git
7 years agoMerge branch 'ab/c-translators-comment-style' into maint
Junio C Hamano [Mon, 5 Jun 2017 00:03:10 +0000 (09:03 +0900)] 
Merge branch 'ab/c-translators-comment-style' into maint

Update the C style recommendation for notes for translators, as
recent versions of gettext tools can work with our style of
multi-line comments.

* ab/c-translators-comment-style:
  C style: use standard style for "TRANSLATORS" comments

7 years agoMerge branch 'ls/travis-doc-asciidoctor' into maint
Junio C Hamano [Mon, 5 Jun 2017 00:03:10 +0000 (09:03 +0900)] 
Merge branch 'ls/travis-doc-asciidoctor' into maint

Travis CI gained a task to format the documentation with both
AsciiDoc and AsciiDoctor.

* ls/travis-doc-asciidoctor:
  travis-ci: check AsciiDoc/AsciiDoctor stderr output
  travis-ci: unset compiler for jobs that do not need one
  travis-ci: parallelize documentation build
  travis-ci: build documentation with AsciiDoc and Asciidoctor

7 years agoPrepare for 2.13.1; more topics to follow
Junio C Hamano [Sun, 4 Jun 2017 01:25:42 +0000 (10:25 +0900)] 
Prepare for 2.13.1; more topics to follow

Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoMerge branch 'tg/stash-push-fixup' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:07 +0000 (10:21 +0900)] 
Merge branch 'tg/stash-push-fixup' into maint

The shell completion script (in contrib/) learned "git stash" has
a new "push" subcommand.

* tg/stash-push-fixup:
  completion: add git stash push

7 years agoMerge branch 'km/log-showsignature-doc' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:07 +0000 (10:21 +0900)] 
Merge branch 'km/log-showsignature-doc' into maint

Doc update.

* km/log-showsignature-doc:
  config.txt: add an entry for log.showSignature

7 years agoMerge branch 'jt/use-trailer-api-in-commands' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:06 +0000 (10:21 +0900)] 
Merge branch 'jt/use-trailer-api-in-commands' into maint

"git cherry-pick" and other uses of the sequencer machinery
mishandled a trailer block whose last line is an incomplete line.
This has been fixed so that an additional sign-off etc. are added
after completing the existing incomplete line.

* jt/use-trailer-api-in-commands:
  sequencer: add newline before adding footers

7 years agoMerge branch 'jt/push-options-doc' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:05 +0000 (10:21 +0900)] 
Merge branch 'jt/push-options-doc' into maint

The receive-pack program now makes sure that the push certificate
records the same set of push options used for pushing.

* jt/push-options-doc:
  receive-pack: verify push options in cert
  docs: correct receive.advertisePushOptions default

7 years agoMerge branch 'js/plug-leaks' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:04 +0000 (10:21 +0900)] 
Merge branch 'js/plug-leaks' into maint

Fix memory leaks pointed out by Coverity (and people).

* js/plug-leaks: (26 commits)
  checkout: fix memory leak
  submodule_uses_worktrees(): plug memory leak
  show_worktree(): plug memory leak
  name-rev: avoid leaking memory in the `deref` case
  remote: plug memory leak in match_explicit()
  add_reflog_for_walk: avoid memory leak
  shallow: avoid memory leak
  line-log: avoid memory leak
  receive-pack: plug memory leak in update()
  fast-export: avoid leaking memory in handle_tag()
  mktree: plug memory leaks reported by Coverity
  pack-redundant: plug memory leak
  setup_discovered_git_dir(): plug memory leak
  setup_bare_git_dir(): help static analysis
  split_commit_in_progress(): simplify & fix memory leak
  checkout: fix memory leak
  cat-file: fix memory leak
  mailinfo & mailsplit: check for EOF while parsing
  status: close file descriptor after reading git-rebase-todo
  difftool: address a couple of resource/memory leaks
  ...

7 years agoMerge branch 'js/eol-on-ourselves' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:04 +0000 (10:21 +0900)] 
Merge branch 'js/eol-on-ourselves' into maint

Make sure our tests would pass when the sources are checked out
with "platform native" line ending convention by default on
Windows.  Some "text" files out tests use and the test scripts
themselves that are meant to be run with /bin/sh, ought to be
checked out with eol=LF even on Windows.

* js/eol-on-ourselves:
  t4051: mark supporting files as requiring LF-only line endings
  Fix the remaining tests that failed with core.autocrlf=true
  t3901: move supporting files into t/t3901/
  completion: mark bash script as LF-only
  git-new-workdir: mark script as LF-only
  Fix build with core.autocrlf=true

7 years agoMerge branch 'jk/update-links-in-docs' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:04 +0000 (10:21 +0900)] 
Merge branch 'jk/update-links-in-docs' into maint

A few http:// links that are redirected to https:// in the
documentation have been updated to https:// links.

* jk/update-links-in-docs:
  doc: use https links to Wikipedia to avoid http redirects

7 years agoMerge branch 'jk/ignore-broken-tags-when-ignoring-missing-links' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:03 +0000 (10:21 +0900)] 
Merge branch 'jk/ignore-broken-tags-when-ignoring-missing-links' into maint

Tag objects, which are not reachable from any ref, that point at
missing objects were mishandled by "git gc" and friends (they
should silently be ignored instead)

* jk/ignore-broken-tags-when-ignoring-missing-links:
  revision.c: ignore broken tags with ignore_missing_links

7 years agoMerge branch 'jk/doc-config-include' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:02 +0000 (10:21 +0900)] 
Merge branch 'jk/doc-config-include' into maint

Clarify documentation for include.path and includeIf.<condition>.path
configuration variables.

* jk/doc-config-include:
  docs/config: consistify include.path examples
  docs/config: avoid the term "expand" for includes
  docs/config: give a relative includeIf example
  docs/config: clarify include/includeIf relationship

7 years agoMerge branch 'jk/disable-pack-reuse-when-broken' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:02 +0000 (10:21 +0900)] 
Merge branch 'jk/disable-pack-reuse-when-broken' into maint

"pack-objects" can stream a slice of an existing packfile out when
the pack bitmap can tell that the reachable objects are all needed
in the output, without inspecting individual objects.  This
strategy however would not work well when "--local" and other
options are in use, and need to be disabled.

* jk/disable-pack-reuse-when-broken:
  t5310: fix "; do" style
  pack-objects: disable pack reuse for object-selection options

7 years agoMerge branch 'jk/am-leakfix' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:01 +0000 (10:21 +0900)] 
Merge branch 'jk/am-leakfix' into maint

The codepath in "git am" that is used when running "git rebase"
leaked memory held for the log message of the commits being rebased.

* jk/am-leakfix:
  am: shorten ident_split variable name in get_commit_info()
  am: simplify allocations in get_commit_info()
  am: fix commit buffer leak in get_commit_info()

7 years agoMerge branch 'jc/read-tree-empty-with-m' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:21:00 +0000 (10:21 +0900)] 
Merge branch 'jc/read-tree-empty-with-m' into maint

"git read-tree -m" (no tree-ish) gave a nonsense suggestion "use
--empty if you want to clear the index".  With "-m", such a request
will still fail anyway, as you'd need to name at least one tree-ish
to be merged.

* jc/read-tree-empty-with-m:
  read-tree: "read-tree -m --empty" does not make sense

7 years agoMerge branch 'jc/apply-fix-mismerge' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:20:59 +0000 (10:20 +0900)] 
Merge branch 'jc/apply-fix-mismerge' into maint

Mismerge fix.

* jc/apply-fix-mismerge:
  apply.c: fix whitespace-only mismerge

7 years agoMerge branch 'ja/do-not-ask-needless-questions' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:20:58 +0000 (10:20 +0900)] 
Merge branch 'ja/do-not-ask-needless-questions' into maint

Git sometimes gives an advice in a rhetorical question that does
not require an answer, which can confuse new users and non native
speakers.  Attempt to rephrase them.

* ja/do-not-ask-needless-questions:
  git-filter-branch: be more direct in an error message
  read-tree -m: make error message for merging 0 trees less smart aleck
  usability: don't ask questions if no reply is required

7 years agoMerge branch 'bw/submodule-with-bs-path' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:20:58 +0000 (10:20 +0900)] 
Merge branch 'bw/submodule-with-bs-path' into maint

A hotfix to a topic that is already in v2.13.

* bw/submodule-with-bs-path:
  t7400: add !CYGWIN prerequisite to 'add with \\ in path'

7 years agoMerge branch 'ah/log-decorate-default-to-auto' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:20:57 +0000 (10:20 +0900)] 
Merge branch 'ah/log-decorate-default-to-auto' into maint

Setting "log.decorate=false" in the configuration file did not take
effect in v2.13, which has been corrected.

* ah/log-decorate-default-to-auto:
  builtin/log: honor log.decorate

7 years agoMerge branch 'ab/fix-poison-tests' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:20:56 +0000 (10:20 +0900)] 
Merge branch 'ab/fix-poison-tests' into maint

Update tests to pass under GETTEXT_POISON (a mechanism to ensure
that output strings that should not be translated are not
translated by mistake), and tell TravisCI to run them.

* ab/fix-poison-tests:
  travis-ci: add job to run tests with GETTEXT_POISON
  travis-ci: setup "prove cache" in "script" step
  tests: fix tests broken under GETTEXT_POISON=YesPlease

7 years agoMerge branch 'ab/doc-replace-gmane-links' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:20:56 +0000 (10:20 +0900)] 
Merge branch 'ab/doc-replace-gmane-links' into maint

The Web interface to gmane news archive is long gone, even though
the articles are still accessible via NTTP.  Replace the links with
ones to public-inbox.org.  Because their message identification is
based on the actual message-id, it is likely that it will be easier
to migrate away from it if/when necessary.

* ab/doc-replace-gmane-links:
  doc: replace more gmane links
  doc: replace a couple of broken gmane links

7 years agoMerge branch 'ab/aix-needs-compat-regex' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:20:56 +0000 (10:20 +0900)] 
Merge branch 'ab/aix-needs-compat-regex' into maint

Build fix.

* ab/aix-needs-compat-regex:
  config.mak.uname: set NO_REGEX=NeedsStartEnd on AIX

7 years agoMerge branch 'pw/rebase-i-regression-fix' into maint
Junio C Hamano [Sun, 4 Jun 2017 01:20:55 +0000 (10:20 +0900)] 
Merge branch 'pw/rebase-i-regression-fix' into maint

Just the first one of three? new tests that follows up a regression
fix.

* pw/rebase-i-regression-fix:
  rebase -i: add missing newline to end of message
  rebase -i: silence stash apply
  rebase -i: fix reflog message

7 years agoC style: use standard style for "TRANSLATORS" comments
Ævar Arnfjörð Bjarmason [Thu, 11 May 2017 21:20:12 +0000 (21:20 +0000)] 
C style: use standard style for "TRANSLATORS" comments

Change all the "TRANSLATORS: [...]" comments in the C code to use the
regular Git coding style, and amend the style guide so that the
example there uses that style.

This custom style was necessary back in 2010 when the gettext support
was initially added, and was subsequently documented in commit
cbcfd4e3ea ("i18n: mention "TRANSLATORS:" marker in
Documentation/CodingGuidelines", 2014-04-18).

GNU xgettext hasn't had the parsing limitation that necessitated this
exception for almost 3 years. Since its 0.19 release on 2014-06-02
it's been able to recognize TRANSLATOR comments in the standard Git
comment syntax[1].

Usually we'd like to keep compatibility with software that's that
young, but in this case literally the only person who needs to be
using a gettext newer than 3 years old is Jiang Xin (the only person
who runs & commits "make pot" results), so I think in this case we can
make an exception.

This xgettext parsing feature was added after a thread on the Git
mailing list[2] which continued on the bug-gettext[3] list, but we
never subsequently changed our style & styleguide, do so.

There are already longstanding changes in git that use the standard
comment style & have their TRANSLATORS comments extracted properly
without getting the literal "*"'s mixed up in the text, as would
happen before xgettext 0.19.

Commit 7ff2683253 ("builtin-am: implement -i/--interactive",
2015-08-04) added one such comment, which in commit df0617bfa7 ("l10n:
git.pot: v2.6.0 round 1 (123 new, 41 removed)", 2015-09-05) got picked
up in the po/git.pot file with the right format, showing that Jiang
already runs a modern xgettext.

The xgettext parser does not handle the sort of non-standard comment
style that I'm amending here in sequencer.c, but that isn't standard
Git comment syntax anyway. With this change to sequencer.c & "make
pot" the comment in the pot file is now correct:

     #. TRANSLATORS: %s will be "revert", "cherry-pick" or
    -#. * "rebase -i".
    +#. "rebase -i".

1. http://git.savannah.gnu.org/cgit/gettext.git/commit/?id=10af7fe6bd
2. <2ce9ec406501d112e032c8208417f8100bed04c6.1397712142.git.worldhello.net@gmail.com>
   (https://public-inbox.org/git/2ce9ec406501d112e032c8208417f8100bed04c6.1397712142.git.worldhello.net@gmail.com/)
3. https://lists.gnu.org/archive/html/bug-gettext/2014-04/msg00016.html

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Acked-by: Jiang Xin <worldhello.net@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agorevision.c: ignore broken tags with ignore_missing_links
Jeff King [Sat, 20 May 2017 08:30:25 +0000 (04:30 -0400)] 
revision.c: ignore broken tags with ignore_missing_links

When peeling a tag for prepare_revision_walk(), we do not
respect the ignore_missing_links flag. This can lead to a
bogus error when pack-objects walks the possibly-broken
unreachable-but-recent part of the object graph.

The other link-following all happens via traverse_commit_list(),
which explains why this case was missed. And our tests
covered only broken links from commits. Let's be more
comprehensive and cover broken tree entries (which do work)
and tags (which shows off this bug).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoconfig.txt: add an entry for log.showSignature
Kyle Meyer [Fri, 19 May 2017 02:44:51 +0000 (22:44 -0400)] 
config.txt: add an entry for log.showSignature

The configuration variable log.showSignature is mentioned in git-log's
manpage.  Document it in git-config's manpage as well.

Signed-off-by: Kyle Meyer <kyle@kyleam.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agorebase -i: add missing newline to end of message
Phillip Wood [Fri, 19 May 2017 14:32:48 +0000 (15:32 +0100)] 
rebase -i: add missing newline to end of message

The message that's printed when auto-stashed changes are successfully
restored was missing '\n' at the end.

Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agorebase -i: silence stash apply
Phillip Wood [Thu, 18 May 2017 10:02:33 +0000 (11:02 +0100)] 
rebase -i: silence stash apply

The shell version of rebase -i silences the status output from 'git
stash apply' when restoring the autostashed changes. The C version
does not.

Having the output from git stash apply on the screen is
distracting as it makes it difficult to find the message from git
rebase saying that the rebase succeeded. Also the status information
that git stash prints talks about looking in .git/rebase-merge/done to
see which commits have been applied. As .git/rebase-merge is removed
shortly after the message is printed before rebase -i exits this is
confusing.

Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agorebase -i: fix reflog message
Phillip Wood [Thu, 18 May 2017 10:02:32 +0000 (11:02 +0100)] 
rebase -i: fix reflog message

When rebase -i was converted to C a bug was introduced into the code
that creates the reflog message. Instead of saying
rebase -i (finish): <head-name> onto <onto>
it says
rebase -i (finish): <head-name> onto <orig-head><onto>
as the strbuf is not reset between reading the value of <orig-head>
and <onto>.

Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agocompletion: add git stash push
Thomas Gummerer [Tue, 16 May 2017 19:59:45 +0000 (20:59 +0100)] 
completion: add git stash push

When introducing git stash push in f5727e26e4 ("stash: introduce push
verb", 2017-02-19), I forgot to add it to the completion code. Add it
now.

Signed-off-by: Thomas Gummerer <t.gummerer@gmail.com>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agodoc: use https links to Wikipedia to avoid http redirects
Sven Strickroth [Sat, 13 May 2017 09:54:51 +0000 (11:54 +0200)] 
doc: use https links to Wikipedia to avoid http redirects

Signed-off-by: Sven Strickroth <email@cs-ware.de>
Reviewed-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agobuiltin/log: honor log.decorate
brian m. carlson [Sun, 14 May 2017 18:00:58 +0000 (18:00 +0000)] 
builtin/log: honor log.decorate

The recent change that introduced autodecorating of refs accidentally
broke the ability of users to set log.decorate = false to override it.
When the git_log_config was traversed a second time with an option other
than log.decorate, the decoration style would be set to the automatic
style, even if the user had already overridden it.  Instead of setting
the option in config parsing, set it in init_log_defaults instead.

Add a test for this case.  The actual additional config option doesn't
matter, but it needs to be something not already set in the
configuration file.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Acked-by: Alex Henrie <alexhenrie24@gmail.com>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agogit-filter-branch: be more direct in an error message
Jean-Noel Avila [Thu, 11 May 2017 12:06:34 +0000 (14:06 +0200)] 
git-filter-branch: be more direct in an error message

git-filter-branch requires the specification of a branch by one way or
another. If no branch appears to have been specified, we know the user
got the usage wrong but we don't know what they were trying to do ---
e.g. maybe they specified the ref to rewrite but in the wrong place.

In this case, just state that the branch specification is missing.

Signed-off-by: Jean-Noel Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoread-tree -m: make error message for merging 0 trees less smart aleck
Jean-Noel Avila [Thu, 11 May 2017 12:06:33 +0000 (14:06 +0200)] 
read-tree -m: make error message for merging 0 trees less smart aleck

"git read-tree -m" requires a tree argument to name the tree to be
merged in.  Git uses a cutesy error message to say so and why:

    $ git read-tree -m
    warning: read-tree: emptying the index with no arguments is
    deprecated; use --empty
    fatal: just how do you expect me to merge 0 trees?
    $ git read-tree -m --empty
    fatal: just how do you expect me to merge 0 trees?

When lucky, that could produce an ah-hah moment for the user, but it's
more likely to irritate and distract them.

Instead, tell the user plainly that the tree argument is
required. Also document this requirement in the git-read-tree(1)
manpage where there is room to explain it in a more straightforward way.

Signed-off-by: Jean-Noel Avila <jn.avila@free.fr>
Helped-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agousability: don't ask questions if no reply is required
Jean-Noel Avila [Thu, 11 May 2017 12:06:32 +0000 (14:06 +0200)] 
usability: don't ask questions if no reply is required

There has been a bug report by a corporate user that stated that
"spelling mistake of stash followed by a yes prints character 'y'
infinite times."

This analysis was false. When the spelling of a command contains
errors, the git program tries to help the user by providing candidates
which are close to the unexisting command. E.g Git prints the
following:

        git: 'stahs' is not a git command. See 'git --help'.
        Did you mean this?

        stash

and then exits.

The problem with this hint is that it is not formally indicated as an
hint and the user is in fact encouraged to reply to the question,
whereas the Git command is already finished.

The user was unlucky enough that it was the command he was looking
for, and replied "yes" on the command line, effectively launching the
`yes` program.

The initial error is that the Git programs, when launched in
command-line mode (without interaction) must not ask questions,
because these questions would normally require a user input as a reply
that they won't handle indeed. That's a source of confusion on UX
level.

To improve the general usability of the Git suite, the following rule
was applied:

if the sentence
 * appears in a non-interactive session
 * is printed last before exit
 * is a question addressing the user ("you")

the sentence is turned into affirmative and proposes the option.

The basic rewording of the question sentences has been extended to
other spots found in the source.

Requested at https://github.com/git/git-scm.com/issues/999 by rpai1

Signed-off-by: Jean-Noel Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agodocs/config: consistify include.path examples
Jeff King [Thu, 11 May 2017 09:14:30 +0000 (05:14 -0400)] 
docs/config: consistify include.path examples

Most of the include examples use "foo.inc", but some use
"foo". Since the string of examples are meant to show
variations and how they differ, it's a good idea to change
only one thing at a time. The filename differences are not
relevant to what we're trying to show.

Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agodocs/config: avoid the term "expand" for includes
Jeff King [Thu, 11 May 2017 09:13:04 +0000 (05:13 -0400)] 
docs/config: avoid the term "expand" for includes

Using the word "expand" to refer to including the contents
of another config file isn't really accurate, since it's a
verbatim insertion. And it can cause confusion with the
expanding of the path itself via things like "~".

Let's clarify when we are referring to the contents versus
the filename, and use appropriate verbs in each case.

Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agodocs/config: give a relative includeIf example
Jeff King [Thu, 11 May 2017 09:11:06 +0000 (05:11 -0400)] 
docs/config: give a relative includeIf example

The changes in the previous commit hopefully clarify that
the evaluation of an include "path" variable is the same no
matter if it's in a conditional section or not. But since
this question came up on the list, let's add an example that
makes it obvious.

Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agodocs/config: clarify include/includeIf relationship
Jeff King [Thu, 11 May 2017 09:10:47 +0000 (05:10 -0400)] 
docs/config: clarify include/includeIf relationship

The "includeIf" directives behave exactly like include ones,
except they only kick in when the conditional is true. That
was mentioned in the "conditional" section, but let's make
it more clear for the whole "includes" section, since people
don't necessarily read the documentation top to bottom.

Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agotravis-ci: add job to run tests with GETTEXT_POISON
Lars Schneider [Fri, 5 May 2017 15:40:53 +0000 (17:40 +0200)] 
travis-ci: add job to run tests with GETTEXT_POISON

Add a job to run Git tests with GETTEXT_POISON. In this job we don't run
the git-p4, git-svn, and HTTPD tests to save resources/time (those tests
are already executed in other jobs). Since we don't run these tests, we
can also skip the "before_install" step (which would install the
necessary dependencies) with an empty override.

Signed-off-by: Lars Schneider <larsxschneider@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agotravis-ci: setup "prove cache" in "script" step
Lars Schneider [Fri, 5 May 2017 15:40:52 +0000 (17:40 +0200)] 
travis-ci: setup "prove cache" in "script" step

The command that made the "prove cache" persistent across builds was
executed in the "before_install" step. Consequently, every job that
wanted to make use of the cache had to run this step.

The "prove cache" is only used in the "script" step for the
"make test" command. Therefore, we should configure the "prove cache"
in this step.

This change is useful for a subsequent patch that adds a job which does
not need the "before_install" step but wants to run the "script" step to
execute the tests.

Signed-off-by: Lars Schneider <larsxschneider@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agotests: fix tests broken under GETTEXT_POISON=YesPlease
Ævar Arnfjörð Bjarmason [Fri, 5 May 2017 18:19:32 +0000 (18:19 +0000)] 
tests: fix tests broken under GETTEXT_POISON=YesPlease

The GETTEXT_POISON=YesPlease compile-time testing option added in my
bb946bba76 ("i18n: add GETTEXT_POISON to simulate unfriendly
translator", 2011-02-22) has been slowly bitrotting as strings have
been marked for translation, and new tests have been added without
running it.

I brought this up on the list ("[BUG] test suite broken with
GETTEXT_POISON=YesPlease", [1]) asking whether this mode was useful at
all anymore. At least one person occasionally uses it, and Lars
Schneider offered to change one of the the Travis builds to run in
this mode, so fix up the failing ones.

My test setup runs most of the tests, with the notable exception of
skipping all the p4 tests, so it's possible that there's still some
lurking regressions I haven't fixed.

1. <CACBZZX62+acvi1dpkknadTL827mtCm_QesGSZ=6+UnyeMpg8+Q@mail.gmail.com>

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoread-tree: "read-tree -m --empty" does not make sense
Junio C Hamano [Thu, 11 May 2017 04:31:54 +0000 (21:31 -0700)] 
read-tree: "read-tree -m --empty" does not make sense

fb1bb965 ("read-tree: deprecate syntax without tree-ish args",
2010-09-10) wanted to deprecate "git read-tree" without any tree,
which used to be the way to empty the index, and encourage use of
"git read-tree --empty" instead.

However, when used with "-m", "--empty" does not make any sense,
either, simply because merging 0 trees will result in a different
error anyway.

Omit the deprecation warning and let the code to emit real error
message diagnose the error.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agot4051: mark supporting files as requiring LF-only line endings
Johannes Schindelin [Tue, 9 May 2017 12:54:32 +0000 (14:54 +0200)] 
t4051: mark supporting files as requiring LF-only line endings

The test t4051-diff-function-context.sh passes on Linux when
core.autocrlf=true even without marking its support files as LF-only,
but they fail when core.autocrlf=true in Git for Windows' SDK.

The reason is that `grep ... >file.c.new` will keep CR/LF line endings
on Linux (obviously treating CRs as if they were regular characters),
but will be converted to LF-only line endings with MSYS2's grep that is
used in Git for Windows.

As we do not want to validate the way the available `grep` works, let's
just mark the input as LF-only and move on.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoFix the remaining tests that failed with core.autocrlf=true
Johannes Schindelin [Tue, 9 May 2017 12:54:27 +0000 (14:54 +0200)] 
Fix the remaining tests that failed with core.autocrlf=true

The test suite is mainly developed on Linux and MacOSX, which is the
reason that nobody thought to mark files as LF-only as needed.

The symptom is a test suite that fails left and right when being checked
out using Git for Windows (which defaults to core.autocrlf=true).

Mostly, the problems stem from Git's (LF-only) output being compared to
hard-coded files that are checked out with line endings according to
core.autocrlf (which is of course incorrect). This includes the two test
files in t/diff-lib/, README and COPYING.

This patch can be validated even on Linux by using this cadence:

git config core.autocrlf true
rm .git/index && git stash
make -j15 DEVELOPER=1 test

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agot3901: move supporting files into t/t3901/
Johannes Schindelin [Tue, 9 May 2017 12:54:24 +0000 (14:54 +0200)] 
t3901: move supporting files into t/t3901/

The current convention is to either generate files on the fly in tests,
or to use supporting files taken from a t/tNNNN/ directory (where NNNN
matches the test's number, or the number of the test from which we
borrow supporting files).

The test t3901-i18n-patch.sh was obviously introduced before that
convention was in full swing, hence its supporting files still lived in
t/t3901-8859-1.txt and t/t3901-utf8.txt, respectively.

Let's adjust to the current convention.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agocompletion: mark bash script as LF-only
Johannes Schindelin [Tue, 9 May 2017 12:54:20 +0000 (14:54 +0200)] 
completion: mark bash script as LF-only

Without this change, the completion script does not work, as Bash expects
its scripts to have line feeds as end-of-line markers (this is
particularly prominent in quoted multi-line strings, where carriage
returns would slip into the strings as verbatim characters otherwise).

This change is required to let t9902-completion pass when Git's source
code is checked out with `core.autocrlf = true`.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agogit-new-workdir: mark script as LF-only
Johannes Schindelin [Tue, 9 May 2017 12:53:25 +0000 (14:53 +0200)] 
git-new-workdir: mark script as LF-only

Bash does not handle scripts with CR/LF line endings correctly, therefore
they *have* to be forced to LF-only line endings.

Funnily enough, this fixes t3000-ls-files-others and
t1021-rerere-in-workdir when git.git was checked out with
core.autocrlf=true, as these test still use git-new-workdir (once `git
worktree` is no longer marked as experimental, both scripts probably
want to be ported to using that command instead).

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoFix build with core.autocrlf=true
Johannes Schindelin [Tue, 9 May 2017 12:53:21 +0000 (14:53 +0200)] 
Fix build with core.autocrlf=true

On Windows, the default line endings are denoted by a Carriage Return
byte followed by a Line Feed byte, while Linux and MacOSX use a single
Line Feed byte to denote a line ending.

To help with this situation, Git introduced several mechanisms over the
last decade, most prominently the `core.autocrlf` setting.

Sometimes, however, a single setting is incorrect, e.g. when certain
files in the source code are to be consumed by software that can handle
only LF line endings, while other files can use whatever is appropriate
for the current platform.

To allow for that, Git added the `eol` option to its .gitattributes
handling, expecting every user of Git to mark their source code
appropriately.

Bash assumes that line-endings of scripts are denoted by a single Line
Feed byte. Therefore, shell scripts in Git's source code are one example
where that `eol=lf` option is *required*.

When generating common-cmds.h, the Unix tools we use generally operate on
the assumption that input and output deliminate their lines using LF-only
line endings. Consequently, they would happily copy the CR byte verbatim
into the strings in common-cmds.h, which in turn makes the C preprocessor
barf (that interprets them as MacOS-style line endings). Therefore, we
have to mark the input files as LF-only: command-list.txt and
Documentation/git-*.txt.

Quite a bit belatedly, this patch brings Git's own source code in line
with those expectations by setting those attributes to allow for a
correct build even when core.autocrlf=true.

This patch can be validated even on Linux, by using this cadence:

git config core.autocrlf true
rm .git/index && git stash
make -j15 DEVELOPER=1

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoreceive-pack: verify push options in cert
Jonathan Tan [Tue, 9 May 2017 19:23:53 +0000 (12:23 -0700)] 
receive-pack: verify push options in cert

In commit f6a4e61 ("push: accept push options", 2016-07-14), send-pack
was taught to include push options both within the signed cert (if the
push is a signed push) and outside the signed cert; however,
receive-pack ignores push options within the cert, only handling push
options outside the cert.

Teach receive-pack, in the case that push options are provided for a
signed push, to verify that the push options both within the cert and
outside the cert are consistent.

This sets in stone the requirement that send-pack redundantly send its
push options in 2 places, but I think that this is better than the
alternatives. Sending push options only within the cert is
backwards-incompatible with existing Git servers (which read push
options only from outside the cert), and sending push options only
outside the cert means that the push options are not signed for.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agodoc: replace more gmane links
Junio C Hamano [Mon, 8 May 2017 01:38:59 +0000 (10:38 +0900)] 
doc: replace more gmane links

Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agocheckout: fix memory leak
Junio C Hamano [Mon, 8 May 2017 04:21:06 +0000 (13:21 +0900)] 
checkout: fix memory leak

When "git checkout -m" does an in-core three-way merge to carry
local modifications forward to check out a different branch, the
code forgot to free the updated contents it has in-core.

Noticed-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoGit 2.13 v2.13.0
Junio C Hamano [Tue, 9 May 2017 14:26:02 +0000 (23:26 +0900)] 
Git 2.13

Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoMerge tag 'l10n-2.13.0-rnd2.1' of git://github.com/git-l10n/git-po
Junio C Hamano [Tue, 9 May 2017 14:25:26 +0000 (23:25 +0900)] 
Merge tag 'l10n-2.13.0-rnd2.1' of git://github.com/git-l10n/git-po

l10n for Git 2.13.0 round 2.1

* tag 'l10n-2.13.0-rnd2.1' of git://github.com/git-l10n/git-po:
  l10n: zh_CN: for git v2.13.0 l10n round 2
  l10n: sv.po: Update Swedish translation (3195t0f0u)
  l10n: zh_CN: review for git v2.13.0 l10n round 1
  l10n: Update Catalan translation
  l10n: bg.po: Updated Bulgarian translation (3195t)
  l10n: fr.po v2.13 rnd 2
  l10n: de.po: translate 4 new messages
  l10n: de.po: update German translation
  l10n: de.po: lower case after semi-colon
  l10n: vi.po(3195t): Update translation for v2.13.0 round 2
  l10n: git.pot: v2.13.0 round 2 (4 new, 7 removed)
  l10n: zh_CN: for git v2.13.0 l10n round 1
  l10n: fr.po v2.13 round 1
  l10n: pt_PT: update Portuguese translation
  l10n: bg.po: Updated Bulgarian translation (3201t)
  l10n: vi.po(3198t): Updated Vietnamese translation for v2.13.0-rc0
  l10n: sv.po: Update Swedish translation (3199t0f0u)
  l10n: git.pot: v2.13.0 round 1 (96 new, 37 removed)

7 years agoMerge branch 'master' of git://github.com/nafmo/git-l10n-sv
Jiang Xin [Tue, 9 May 2017 14:12:34 +0000 (22:12 +0800)] 
Merge branch 'master' of git://github.com/nafmo/git-l10n-sv

* 'master' of git://github.com/nafmo/git-l10n-sv:
  l10n: sv.po: Update Swedish translation (3195t0f0u)

7 years agol10n: zh_CN: for git v2.13.0 l10n round 2
Jiang Xin [Tue, 9 May 2017 13:55:38 +0000 (21:55 +0800)] 
l10n: zh_CN: for git v2.13.0 l10n round 2

Translate 4 messages (3195t0f0u) for git v2.13.0-rc2.

Signed-off-by: Jiang Xin <worldhello.net@gmail.com>
7 years agol10n: sv.po: Update Swedish translation (3195t0f0u)
Peter Krefting [Tue, 9 May 2017 07:05:09 +0000 (08:05 +0100)] 
l10n: sv.po: Update Swedish translation (3195t0f0u)

Signed-off-by: Peter Krefting <peter@softwolves.pp.se>
7 years agoSync with v2.12.3
Junio C Hamano [Tue, 9 May 2017 03:20:21 +0000 (20:20 -0700)] 
Sync with v2.12.3

Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoMerge branch 'jh/verify-index-checksum-only-in-fsck'
Junio C Hamano [Tue, 9 May 2017 03:17:42 +0000 (12:17 +0900)] 
Merge branch 'jh/verify-index-checksum-only-in-fsck'

* jh/verify-index-checksum-only-in-fsck:
  t1450: avoid use of "sed" on the index, which is a binary file

7 years agodocs: correct receive.advertisePushOptions default
Jonathan Tan [Mon, 8 May 2017 21:33:50 +0000 (14:33 -0700)] 
docs: correct receive.advertisePushOptions default

In commit c714e45 ("receive-pack: implement advertising and receiving
push options", 2016-07-14), receive-pack was taught to (among other
things) advertise that it understood push options, depending on
configuration. It was documented that it advertised such ability by
default; however, it actually does not. (In that commit, notice that
advertise_push_options defaults to 0, unlike advertise_atomic_push which
defaults to 1.)

Update the documentation to state that it does not advertise the ability
by default.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agot5310: fix "; do" style
Jeff King [Tue, 9 May 2017 02:59:46 +0000 (22:59 -0400)] 
t5310: fix "; do" style

Our usual shell style is to put the "do" of a loop on its
own line, like:

  while $cond
  do
          something
  done

instead of:

  while $cond; do
          something
  done

We have a bit of both in our code base, but the former is
what's in CodingGuidelines (and outnumbers the latter in t/
by about 6:1).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agopack-objects: disable pack reuse for object-selection options
Jeff King [Tue, 9 May 2017 02:54:13 +0000 (22:54 -0400)] 
pack-objects: disable pack reuse for object-selection options

If certain options like --honor-pack-keep, --local, or
--incremental are used with pack-objects, then we need to
feed each potential object to want_object_in_pack() to see
if it should be filtered out. But when the bitmap
reuse_packfile optimization is in effect, we do not call
that function at all, and in fact skip adding the objects to
the to_pack list entirely.  This means we have a bug: for
certain requests we will silently ignore those options and
include objects in that pack that should not be there.

The problem has been present since the inception of the
pack-reuse code in 6b8fda2db (pack-objects: use bitmaps when
packing objects, 2013-12-21), but it was unlikely to come up
in practice.  These options are generally used for on-disk
packing, not transfer packs (which go to stdout), but we've
never allowed pack reuse for non-stdout packs (until
645c432d6, we did not even use bitmaps, which the reuse
optimization relies on; after that, we explicitly turned it
off when not packing to stdout).

We can fix this by just disabling the reuse_packfile
optimization when the options are in use. In theory we could
teach the pack-reuse code to satisfy these checks, but it's
not worth the complexity. The purpose of the optimization is
to keep the amount of per-object work we do to a minimum.
But these options inherently require us to search for other
copies of each object, drowning out any benefit of the
pack-reuse optimization. But note that the optimizations
from 56dfeb626 (pack-objects: compute local/ignore_pack_keep
early, 2016-07-29) happen before pack-reuse, meaning that
specifying "--honor-pack-keep" in a repository with no .keep
files can still follow the fast path.

There are tests in t5310 that check these options with
bitmaps and --stdout, but they didn't catch the bug, and
it's hard to adapt them to do so.

One problem is that they don't use --delta-base-offset;
without that option, we always disable the reuse
optimization entirely. It would be fine to add it in (it
actually makes the test more realistic), but that still
isn't quite enough.

The other problem is that the reuse code is very picky; it
only kicks in when it can reuse most of a pack, starting
from the first byte. So we'd have to start from a fully
repacked and bitmapped state to trigger it. But the tests
for these options use a much more subtle state; they want to
be sure that the want_object_in_pack() code is allowing some
objects but not others. Doing a full repack runs counter to
that.

So this patch adds new tests at the end of the script which
create the fully-packed state and make sure that each option
is not fooled by reusable pack.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoapply.c: fix whitespace-only mismerge
Junio C Hamano [Tue, 9 May 2017 02:30:24 +0000 (19:30 -0700)] 
apply.c: fix whitespace-only mismerge

4af9a7d3 ("Merge branch 'bc/object-id'", 2016-09-19) involved
merging a lot of changes made to builtin/apply.c on the side branch
manually to apply.c as an intervening commit 13b5af22 ("apply: move
libified code from builtin/apply.c to apply.{c,h}", 2016-04-22)
moved a lot of the lines changed on the side branch to a different
file apply.c at the top-level, requiring manual patching of it.
Apparently, the maintainer screwed up and made the code indent in a
funny way while doing so.

Reported-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agol10n: zh_CN: review for git v2.13.0 l10n round 1
Ray Chen [Tue, 2 May 2017 15:42:43 +0000 (23:42 +0800)] 
l10n: zh_CN: review for git v2.13.0 l10n round 1

Signed-off-by: Ray Chen <oldsharp@gmail.com>
7 years agoMerge branch 'master' of https://github.com/vnwildman/git
Jiang Xin [Mon, 8 May 2017 22:39:31 +0000 (06:39 +0800)] 
Merge branch 'master' of https://github.com/vnwildman/git

* 'master' of https://github.com/vnwildman/git:
  l10n: vi.po(3195t): Update translation for v2.13.0 round 2

7 years agol10n: Update Catalan translation
Jordi Mas [Sun, 7 May 2017 08:12:01 +0000 (10:12 +0200)] 
l10n: Update Catalan translation

Signed-off-by: Jordi Mas <jmas@softcatala.org>
7 years agol10n: bg.po: Updated Bulgarian translation (3195t)
Alexander Shopov [Sun, 7 May 2017 07:25:19 +0000 (09:25 +0200)] 
l10n: bg.po: Updated Bulgarian translation (3195t)

Signed-off-by: Alexander Shopov <ash@kambanaria.org>
7 years agoMerge branch 'fr_l10n_v2.13_rnd2' of git://github.com/jnavila/git
Jiang Xin [Mon, 8 May 2017 22:18:53 +0000 (06:18 +0800)] 
Merge branch 'fr_l10n_v2.13_rnd2' of git://github.com/jnavila/git

* 'fr_l10n_v2.13_rnd2' of git://github.com/jnavila/git:
  l10n: fr.po v2.13 rnd 2

7 years agosubmodule_uses_worktrees(): plug memory leak
Johannes Schindelin [Thu, 4 May 2017 13:59:19 +0000 (15:59 +0200)] 
submodule_uses_worktrees(): plug memory leak

There is really no reason why we would need to hold onto the allocated
string longer than necessary.

Reported by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoshow_worktree(): plug memory leak
Johannes Schindelin [Thu, 4 May 2017 13:59:13 +0000 (15:59 +0200)] 
show_worktree(): plug memory leak

The buffer allocated by shorten_unambiguous_ref() needs to be released.

Discovered by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoname-rev: avoid leaking memory in the `deref` case
Johannes Schindelin [Thu, 4 May 2017 13:59:06 +0000 (15:59 +0200)] 
name-rev: avoid leaking memory in the `deref` case

When the `name_rev()` function is asked to dereference the tip name, it
allocates memory. But when it turns out that another tip already
described the commit better than the current one, we forgot to release
the memory.

Pointed out by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoremote: plug memory leak in match_explicit()
Johannes Schindelin [Thu, 4 May 2017 13:59:01 +0000 (15:59 +0200)] 
remote: plug memory leak in match_explicit()

The `guess_ref()` returns an allocated buffer of which `make_linked_ref()`
does not take custody (`alloc_ref()` makes a copy), therefore we need to
release the buffer afterwards.

Noticed via Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoadd_reflog_for_walk: avoid memory leak
Johannes Schindelin [Thu, 4 May 2017 13:58:42 +0000 (15:58 +0200)] 
add_reflog_for_walk: avoid memory leak

We free()d the `log` buffer when dwim_log() returned 1, but not when it
returned a larger value (which meant that it still allocated the buffer
but we simply ignored it).

While in the vicinity, make sure that the `reflogs` structure as well as
the `branch` variable are released properly, too.

Identified by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoshallow: avoid memory leak
Johannes Schindelin [Thu, 4 May 2017 13:58:35 +0000 (15:58 +0200)] 
shallow: avoid memory leak

Reported by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoline-log: avoid memory leak
Johannes Schindelin [Thu, 4 May 2017 13:58:01 +0000 (15:58 +0200)] 
line-log: avoid memory leak

Discovered by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoreceive-pack: plug memory leak in update()
Johannes Schindelin [Thu, 4 May 2017 13:57:55 +0000 (15:57 +0200)] 
receive-pack: plug memory leak in update()

Reported via Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agofast-export: avoid leaking memory in handle_tag()
Johannes Schindelin [Thu, 4 May 2017 13:57:33 +0000 (15:57 +0200)] 
fast-export: avoid leaking memory in handle_tag()

Reported by, you guessed it, Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agomktree: plug memory leaks reported by Coverity
Johannes Schindelin [Thu, 4 May 2017 13:57:28 +0000 (15:57 +0200)] 
mktree: plug memory leaks reported by Coverity

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agopack-redundant: plug memory leak
Johannes Schindelin [Thu, 4 May 2017 13:56:54 +0000 (15:56 +0200)] 
pack-redundant: plug memory leak

Identified via Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agosetup_discovered_git_dir(): plug memory leak
Johannes Schindelin [Thu, 4 May 2017 13:56:51 +0000 (15:56 +0200)] 
setup_discovered_git_dir(): plug memory leak

The setup_explicit_git_dir() function does not take custody of the string
passed as first parameter; we have to release it if we turned the value of
git_dir into an absolute path.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agosetup_bare_git_dir(): help static analysis
Johannes Schindelin [Thu, 4 May 2017 13:56:47 +0000 (15:56 +0200)] 
setup_bare_git_dir(): help static analysis

Coverity reported a memory leak in this function. However, it can only
be called once, as setup_git_directory() changes global state and hence
is not reentrant.

Mark the variable as static to indicate that this is a singleton.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agosplit_commit_in_progress(): simplify & fix memory leak
Johannes Schindelin [Thu, 4 May 2017 13:56:44 +0000 (15:56 +0200)] 
split_commit_in_progress(): simplify & fix memory leak

This function did a whole lot of unnecessary work, such as reading in
four files just to figure out that, oh, hey, we do not need to look at
them after all because the HEAD is not detached.

Simplify the entire function to return early when possible, to read in
the files only when necessary, and to release the allocated memory
always (there was a leak, reported via Coverity, where we failed to
release the allocated strings if the HEAD is not detached).

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agocheckout: fix memory leak
Johannes Schindelin [Thu, 4 May 2017 13:56:40 +0000 (15:56 +0200)] 
checkout: fix memory leak

This change addresses part of the NEEDSWORK comment above the code,
therefore the comment needs to be adjusted, too.

Discovered via Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agocat-file: fix memory leak
Johannes Schindelin [Thu, 4 May 2017 13:56:17 +0000 (15:56 +0200)] 
cat-file: fix memory leak

Discovered by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agomailinfo & mailsplit: check for EOF while parsing
Johannes Schindelin [Thu, 4 May 2017 13:56:14 +0000 (15:56 +0200)] 
mailinfo & mailsplit: check for EOF while parsing

While POSIX states that it is okay to pass EOF to isspace() (and it seems
to be implied that EOF should *not* be treated as whitespace), and also to
pass EOF to ungetc() (which seems to be intended to fail without buffering
the character), it is much better to handle these cases explicitly. Not
only does it reduce head-scratching (and helps static analysis avoid
reporting false positives), it also lets us handle files containing
nothing but whitespace by erroring out.

Reported via Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agostatus: close file descriptor after reading git-rebase-todo
Johannes Schindelin [Thu, 4 May 2017 13:55:52 +0000 (15:55 +0200)] 
status: close file descriptor after reading git-rebase-todo

Reported via Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agodifftool: address a couple of resource/memory leaks
Johannes Schindelin [Thu, 4 May 2017 13:55:48 +0000 (15:55 +0200)] 
difftool: address a couple of resource/memory leaks

This change plugs a couple of memory leaks and makes sure that the file
descriptor is closed in run_dir_diff().

Spotted by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoget_mail_commit_oid(): avoid resource leak
Johannes Schindelin [Thu, 4 May 2017 13:55:45 +0000 (15:55 +0200)] 
get_mail_commit_oid(): avoid resource leak

When we fail to read, or parse, the file, we still want to close the file
descriptor and release the strbuf.

Reported via Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agogit_config_rename_section_in_file(): avoid resource leak
Johannes Schindelin [Thu, 4 May 2017 13:55:41 +0000 (15:55 +0200)] 
git_config_rename_section_in_file(): avoid resource leak

In case of errors, we really want the file descriptor to be closed.

Discovered by a Coverity scan.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agoadd_commit_patch_id(): avoid allocating memory unnecessarily
Johannes Schindelin [Thu, 4 May 2017 13:55:38 +0000 (15:55 +0200)] 
add_commit_patch_id(): avoid allocating memory unnecessarily

It would appear that we allocate (and forget to release) memory if the
patch ID is not even defined.

Reported by the Coverity tool.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agowinansi: avoid buffer overrun
Johannes Schindelin [Thu, 4 May 2017 13:55:34 +0000 (15:55 +0200)] 
winansi: avoid buffer overrun

When we could not convert the UTF-8 sequence into Unicode for writing to
the Console, we should not try to write an insanely-long sequence of
invalid wide characters (mistaking the negative return value for an
unsigned length).

Reported by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agowinansi: avoid use of uninitialized value
Johannes Schindelin [Thu, 4 May 2017 13:55:29 +0000 (15:55 +0200)] 
winansi: avoid use of uninitialized value

To initialize the foreground color attributes of "plain text", our ANSI
emulation tries to infer them from the currently attached console while
running the is_console() function. This function first tries to detect any
console attached to stdout, then it is called with stderr.

If neither stdout nor stderr has any console attached, it does not
actually matter what we use for "plain text" attributes, as we never need
to output any text to any console in that case.

However, after working on stdout and stderr, is_console() is called with
stdin, and it still tries to initialize the "plain text" attributes if
they had not been initialized earlier. In this case, we cannot detect any
attributes, and we used an uninitialized value for them.

Naturally, Coverity complained about this use case because it could not
reason about the code deeply enough to figure out that we do not even use
those attributes in that case.

Let's just initialize the value to 0 in that case, both to avoid future
Coverity reports, and to help catch future regressions in case anybody
changes the order of the is_console() calls (which would make the text
black on black).

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agomingw: avoid memory leak when splitting PATH
Johannes Schindelin [Thu, 4 May 2017 13:55:24 +0000 (15:55 +0200)] 
mingw: avoid memory leak when splitting PATH

In the (admittedly, concocted) case that PATH consists only of path
delimiters, we would leak the duplicated string.

Reported by Coverity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agodoc: replace a couple of broken gmane links
Ævar Arnfjörð Bjarmason [Fri, 5 May 2017 10:08:03 +0000 (10:08 +0000)] 
doc: replace a couple of broken gmane links

Replace a couple of broken links to gmane with links to other
archives. See commit 54471fdcc3 ("README: replace gmane link with
public-inbox", 2016-12-15) for prior art.

With this change there's still 4 references left in the code:

    $ git grep -E '(article|thread)\.gmane.org' -- |grep -v RelNotes|wc -l
    4

I couldn't find alternative links for those.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
7 years agol10n: fr.po v2.13 rnd 2
Jean-Noel Avila [Fri, 5 May 2017 09:55:14 +0000 (11:55 +0200)] 
l10n: fr.po v2.13 rnd 2

Signed-off-by: Jean-Noel Avila <jn.avila@free.fr>
7 years agol10n: de.po: translate 4 new messages
Ralf Thielow [Fri, 5 May 2017 09:23:53 +0000 (11:23 +0200)] 
l10n: de.po: translate 4 new messages

Translate 4 new messages came from git.pot update in 28e1aaa48 (l10n:
git.pot: v2.13.0 round 2 (4 new, 7 removed)).

Signed-off-by: Ralf Thielow <ralf.thielow@gmail.com>
Acked-by: Matthias Rüster <matthias.ruester@gmail.com>
7 years agol10n: de.po: update German translation
Ralf Thielow [Thu, 13 Apr 2017 14:45:29 +0000 (16:45 +0200)] 
l10n: de.po: update German translation

Translate 96 new messages came from git.pot update in dfc182b (l10n:
git.pot: v2.13.0 round 1 (96 new, 37 removed)).

Signed-off-by: Ralf Thielow <ralf.thielow@gmail.com>
Acked-by: Matthias Rüster <matthias.ruester@gmail.com>
7 years agol10n: de.po: lower case after semi-colon
Michael J Gruber [Fri, 17 Mar 2017 15:04:19 +0000 (16:04 +0100)] 
l10n: de.po: lower case after semi-colon

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Ralf Thielow <ralf.thielow@gmail.com>
7 years agol10n: vi.po(3195t): Update translation for v2.13.0 round 2
Tran Ngoc Quan [Fri, 5 May 2017 06:41:32 +0000 (13:41 +0700)] 
l10n: vi.po(3195t): Update translation for v2.13.0 round 2

Signed-off-by: Tran Ngoc Quan <vnwildman@gmail.com>
7 years agoGit 2.12.3 v2.12.3
Junio C Hamano [Fri, 5 May 2017 04:33:22 +0000 (13:33 +0900)] 
Git 2.12.3

Signed-off-by: Junio C Hamano <gitster@pobox.com>