repo" in t5800-remote-helpers can hang
due to a race condition in git-remote-testgit.  Fix it by
setting stdin to unbuffered.

On the writer side, "git push" invokes push_refs_with_export(),
which sends to stdout the command "export\n" and immediately
starts up "git fast-export".  The latter writes its output stream
to the same stdout.

On the reader side, remote helper "git-remote-testgit" reads from
stdin to get its next command.  It uses getc() to read characters
from libc up until \n.  Libc has buffered a potentially much
larger chunk of stdin.  When it sees the "export\n" command, it
forks "git fast-import" to read the stream.

If fast-export finishes before git fast-import starts, the
fast-export output can end up in libc's buffer in
git-remote-testgit, rather than in git fast-import.  The latter
hangs indefinitely on a now-empty stdin.

Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-cherry-pick: Add test to validate new options
Neil Horman [Fri, 20 Apr 2012 14:36:16 +0000 (10:36 -0400)]
git-cherry-pick: Add test to validate new options

Since we've added the --allow-empty and --keep-redundant-commits
options to git cherry-pick we should also add a test to ensure that its working
properly.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-cherry-pick: Add keep-redundant-commits option
Neil Horman [Fri, 20 Apr 2012 14:36:15 +0000 (10:36 -0400)]
git-cherry-pick: Add keep-redundant-commits option

The git-cherry-pick --allow-empty command by default only preserves empty
commits that were originally empty, i.e only those commits for which
<commit>^{tree} and <commit>^^{tree} are equal.  By default commits which are
non-empty, but were made empty by the inclusion of a prior commit on the current
history are filtered out.  This option allows us to override that behavior and
include redundant commits as empty commits in the change history.

Note that this patch changes the default behavior of git cherry-pick slightly.
Prior to this patch all commits in a cherry-pick sequence were applied and git
commit was run.  The implication here was that, if a commit was redundant, and
the commit did not trigger the fast forward logic, the git commit operation, and
therefore the git cherry-pick operation would fail, displaying the cherry pick
advice (i.e. run git commit --allow-empty).  With this patch however, such
redundant commits are automatically skipped without stopping, unless
--keep-redundant-commits is specified, in which case, they are automatically
applied as empty commits.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoThe fifth batch of topics graduated to 'master'
Junio C Hamano [Tue, 24 Apr 2012 21:50:06 +0000 (14:50 -0700)]
The fifth batch of topics graduated to 'master'

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'fc/completion-tests'
Junio C Hamano [Tue, 24 Apr 2012 21:41:22 +0000 (14:41 -0700)]
Merge branch 'fc/completion-tests'

By Felipe Contreras (4) and others
* fc/completion-tests:
  completion: fix completion after 'git --option <TAB>'
  completion: avoid trailing space for --exec-path
  completion: add missing general options
  completion: simplify by using $prev
  completion: simplify __gitcomp_1
  tests: add tests for the __gitcomp() completion helper function
  tests: add initial bash completion tests

12 years agoMerge branch 'cc/fix-missing-va-end-in-revert'
Junio C Hamano [Tue, 24 Apr 2012 21:41:16 +0000 (14:41 -0700)]
Merge branch 'cc/fix-missing-va-end-in-revert'

By Christian Couder
* cc/fix-missing-va-end-in-revert:
  revert: add missing va_end

12 years agoMerge branch 'zj/upstream-error-message'
Junio C Hamano [Tue, 24 Apr 2012 21:41:08 +0000 (14:41 -0700)]
Merge branch 'zj/upstream-error-message'

Error message given when @{u} is used for a branch without its
upstream configured have been clatified.

By Zbigniew Jędrzejewski-Szmek
* zj/upstream-error-message:
  i18n: mark @{upstream} error messages for translation
  Be more specific if upstream branch is not tracked
  Provide better message for barnhc_wiht_tpyo@{u}
  Provide branch name in error message when using @{u}
  t1507: add tests to document @{upstream} behaviour

12 years agoMerge branch 'mk/gitweb-diff-hl'
Junio C Hamano [Tue, 24 Apr 2012 21:41:01 +0000 (14:41 -0700)]
Merge branch 'mk/gitweb-diff-hl'

"gitweb" learns to highlight the patch it outputs even more.

By Michał Kiedrowicz (7) and Jakub Narębski (1)
* mk/gitweb-diff-hl:
  gitweb: Refinement highlightning in combined diffs
  gitweb: Highlight interesting parts of diff
  gitweb: Push formatting diff lines to print_diff_chunk()
  gitweb: Use print_diff_chunk() for both side-by-side and inline diffs
  gitweb: Extract print_sidebyside_diff_lines()
  gitweb: Pass esc_html_hl_regions() options to esc_html()
  gitweb: esc_html_hl_regions(): Don't create empty <span> elements
  gitweb: Use descriptive names in esc_html_hl_regions()

12 years agoMerge branch 'it/fetch-pack-many-refs'
Junio C Hamano [Tue, 24 Apr 2012 21:40:51 +0000 (14:40 -0700)]
Merge branch 'it/fetch-pack-many-refs'

When "git fetch" encounters repositories with too many references, the
command line of "fetch-pack" that is run by a helper e.g. remote-curl,
may fail to hold all of them. Now such an internal invocation can feed
the references through the standard input of "fetch-pack".

By Ivan Todoroski
* it/fetch-pack-many-refs:
  remote-curl: main test case for the OS command line overflow
  fetch-pack: test cases for the new --stdin option
  remote-curl: send the refs to fetch-pack on stdin
  fetch-pack: new --stdin option to read refs from stdin

12 years agoMerge branch 'pw/fast-import-dataref-parsing'
Junio C Hamano [Tue, 24 Apr 2012 21:40:45 +0000 (14:40 -0700)]
Merge branch 'pw/fast-import-dataref-parsing'

The parser in "fast-import" did not diagnose ":9" style references
that is not followed by required SP/LF as an error.

By Pete Wyckoff
* pw/fast-import-dataref-parsing:
  fast-import: tighten parsing of datarefs

12 years agoMerge branch 'mh/ref-api'
Junio C Hamano [Tue, 24 Apr 2012 21:40:28 +0000 (14:40 -0700)]
Merge branch 'mh/ref-api'

Allows walking only a sub-hierarchy in refs/*.

By Michael Haggerty
* mh/ref-api:
  do_for_each_ref(): only iterate over the subtree that was requested
  refs: store references hierarchically
  sort_ref_dir(): simplify logic
  refs.c: rename ref_array -> ref_dir
  struct ref_entry: nest the value part in a union
  check_refname_component(): return 0 for zero-length components
  free_ref_entry(): new function
  names_conflict(): simplify implementation
  repack_without_ref(): reimplement using do_for_each_ref_in_array()
  do_for_each_ref_in_arrays(): new function
  do_for_each_ref_in_array(): new function
  refs: manage current_ref within do_one_ref()
  refs.c: reorder definitions more logically

12 years agoMerge branch 'hv/submodule-recurse-push'
Junio C Hamano [Tue, 24 Apr 2012 21:40:20 +0000 (14:40 -0700)]
Merge branch 'hv/submodule-recurse-push'

"git push --recurse-submodules" learns to optionally look into the
histories of submodules bound to the superproject and push them out.

By Heiko Voigt
* hv/submodule-recurse-push:
  push: teach --recurse-submodules the on-demand option
  Refactor submodule push check to use string list instead of integer
  Teach revision walking machinery to walk multiple times sequencially

12 years agogit-svn: drop redundant blocking of SIGPIPE
Roman Kagan [Mon, 23 Apr 2012 16:26:56 +0000 (20:26 +0400)]
git-svn: drop redundant blocking of SIGPIPE

Now that SIGPIPE is ignored there's no point blocking it.

Signed-off-by: Roman Kagan <rkagan@mail.ru>
Acked-by: Eric Wong <normalperson@yhbt.net>
12 years agogit-svn: ignore SIGPIPE
Roman Kagan [Mon, 2 Apr 2012 13:52:34 +0000 (17:52 +0400)]
git-svn: ignore SIGPIPE

In HTTP with keep-alive it's not uncommon for the client to notice that
the server decided to stop maintaining the current connection only when
sending a new request.  This naturally results in -EPIPE and possibly
SIGPIPE.

The subversion library itself makes no provision for SIGPIPE.  Some
combinations of the underlying libraries do (typically SIG_IGN-ing it),
some don't.

Presumably for that reason all subversion commands set SIGPIPE to
SIG_IGN early in their main()-s.

So should we.

This, together with the previous patch, fixes the notorious "git-svn
died of signal 13" problem (see e.g.
http://thread.gmane.org/gmane.comp.version-control.git/134936).

Signed-off-by: Roman Kagan <rkagan@mail.ru>
Acked-by: Eric Wong <normalperson@yhbt.net>
12 years agogit-svn: use POSIX::sigprocmask to block signals
Roman Kagan [Mon, 2 Apr 2012 13:29:32 +0000 (17:29 +0400)]
git-svn: use POSIX::sigprocmask to block signals

In order to maintain consistency of the database mapping svn revision
numbers to git commit ids, rev_map_set() defers signal processing until
it's finished with an append transaction.[*]

The conventional way to achieve this is through sigprocmask(), which is
available in perl in the standard POSIX module.

This is implemented by this patch.  One important consequence of it is
that the signal handlers won't be unconditionally set to SIG_DFL anymore
upon the first invocation of rev_map_set() as they used to.  As a
result, the signals ignored by git-svn parent will remain ignored;
otherwise the behavior remains the same.

This patch paves the way to ignoring SIGPIPE throughout git-svn which
will be done in the followup patch.

[*] Deferring signals is not enough to ensure the database consistency:
the program may die on SIGKILL or power loss, run out of disk space,
etc.  However that's a separate issue that this patch doesn't address.

Signed-off-by: Roman Kagan <rkagan@mail.ru>
Acked-by: Eric Wong <normalperson@yhbt.net>
12 years agoRelNotes: the fourth batch of topics graduated to 'master'
Junio C Hamano [Mon, 23 Apr 2012 20:30:22 +0000 (13:30 -0700)]
RelNotes: the fourth batch of topics graduated to 'master'

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'bw/test-fix-grep-gnuism'
Junio C Hamano [Mon, 23 Apr 2012 20:02:15 +0000 (13:02 -0700)]
Merge branch 'bw/test-fix-grep-gnuism'

Fix two places that were the only place in the test suite that gave "a\+"
to platform grep and expected it to mean one or more "a", which is a
blatant GNUism.

* bw/test-fix-grep-gnuism:
  t9400: fix gnuism in grep

12 years agoMerge branch 'rj/submodule-mswin-path'
Junio C Hamano [Mon, 23 Apr 2012 20:01:49 +0000 (13:01 -0700)]
Merge branch 'rj/submodule-mswin-path'

By Ramsay Jones
* rj/submodule-mswin-path:
  git-submodule.sh: Don't use $path variable in eval_gettext string

12 years agoMerge branch 'rj/sh-setup-mswin-pwd'
Junio C Hamano [Mon, 23 Apr 2012 20:01:46 +0000 (13:01 -0700)]
Merge branch 'rj/sh-setup-mswin-pwd'

By Ramsay Jones
* rj/sh-setup-mswin-pwd:
  git-sh-setup.sh: Add an pwd() function for MinGW

12 years agoMerge branch 'rj/mingw-isguid'
Junio C Hamano [Mon, 23 Apr 2012 20:01:42 +0000 (13:01 -0700)]
Merge branch 'rj/mingw-isguid'

By Ramsay Jones
* rj/mingw-isguid:
  compat/mingw.h: Set S_ISUID to prevent a fast-import test failure

12 years agoMerge branch 'jn/fast-import-doc'
Junio C Hamano [Mon, 23 Apr 2012 20:01:32 +0000 (13:01 -0700)]
Merge branch 'jn/fast-import-doc'

By Jonathan Nieder
* jn/fast-import-doc:
  fast-import doc: cat-blob and ls responses need to be consumed quickly

12 years agoMerge branch 'jn/more-i18ncmp'
Junio C Hamano [Mon, 23 Apr 2012 20:01:28 +0000 (13:01 -0700)]
Merge branch 'jn/more-i18ncmp'

By Jonathan Nieder
* jn/more-i18ncmp:
  test: am of empty patch should not succeed
  test: use test_i18ncmp for "Patch format detection failed" message
  test: do not rely on US English tracking-info messages

12 years agoMerge branch 'jl/maint-submodule-recurse-fetch'
Junio C Hamano [Mon, 23 Apr 2012 19:58:28 +0000 (12:58 -0700)]
Merge branch 'jl/maint-submodule-recurse-fetch'

"git fetch" that recurses into submodules on demand did not check if
it needs to go into submodules when non branches (most notably, tags)
are fetched.

By Jens Lehmann
* jl/maint-submodule-recurse-fetch:
  submodules: recursive fetch also checks new tags for submodule commits

12 years agoMerge branch 'jc/maint-blame-minimal'
Junio C Hamano [Mon, 23 Apr 2012 19:58:22 +0000 (12:58 -0700)]
Merge branch 'jc/maint-blame-minimal'

"git blame" started missing quite a few changes from the origin since we
stopped using the diff minimalization by default in v1.7.2 era.

* jc/maint-blame-minimal:
  blame: accept --need-minimal

12 years agoMerge branch 'lp/maint-diff-three-dash-with-graph'
Junio C Hamano [Mon, 23 Apr 2012 19:57:20 +0000 (12:57 -0700)]
Merge branch 'lp/maint-diff-three-dash-with-graph'

"log -p --graph" used with "--stat" had a few formatting error.

By Lucian Poston
* lp/maint-diff-three-dash-with-graph:
  t4202: add test for "log --graph --stat -p" separator lines
  log --graph: fix break in graph lines
  log --graph --stat: three-dash separator should come after graph lines

12 years agoMerge branch 'rs/commit-list-sort-in-batch'
Junio C Hamano [Mon, 23 Apr 2012 19:52:54 +0000 (12:52 -0700)]
Merge branch 'rs/commit-list-sort-in-batch'

Setting up a revision traversal with many starting points was inefficient
as these were placed in a date-order priority queue one-by-one.

By René Scharfe (3) and Junio C Hamano (1)
* rs/commit-list-sort-in-batch:
  mergesort: rename it to llist_mergesort()
  revision: insert unsorted, then sort in prepare_revision_walk()
  commit: use mergesort() in commit_list_sort_by_date()
  add mergesort() for linked lists

12 years agoMerge branch 'jh/apply-free-patch'
Junio C Hamano [Mon, 23 Apr 2012 19:52:18 +0000 (12:52 -0700)]
Merge branch 'jh/apply-free-patch'

Valgrind reports quite a lot of discarded memory inside apply.
Fix them, audit and document the buffer ownership rules.

By Junio C Hamano (8) and Jared Hance (1)
* jh/apply-free-patch:
  apply: document buffer ownership rules across functions
  apply: tighten constness of line buffer
  apply: drop unused macro
  apply: free unused fragments for submodule patch
  apply: free patch->result
  apply: release memory for fn_table
  apply: free patch->{def,old,new}_name fields
  apply: rename free_patch() to free_patch_list()
  apply: do not leak patches and fragments

12 years agoMerge branch 'sl/test-wc-l-line-count'
Junio C Hamano [Mon, 23 Apr 2012 19:43:10 +0000 (12:43 -0700)]
Merge branch 'sl/test-wc-l-line-count'

By Stefano Lattarini
* sl/test-wc-l-line-count:
  tests: modernise style: more uses of test_line_count

12 years agoMerge branch 'jn/debian-customizes-default-editor'
Junio C Hamano [Mon, 23 Apr 2012 19:41:15 +0000 (12:41 -0700)]
Merge branch 'jn/debian-customizes-default-editor'

Make it easier for distros to document custom pager and editor they
used when building their binary releases in "git var" documentation.

By Jonathan Nieder
* jn/debian-customizes-default-editor:
  var doc: advertise current DEFAULT_PAGER and DEFAULT_EDITOR settings
  var doc: default editor and pager are configurable at build time

12 years agoMerge branch 'rl/show-empty-prefix'
Junio C Hamano [Mon, 23 Apr 2012 19:40:08 +0000 (12:40 -0700)]
Merge branch 'rl/show-empty-prefix'

"git rev-parse --show-prefix" emitted nothing when run at the
top-level of the working tree, while "git rev-parse --show-cdup" gave
an empty line.  Make them consistent.

By Ross Lagerwall
* rl/show-empty-prefix:
  rev-parse --show-prefix: add in trailing newline

12 years agoMerge branch 'pw/git-p4'
Junio C Hamano [Mon, 23 Apr 2012 19:40:02 +0000 (12:40 -0700)]
Merge branch 'pw/git-p4'

By Pete Wyckoff
* pw/git-p4:
  git p4: use "git p4" directly in tests
  git p4: update name in script
  git-p4: move to toplevel

12 years agoMerge branch 'ms/remote-usage-string'
Junio C Hamano [Mon, 23 Apr 2012 19:39:17 +0000 (12:39 -0700)]
Merge branch 'ms/remote-usage-string'

Adds some subcommands that were not listed in "git remote --help"
usage strings.

As an independent follow-up, we may want to rethink how the overall
usage string and subcommand usage strings are maintained.

By Michael Schubert
* ms/remote-usage-string:
  remote: update builtin usage

12 years agoMerge branch 'jk/rebase-i-submodule-conflict-only'
Junio C Hamano [Mon, 23 Apr 2012 19:39:04 +0000 (12:39 -0700)]
Merge branch 'jk/rebase-i-submodule-conflict-only'

Giving "--continue" to a conflicted "rebase -i" session skipped a commit
that only results in changes to submodules.

By John Keeping
* jk/rebase-i-submodule-conflict-only:
  rebase -i continue: don't skip commits that only change submodules

12 years agocompletion: fix completion after 'git --option <TAB>'
SZEDER Gábor [Sun, 15 Apr 2012 19:44:20 +0000 (22:44 +0300)]
completion: fix completion after 'git --option <TAB>'

The bash completion doesn't work when certain options to git itself are
specified, e.g. 'git --no-pager <TAB>' errors out with

    error: invalid key: alias.--no-pager

The main _git() completion function finds out the git command name by
looping through all the words on the command line and searching for
the first word that is not a known option for the git command.

Unfortunately the list of known git options was not updated in a long
time, and newer options are not skipped but mistaken for a git command.
Such a misrecognized "command" is then passed to __git_aliased_command(),
which in turn passes it to a 'git config' query, hence the error.

Currently the following options are misrecognized for a git command:

  -c --no-pager --exec-path --html-path --man-path --info-path
  --no-replace-objects --work-tree= --namespace=

To fix this we could just update the list of options to be skipped,
but the same issue will likely arise, if the git command learns a new
option in the future.  Therefore, to make it more future proof against
new options, this patch changes that loop to skip all option-looking
words, i.e. words starting with a dash.

We also have to handle the '-c' option specially, because it takes a
configutation parameter in a separate word, which must be skipped,
too.

[fc: added tests]

Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: avoid trailing space for --exec-path
Jonathan Nieder [Sun, 15 Apr 2012 19:44:19 +0000 (22:44 +0300)]
completion: avoid trailing space for --exec-path

"--exec-path" looks to the completion script like an unambiguous
successful completion, but it is wrong to emit a SP after it as if
declaring that we are done with completion; the user could be trying
to do

git --exec-path; # print name of helper directory

or

git --exec-path=/path/to/alternative/helper/dir <subcommand>

so the most helpful thing to do is to leave out the trailing space and
leave it to the operator to type an equal sign or carriage return
according to the situation.

[fc: added tests]

Cc: Andreas Schwab <schwab@linux-m68k.org>
Reported-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: add missing general options
Felipe Contreras [Sun, 15 Apr 2012 19:44:18 +0000 (22:44 +0300)]
completion: add missing general options

And add relevant tests.

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: simplify by using $prev
Felipe Contreras [Sun, 15 Apr 2012 19:44:17 +0000 (22:44 +0300)]
completion: simplify by using $prev

cword-1 is the previous word ($prev).

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: simplify __gitcomp_1
Felipe Contreras [Sun, 15 Apr 2012 19:44:16 +0000 (22:44 +0300)]
completion: simplify __gitcomp_1

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agotests: add tests for the __gitcomp() completion helper function
SZEDER Gábor [Tue, 17 Apr 2012 10:27:15 +0000 (12:27 +0200)]
tests: add tests for the __gitcomp() completion helper function

These tests check that trailing space, prefix, and suffix are added
correctly.

Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agorevert: add missing va_end
Christian Couder [Sat, 21 Apr 2012 21:15:19 +0000 (23:15 +0200)]
revert: add missing va_end

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoRelNotes: the third batch
Junio C Hamano [Fri, 20 Apr 2012 22:58:40 +0000 (15:58 -0700)]
RelNotes: the third batch

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'js/spawn-via-shell-path-fix'
Junio C Hamano [Fri, 20 Apr 2012 22:51:18 +0000 (15:51 -0700)]
Merge branch 'js/spawn-via-shell-path-fix'

Mops up an unfortunate fallout from bw/spawn-via-shell-path topic.

By Johannes Sixt
* js/spawn-via-shell-path-fix:
  Do not use SHELL_PATH from build system in prepare_shell_cmd on Windows

12 years agoMerge branch 'ct/advise-push-default'
Junio C Hamano [Fri, 20 Apr 2012 22:50:37 +0000 (15:50 -0700)]
Merge branch 'ct/advise-push-default'

Break down the cases in which "git push" fails due to non-ff into
three categories, and give separate advise messages for each case.

By Christopher Tiwald (2) and Jeff King (1)
* ct/advise-push-default:
  Fix httpd tests that broke when non-ff push advice changed
  clean up struct ref's nonfastforward field
  push: Provide situational hints for non-fast-forward errors

12 years agoMerge branch 'bw/submodule-sed-solaris'
Junio C Hamano [Fri, 20 Apr 2012 22:50:13 +0000 (15:50 -0700)]
Merge branch 'bw/submodule-sed-solaris'

By Ben Walton
* bw/submodule-sed-solaris:
  Avoid bug in Solaris xpg4/sed as used in submodule

12 years agoMerge branch 'jk/run-command-eacces'
Junio C Hamano [Fri, 20 Apr 2012 22:50:03 +0000 (15:50 -0700)]
Merge branch 'jk/run-command-eacces'

When PATH contains an unreadable directory, alias expansion code did not
kick in, and failed with an error that said "git-subcmd" was not found.

By Jeff King (1) and Ramsay Jones (1)
* jk/run-command-eacces:
  run-command: treat inaccessible directories as ENOENT
  compat/mingw.[ch]: Change return type of exec functions to int

12 years agoMerge branch 'jc/push-upstream-sanity'
Junio C Hamano [Fri, 20 Apr 2012 22:49:51 +0000 (15:49 -0700)]
Merge branch 'jc/push-upstream-sanity'

Fix broken 'push to upstream' implementation.  "git push $there" without
refspec, when the current branch is set to push to a remote different from
$there, used to push to $there using the upstream information to a remote
unreleated to $there.

* jc/push-upstream-sanity:
  push: error out when the "upstream" semantics does not make sense