OSDN Git Service

git-core/git.git
8 years agoMerge branch 'mj/pull-rebase-autostash'
Junio C Hamano [Wed, 13 Apr 2016 21:12:36 +0000 (14:12 -0700)]
Merge branch 'mj/pull-rebase-autostash'

"git pull --rebase" learned "--[no-]autostash" option, so that
the rebase.autostash configuration variable set to true can be
overridden from the command line.

* mj/pull-rebase-autostash:
  t5520: test --[no-]autostash with pull.rebase=true
  t5520: reduce commom lines of code
  t5520: factor out common "failing autostash" code
  t5520: factor out common "successful autostash" code
  t5520: use better test to check stderr output
  t5520: ensure consistent test conditions
  t5520: use consistent capitalization in test titles
  pull --rebase: add --[no-]autostash flag
  git-pull.c: introduce git_pull_config()

8 years agoMerge branch 'jn/mergetools-examdiff'
Junio C Hamano [Wed, 13 Apr 2016 21:12:36 +0000 (14:12 -0700)]
Merge branch 'jn/mergetools-examdiff'

"git mergetools" learned to drive ExamDiff.

* jn/mergetools-examdiff:
  mergetools: add support for ExamDiff
  mergetools: create mergetool_find_win32_cmd() helper function for winmerge

8 years agoMerge branch 'es/format-patch-doc-hide-no-patch'
Junio C Hamano [Wed, 13 Apr 2016 21:12:35 +0000 (14:12 -0700)]
Merge branch 'es/format-patch-doc-hide-no-patch'

"git format-patch --help" showed `-s` and `--no-patch` as if these
are valid options to the command.  We already hide `--patch` option
from the documentation, because format-patch is about showing the
diff, and the documentation now hides these options as well.

* es/format-patch-doc-hide-no-patch:
  git-format-patch.txt: don't show -s as shorthand for multiple options

8 years agoMerge branch 'js/mingw-tests-2.8'
Junio C Hamano [Wed, 13 Apr 2016 21:12:34 +0000 (14:12 -0700)]
Merge branch 'js/mingw-tests-2.8'

Code clean-up.

* js/mingw-tests-2.8:
  Windows: shorten code by re-using convert_slashes()

8 years agoMerge branch 'cc/apply'
Junio C Hamano [Wed, 13 Apr 2016 21:12:34 +0000 (14:12 -0700)]
Merge branch 'cc/apply'

Minor code clean-up.

* cc/apply:
  builtin/apply: free patch when parse_chunk() fails
  builtin/apply: handle parse_binary() failure
  apply: remove unused call to free() in gitdiff_{old,new}name()
  builtin/apply: get rid of useless 'name' variable

8 years agoMerge branch 'sb/misc-cleanups'
Junio C Hamano [Wed, 13 Apr 2016 21:12:34 +0000 (14:12 -0700)]
Merge branch 'sb/misc-cleanups'

Assorted minor clean-ups.

* sb/misc-cleanups:
  credential-cache, send_request: close fd when done
  bundle: don't leak an fd in case of early return
  abbrev_sha1_in_line: don't leak memory
  notes: don't leak memory in git_config_get_notes_strategy

8 years agoMerge branch 'sk/send-pack-all-fix'
Junio C Hamano [Wed, 13 Apr 2016 21:12:33 +0000 (14:12 -0700)]
Merge branch 'sk/send-pack-all-fix'

"git send-pack --all <there>" was broken when its command line
option parsing was written in the 2.6 timeframe.

* sk/send-pack-all-fix:
  git-send-pack: fix --all option when used with directory

8 years agoMerge branch 'sg/diff-multiple-identical-renames'
Junio C Hamano [Wed, 13 Apr 2016 21:12:32 +0000 (14:12 -0700)]
Merge branch 'sg/diff-multiple-identical-renames'

"git diff -M" used to work better when two originally identical
files A and B got renamed to X/A and X/B by pairing A to X/A and B
to X/B, but this was broken in the 2.0 timeframe.

* sg/diff-multiple-identical-renames:
  diffcore: fix iteration order of identical files during rename detection

8 years agoMerge branch 'kn/for-each-tag-branch'
Junio C Hamano [Wed, 13 Apr 2016 21:12:31 +0000 (14:12 -0700)]
Merge branch 'kn/for-each-tag-branch'

A minor documentation update.

* kn/for-each-tag-branch:
  for-each-ref: fix description of '--contains' in manpage

8 years agoMerge branch 'ky/branch-d-worktree'
Junio C Hamano [Wed, 13 Apr 2016 21:12:30 +0000 (14:12 -0700)]
Merge branch 'ky/branch-d-worktree'

When "git worktree" feature is in use, "git branch -d" allowed
deletion of a branch that is checked out in another worktree

* ky/branch-d-worktree:
  branch -d: refuse deleting a branch which is currently checked out

8 years agoMerge branch 'rz/worktree-no-checkout'
Junio C Hamano [Wed, 13 Apr 2016 21:12:30 +0000 (14:12 -0700)]
Merge branch 'rz/worktree-no-checkout'

"git worktree add" can be given "--no-checkout" option to only
create an empty worktree without checking out the files.

* rz/worktree-no-checkout:
  worktree: add: introduce --checkout option

8 years agoMerge branch 'rt/rebase-i-shorten-stop-report'
Junio C Hamano [Wed, 13 Apr 2016 21:12:30 +0000 (14:12 -0700)]
Merge branch 'rt/rebase-i-shorten-stop-report'

The commit object name reported when "rebase -i" stops has been
shortened.

* rt/rebase-i-shorten-stop-report:
  rebase-i: print an abbreviated hash when stop for editing

8 years agoMerge branch 'rt/completion-help'
Junio C Hamano [Wed, 13 Apr 2016 21:12:29 +0000 (14:12 -0700)]
Merge branch 'rt/completion-help'

Shell completion (in contrib/) updates.

* rt/completion-help:
  completion: add 'revisions' and 'everyday' to 'git help'
  completion: add option '--guides' to 'git help'

8 years agoMerge branch 'ak/use-hashmap-iter-first-in-submodule-config'
Junio C Hamano [Wed, 13 Apr 2016 21:12:29 +0000 (14:12 -0700)]
Merge branch 'ak/use-hashmap-iter-first-in-submodule-config'

Minor code cleanup.

* ak/use-hashmap-iter-first-in-submodule-config:
  submodule-config: use hashmap_iter_first()

8 years agoMerge branch 'jk/check-repository-format'
Junio C Hamano [Wed, 13 Apr 2016 21:12:28 +0000 (14:12 -0700)]
Merge branch 'jk/check-repository-format'

The repository set-up sequence has been streamlined (the biggest
change is that there is no longer git_config_early()), so that we
do not attempt to look into refs/* when we know we do not have a
Git repository.

* jk/check-repository-format:
  verify_repository_format: mark messages for translation
  setup: drop repository_format_version global
  setup: unify repository version callbacks
  init: use setup.c's repo version verification
  setup: refactor repo format reading and verification
  config: drop git_config_early
  check_repository_format_gently: stop using git_config_early
  lazily load core.sharedrepository
  wrap shared_repository global in get/set accessors
  setup: document check_repository_format()

8 years agoThird batch for post 2.8 cycle
Junio C Hamano [Fri, 8 Apr 2016 21:34:11 +0000 (14:34 -0700)]
Third batch for post 2.8 cycle

Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoMerge branch 'ss/msvc'
Junio C Hamano [Fri, 8 Apr 2016 21:29:13 +0000 (14:29 -0700)]
Merge branch 'ss/msvc'

Build updates for MSVC.

* ss/msvc:
  MSVC: use shipped headers instead of fallback definitions
  MSVC: vsnprintf in Visual Studio 2015 doesn't need SNPRINTF_SIZE_CORR any more

8 years agoMerge branch 'oa/doc-diff-check'
Junio C Hamano [Fri, 8 Apr 2016 21:29:13 +0000 (14:29 -0700)]
Merge branch 'oa/doc-diff-check'

A minor documentation update.

* oa/doc-diff-check:
  Documentation: git diff --check detects conflict markers

8 years agoMerge branch 'pb/opt-cmdmode-doc'
Junio C Hamano [Fri, 8 Apr 2016 21:29:12 +0000 (14:29 -0700)]
Merge branch 'pb/opt-cmdmode-doc'

Minor API documentation update.

* pb/opt-cmdmode-doc:
  api-parse-options.txt: document OPT_CMDMODE()

8 years agoMerge branch 'nd/apply-report-skip'
Junio C Hamano [Fri, 8 Apr 2016 21:29:12 +0000 (14:29 -0700)]
Merge branch 'nd/apply-report-skip'

"git apply -v" learned to report paths in the patch that were
skipped via --include/--exclude mechanism or being outside the
current working directory.

* nd/apply-report-skip:
  apply: report patch skipping in verbose mode

8 years agoMerge branch 'nd/apply-doc'
Junio C Hamano [Fri, 8 Apr 2016 21:29:12 +0000 (14:29 -0700)]
Merge branch 'nd/apply-doc'

A minor documentation update.

* nd/apply-doc:
  git-apply.txt: mention the behavior inside a subdir
  git-apply.txt: remove a space

8 years agoMerge branch 'jc/merge-refuse-new-root'
Junio C Hamano [Fri, 8 Apr 2016 21:29:11 +0000 (14:29 -0700)]
Merge branch 'jc/merge-refuse-new-root'

"git merge" used to allow merging two branches that have no common
base by default, which led to a brand new history of an existing
project created and then get pulled by an unsuspecting maintainer,
which allowed an unnecessary parallel history merged into the
existing project.  The command has been taught not to allow this by
default, with an escape hatch "--allow-unrelated-histories" option
to be used in a rare event that merges histories of two projects
that started their lives independently.

* jc/merge-refuse-new-root:
  merge: refuse to create too cool a merge by default

8 years agoSecond batch for post 2.8 cycle
Junio C Hamano [Wed, 6 Apr 2016 18:44:24 +0000 (11:44 -0700)]
Second batch for post 2.8 cycle

Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoMerge branch 'la/tag-force-signing-annotated-tags'
Junio C Hamano [Wed, 6 Apr 2016 18:39:13 +0000 (11:39 -0700)]
Merge branch 'la/tag-force-signing-annotated-tags'

"git tag" can create an annotated tag without explicitly given an
"-a" (or "-s") option (i.e. when a tag message is given).  A new
configuration variable, tag.forceSignAnnotated, can be used to tell
the command to create signed tag in such a situation.

* la/tag-force-signing-annotated-tags:
  tag: add the option to force signing of annotated tags

8 years agoMerge branch 'jk/submodule-c-credential'
Junio C Hamano [Wed, 6 Apr 2016 18:39:12 +0000 (11:39 -0700)]
Merge branch 'jk/submodule-c-credential'

"git -c credential.<var>=<value> submodule" can now be used to
propagate configuration variables related to credential helper
down to the submodules.

* jk/submodule-c-credential:
  git_config_push_parameter: handle empty GIT_CONFIG_PARAMETERS
  git: submodule honor -c credential.* from command line
  quote: implement sq_quotef()
  submodule: fix segmentation fault in submodule--helper clone
  submodule: fix submodule--helper clone usage
  submodule: check argc count for git submodule--helper clone
  submodule: don't pass empty string arguments to submodule--helper clone

8 years agoMerge branch 'jv/merge-nothing-into-void'
Junio C Hamano [Wed, 6 Apr 2016 18:39:11 +0000 (11:39 -0700)]
Merge branch 'jv/merge-nothing-into-void'

"git merge FETCH_HEAD" dereferenced NULL pointer when merging
nothing into an unborn history (which is arguably unusual usage,
which perhaps was the reason why nobody noticed it).

* jv/merge-nothing-into-void:
  merge: fix NULL pointer dereference when merging nothing into void

8 years agoMerge branch 'ss/commit-squash-msg'
Junio C Hamano [Wed, 6 Apr 2016 18:39:10 +0000 (11:39 -0700)]
Merge branch 'ss/commit-squash-msg'

When "git merge --squash" stopped due to conflict, the concluding
"git commit" failed to read in the SQUASH_MSG that shows the log
messages from all the squashed commits.

* ss/commit-squash-msg:
  commit: do not lose SQUASH_MSG contents

8 years agoMerge branch 'sb/rebase-x'
Junio C Hamano [Wed, 6 Apr 2016 18:39:09 +0000 (11:39 -0700)]
Merge branch 'sb/rebase-x'

"git rebase -x" can be used without passing "-i" option.

* sb/rebase-x:
  t3404: cleanup double empty lines between tests
  rebase: decouple --exec from --interactive

8 years agoMerge branch 'jk/test-httpd-config-nosystem'
Junio C Hamano [Wed, 6 Apr 2016 18:39:08 +0000 (11:39 -0700)]
Merge branch 'jk/test-httpd-config-nosystem'

The tests that involve running httpd leaked the system-wide
configuration in /etc/gitconfig to the tested environment.

* jk/test-httpd-config-nosystem:
  t/lib-httpd: pass through GIT_CONFIG_NOSYSTEM env

8 years agoMerge branch 'jk/send-email-rtrim-mailrc-alias'
Junio C Hamano [Wed, 6 Apr 2016 18:39:07 +0000 (11:39 -0700)]
Merge branch 'jk/send-email-rtrim-mailrc-alias'

"git send-email" had trouble parsing alias file in mailrc format
when lines in it had trailing whitespaces on them.

* jk/send-email-rtrim-mailrc-alias:
  send-email: ignore trailing whitespace in mailrc alias file

8 years agoMerge branch 'jk/credential-cache-comment-exit'
Junio C Hamano [Wed, 6 Apr 2016 18:39:06 +0000 (11:39 -0700)]
Merge branch 'jk/credential-cache-comment-exit'

A code clarification.

* jk/credential-cache-comment-exit:
  credential-cache--daemon: clarify "exit" action semantics

8 years agoMerge branch 'sb/clone-t57-t56'
Junio C Hamano [Wed, 6 Apr 2016 18:39:05 +0000 (11:39 -0700)]
Merge branch 'sb/clone-t57-t56'

Rename bunch of tests on "git clone" for better organization.

* sb/clone-t57-t56:
  clone tests: rename t57* => t56*

8 years agoMerge branch 'ls/p4-map-user'
Junio C Hamano [Wed, 6 Apr 2016 18:39:05 +0000 (11:39 -0700)]
Merge branch 'ls/p4-map-user'

"git p4" now allows P4 author names to be mapped to Git author
names.

* ls/p4-map-user:
  git-p4: map a P4 user to Git author name and email address

8 years agoMerge branch 'cc/doc-recommend-performance-trace-to-file'
Junio C Hamano [Wed, 6 Apr 2016 18:39:04 +0000 (11:39 -0700)]
Merge branch 'cc/doc-recommend-performance-trace-to-file'

A minor documentation update.

* cc/doc-recommend-performance-trace-to-file:
  Documentation: talk about pager in api-trace.txt

8 years agoMerge branch 'pb/t7502-drop-dup'
Junio C Hamano [Wed, 6 Apr 2016 18:39:03 +0000 (11:39 -0700)]
Merge branch 'pb/t7502-drop-dup'

Code clean-up.

* pb/t7502-drop-dup:
  t/t7502 : drop duplicate test

8 years agoMerge branch 'da/mergetool-delete-delete-conflict'
Junio C Hamano [Wed, 6 Apr 2016 18:39:02 +0000 (11:39 -0700)]
Merge branch 'da/mergetool-delete-delete-conflict'

"git mergetool" did not work well with conflicts that both sides
deleted.

* da/mergetool-delete-delete-conflict:
  mergetool: honor tempfile configuration when resolving delete conflicts
  mergetool: support delete/delete conflicts

8 years agoMerge branch 'sb/submodule-parallel-update'
Junio C Hamano [Wed, 6 Apr 2016 18:39:01 +0000 (11:39 -0700)]
Merge branch 'sb/submodule-parallel-update'

A major part of "git submodule update" has been ported to C to take
advantage of the recently added framework to run download tasks in
parallel.

* sb/submodule-parallel-update:
  clone: allow an explicit argument for parallel submodule clones
  submodule update: expose parallelism to the user
  submodule helper: remove double 'fatal: ' prefix
  git submodule update: have a dedicated helper for cloning
  run_processes_parallel: rename parameters for the callbacks
  run_processes_parallel: treat output of children as byte array
  submodule update: direct error message to stderr
  fetching submodules: respect `submodule.fetchJobs` config option
  submodule-config: drop check against NULL
  submodule-config: keep update strategy around

8 years agoMerge branch 'ss/receive-pack-parse-options'
Junio C Hamano [Wed, 6 Apr 2016 18:38:59 +0000 (11:38 -0700)]
Merge branch 'ss/receive-pack-parse-options'

The command line argument parser for "receive-pack" has been
rewritten to use parse-options.

* ss/receive-pack-parse-options:
  builtin/receive-pack.c: use parse_options API

8 years agoMerge branch 'ss/exc-flag-is-a-collection-of-bits'
Junio C Hamano [Wed, 6 Apr 2016 18:38:59 +0000 (11:38 -0700)]
Merge branch 'ss/exc-flag-is-a-collection-of-bits'

Code clean-up.

* ss/exc-flag-is-a-collection-of-bits:
  dir: store EXC_FLAG_* values in unsigned integers

8 years agoWindows: shorten code by re-using convert_slashes()
Johannes Sixt [Sat, 2 Apr 2016 19:03:14 +0000 (21:03 +0200)]
Windows: shorten code by re-using convert_slashes()

Make a few more spots more readable by using the recently introduced,
Windows-specific helper.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agogit-format-patch.txt: don't show -s as shorthand for multiple options
Eric Sunshine [Sun, 27 Mar 2016 21:26:07 +0000 (17:26 -0400)]
git-format-patch.txt: don't show -s as shorthand for multiple options

git-format-patch recognizes -s as shorthand only for --signoff, however,
its documentation shows -s as shorthand for both --signoff and
--no-patch. Resolve this confusion by suppressing the bogus -s shorthand
for --no-patch.

While here, also avoid showing the --no-patch option in git-format-patch
documentation since it doesn't make sense to ask to suppress the patch
while at the same time explicitly asking to format the patch (which,
after all, is the purpose of git-format-patch).

Reported-by: Kevin Brodsky <corax26@gmail.com>
Signed-off-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agot5520: test --[no-]autostash with pull.rebase=true
Mehul Jain [Sat, 2 Apr 2016 17:58:32 +0000 (23:28 +0530)]
t5520: test --[no-]autostash with pull.rebase=true

The "--[no-]autostash" options for git-pull are only valid in
rebase mode (i.e. either --rebase is used or pull.rebase=true).
Existing tests already check the cases when --rebase is used but
fail to check for pull.rebase=true case.

Add two new tests to check that the --[no-]autostash options work
with pull.rebase=true.

Signed-off-by: Mehul Jain <mehul.jain2029@gmail.com>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agot5520: reduce commom lines of code
Mehul Jain [Sat, 2 Apr 2016 17:58:31 +0000 (23:28 +0530)]
t5520: reduce commom lines of code

These two tests are almost similar and thus can be folded in a for-loop.

Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Mehul Jain <mehul.jain2029@gmail.com>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agot5520: factor out common "failing autostash" code
Mehul Jain [Sat, 2 Apr 2016 17:58:30 +0000 (23:28 +0530)]
t5520: factor out common "failing autostash" code

Three tests contains repetitive lines of code.

Factor out common code into test_pull_autostash_fail() and then call it in
these tests.

Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Mehul Jain <mehul.jain2029@gmail.com>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agot5520: factor out common "successful autostash" code
Mehul Jain [Sat, 2 Apr 2016 17:58:29 +0000 (23:28 +0530)]
t5520: factor out common "successful autostash" code

Four tests contains repetitive lines of code.

Factor out common code into test_pull_autostash() and then call it in
these tests.

Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Mehul Jain <mehul.jain2029@gmail.com>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agot5520: use better test to check stderr output
Mehul Jain [Sat, 2 Apr 2016 17:58:28 +0000 (23:28 +0530)]
t5520: use better test to check stderr output

Checking stderr output using test_i18ncmp may lead to test failure as
some shells write trace output to stderr when run under 'set -x'.

Use test_i18ngrep instead of test_i18ncmp.

Signed-off-by: Mehul Jain <mehul.jain2029@gmail.com>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agot5520: ensure consistent test conditions
Mehul Jain [Sat, 2 Apr 2016 17:58:27 +0000 (23:28 +0530)]
t5520: ensure consistent test conditions

Test title says that tests are done with rebase.autostash unset,
but does not take any action to make sure that it is indeed unset.
This may lead to test failure if future changes somehow pollutes
the configuration globally.

Ensure consistent test conditions by explicitly unsetting
rebase.autostash.

Signed-off-by: Mehul Jain <mehul.jain2029@gmail.com>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agot5520: use consistent capitalization in test titles
Mehul Jain [Sat, 2 Apr 2016 17:58:26 +0000 (23:28 +0530)]
t5520: use consistent capitalization in test titles

Signed-off-by: Mehul Jain <mehul.jain2029@gmail.com>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agomergetools: add support for ExamDiff
Jacob Nisnevich [Fri, 25 Mar 2016 23:17:57 +0000 (16:17 -0700)]
mergetools: add support for ExamDiff

Signed-off-by: Jacob Nisnevich <jacob.nisnevich@gmail.com>
Acked-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agomergetools: create mergetool_find_win32_cmd() helper function for winmerge
Jacob Nisnevich [Fri, 25 Mar 2016 23:17:56 +0000 (16:17 -0700)]
mergetools: create mergetool_find_win32_cmd() helper function for winmerge

Signed-off-by: Jacob Nisnevich <jacob.nisnevich@gmail.com>
Acked-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoFirst batch for post 2.8 cycle
Junio C Hamano [Sun, 3 Apr 2016 17:55:36 +0000 (10:55 -0700)]
First batch for post 2.8 cycle

Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoSync with Git 2.8.1
Junio C Hamano [Sun, 3 Apr 2016 17:54:38 +0000 (10:54 -0700)]
Sync with Git 2.8.1

8 years agoMerge branch 'jk/startup-info'
Junio C Hamano [Sun, 3 Apr 2016 17:29:35 +0000 (10:29 -0700)]
Merge branch 'jk/startup-info'

The startup_info data, which records if we are working inside a
repository (among other things), are now uniformly available to Git
subcommand implementations, and Git avoids attempting to touch
references when we are not in a repository.

* jk/startup-info:
  use setup_git_directory() in test-* programs
  grep: turn off gitlink detection for --no-index
  mailmap: do not resolve blobs in a non-repository
  remote: don't resolve HEAD in non-repository
  setup: set startup_info->have_repository more reliably
  setup: make startup_info available everywhere

8 years agoMerge branch 'es/test-gpg-tags'
Junio C Hamano [Sun, 3 Apr 2016 17:29:35 +0000 (10:29 -0700)]
Merge branch 'es/test-gpg-tags'

A test for tags has been restructured so that more parts of it can
easily be run on a platform without a working GnuPG.

* es/test-gpg-tags:
  t6302: skip only signed tags rather than all tests when GPG is missing
  t6302: also test annotated in addition to signed tags
  t6302: normalize names and descriptions of signed tags
  lib-gpg: drop unnecessary "missing GPG" warning

8 years agoMerge branch 'jk/getwholeline-getdelim-empty'
Junio C Hamano [Sun, 3 Apr 2016 17:29:34 +0000 (10:29 -0700)]
Merge branch 'jk/getwholeline-getdelim-empty'

strbuf_getwholeline() did not NUL-terminate the buffer on certain
corner cases in its error codepath.

* jk/getwholeline-getdelim-empty:
  strbuf_getwholeline: NUL-terminate getdelim buffer on error

8 years agoMerge branch 'rj/xdiff-prepare-plug-leak-on-error-codepath'
Junio C Hamano [Sun, 3 Apr 2016 17:29:33 +0000 (10:29 -0700)]
Merge branch 'rj/xdiff-prepare-plug-leak-on-error-codepath'

A small memory leak in an error codepath has been plugged in xdiff
code.

* rj/xdiff-prepare-plug-leak-on-error-codepath:
  xdiff/xprepare: fix a memory leak
  xdiff/xprepare: use the XDF_DIFF_ALG() macro to access flag bits

8 years agoMerge branch 'jc/index-pack'
Junio C Hamano [Sun, 3 Apr 2016 17:29:31 +0000 (10:29 -0700)]
Merge branch 'jc/index-pack'

Code clean-up.

* jc/index-pack:
  index-pack: add a helper function to derive .idx/.keep filename

8 years agoMerge branch 'gf/fetch-pack-direct-object-fetch'
Junio C Hamano [Sun, 3 Apr 2016 17:29:29 +0000 (10:29 -0700)]
Merge branch 'gf/fetch-pack-direct-object-fetch'

Fetching of history by naming a commit object name directly didn't
work across remote-curl transport.

* gf/fetch-pack-direct-object-fetch:
  fetch-pack: update the documentation for "<refs>..." arguments
  fetch-pack: fix object_id of exact sha1

8 years agoMerge branch 'jc/maint-index-pack-keep'
Junio C Hamano [Sun, 3 Apr 2016 17:29:29 +0000 (10:29 -0700)]
Merge branch 'jc/maint-index-pack-keep'

"git index-pack --keep[=<msg>] pack-$name.pack" simply did not work.

* jc/maint-index-pack-keep:
  index-pack: correct --keep[=<msg>]

8 years agoMerge branch 'mm/lockfile-error-message'
Junio C Hamano [Sun, 3 Apr 2016 17:29:27 +0000 (10:29 -0700)]
Merge branch 'mm/lockfile-error-message'

* mm/lockfile-error-message:
  lockfile: improve error message when lockfile exists
  lockfile: mark strings for translation

8 years agoMerge branch 'jk/rev-parse-local-env-vars'
Junio C Hamano [Sun, 3 Apr 2016 17:29:26 +0000 (10:29 -0700)]
Merge branch 'jk/rev-parse-local-env-vars'

The "--local-env-vars" and "--resolve-git-dir" options of "git
rev-parse" failed to work outside a repository when the command's
option parsing was rewritten in 1.8.5 era.

* jk/rev-parse-local-env-vars:
  rev-parse: let some options run outside repository
  t1515: add tests for rev-parse out-of-repo helpers

8 years agoMerge branch 'jk/config-get-urlmatch'
Junio C Hamano [Sun, 3 Apr 2016 17:29:26 +0000 (10:29 -0700)]
Merge branch 'jk/config-get-urlmatch'

"git config --get-urlmatch", unlike other variants of the "git
config --get" family, did not signal error with its exit status
when there was no matching configuration.

* jk/config-get-urlmatch:
  Documentation/git-config: fix --get-all description
  Documentation/git-config: use bulleted list for exit codes
  config: fail if --get-urlmatch finds no value

8 years agoMerge branch 'jk/add-i-highlight'
Junio C Hamano [Sun, 3 Apr 2016 17:29:25 +0000 (10:29 -0700)]
Merge branch 'jk/add-i-highlight'

* jk/add-i-highlight:
  add--interactive: allow custom diff highlighting programs

8 years agoMerge branch 'jk/credential-clear-config'
Junio C Hamano [Sun, 3 Apr 2016 17:29:24 +0000 (10:29 -0700)]
Merge branch 'jk/credential-clear-config'

The credential.helper configuration variable is cumulative and
there is no good way to override it from the command line.  As
a special case, giving an empty string as its value now serves
as the signal to clear the values specified in various files.

* jk/credential-clear-config:
  credential: let empty credential specs reset helper list

8 years agoMerge branch 'mp/upload-pack-use-embedded-args'
Junio C Hamano [Sun, 3 Apr 2016 17:29:23 +0000 (10:29 -0700)]
Merge branch 'mp/upload-pack-use-embedded-args'

The embedded args argv-array in the child process is used to build
the command line to run pack-objects instead of using a separate
array of strings.

* mp/upload-pack-use-embedded-args:
  upload-pack: use argv_array for pack_objects

8 years agoMerge branch 'mm/diff-renames-default'
Junio C Hamano [Sun, 3 Apr 2016 17:29:22 +0000 (10:29 -0700)]
Merge branch 'mm/diff-renames-default'

The end-user facing Porcelain level commands like "diff" and "log"
now enables the rename detection by default.

* mm/diff-renames-default:
  diff: activate diff.renames by default
  log: introduce init_log_defaults()
  t: add tests for diff.renames (true/false/unset)
  t4001-diff-rename: wrap file creations in a test
  Documentation/diff-config: fix description of diff.renames

8 years agoMerge branch 'mm/readme-markdown'
Junio C Hamano [Sun, 3 Apr 2016 17:27:22 +0000 (10:27 -0700)]
Merge branch 'mm/readme-markdown'

Fix a few broken links in README.md and also teach rpmbuild
that there is no README.

* mm/readme-markdown:
  README.md: don't take 'commandname' literally
  git.spec.in: use README.md, not README

8 years agoGit 2.8.1 v2.8.1
Junio C Hamano [Sun, 3 Apr 2016 17:11:35 +0000 (10:11 -0700)]
Git 2.8.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoMerge branch 'mm/readme-markdown' into maint
Junio C Hamano [Sun, 3 Apr 2016 17:13:09 +0000 (10:13 -0700)]
Merge branch 'mm/readme-markdown' into maint

* 'mm/readme-markdown':
  git.spec.in: use README.md, not README

8 years agoREADME.md: don't take 'commandname' literally
Matthieu Moy [Sat, 27 Feb 2016 10:46:12 +0000 (11:46 +0100)]
README.md: don't take 'commandname' literally

The link to Documentation/git-commandname.txt was obviously broken.
Remove the link and make it clear that it is not a literal path name by
using *italics* in makdown.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agogit.spec.in: use README.md, not README
Matthieu Moy [Fri, 1 Apr 2016 07:10:51 +0000 (09:10 +0200)]
git.spec.in: use README.md, not README

The file was renamed in 4ad21f5 (README: use markdown syntax,
2016-02-25), but that commit forgot to update git.spec.in, which
caused the rpmbuild target in the Makefile to fail.

Reported-by: Ron Isaacson <isaacson.ljits@gmail.com>
Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agocredential-cache, send_request: close fd when done
Stefan Beller [Fri, 1 Apr 2016 00:35:46 +0000 (17:35 -0700)]
credential-cache, send_request: close fd when done

No need to keep it open any further.

Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agobundle: don't leak an fd in case of early return
Stefan Beller [Fri, 1 Apr 2016 00:35:45 +0000 (17:35 -0700)]
bundle: don't leak an fd in case of early return

In successful operation `write_pack_data` will close the `bundle_fd`,
but when we exit early, we need to take care of the file descriptor
as well as the lock file ourselves. The lock file may be deleted at the
end of running the program, but we are in library code, so we should
not rely on that.

Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoabbrev_sha1_in_line: don't leak memory
Stefan Beller [Fri, 1 Apr 2016 00:35:44 +0000 (17:35 -0700)]
abbrev_sha1_in_line: don't leak memory

`split` is of type `struct strbuf **`, and currently we are leaking split
itself as well as each element in split[i]. We have a dedicated free
function for `struct strbuf **`, which takes care of freeing all
related memory.

Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agonotes: don't leak memory in git_config_get_notes_strategy
Stefan Beller [Fri, 1 Apr 2016 00:35:43 +0000 (17:35 -0700)]
notes: don't leak memory in git_config_get_notes_strategy

This function asks for the value of a configuration and after
using the value does not have to retain ownership of it.
git_config_get_string_const() however is a function to get a
copy of the value, but we forget to free it before we return.

Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agobuiltin/apply: free patch when parse_chunk() fails
Christian Couder [Wed, 16 Mar 2016 19:35:11 +0000 (20:35 +0100)]
builtin/apply: free patch when parse_chunk() fails

When parse_chunk() fails it can return -1, for example
when find_header() doesn't find a patch header.

In this case it's better in apply_patch() to free the
"struct patch" that we just allocated instead of
leaking it.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agobuiltin/apply: handle parse_binary() failure
Christian Couder [Fri, 18 Mar 2016 12:30:41 +0000 (13:30 +0100)]
builtin/apply: handle parse_binary() failure

In parse_binary() there is:

forward = parse_binary_hunk(&buffer, &size, &status, &used);
if (!forward && !status)
/* there has to be one hunk (forward hunk) */
return error(_("unrecognized binary patch at line %d"), linenr-1);

so parse_binary() can return -1, because that's what error() returns.

Also parse_binary_hunk() sets "status" to -1 in case of error and
parse_binary() does "if (status) return status;".

In this case parse_chunk() should not add -1 to the patchsize it computes.
It is better for future libification efforts to make it just return -1.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agogit-send-pack: fix --all option when used with directory
Stanislav Kolotinskiy [Thu, 31 Mar 2016 13:55:09 +0000 (16:55 +0300)]
git-send-pack: fix --all option when used with directory

When using git send-pack with --all option
and a target repository specification ([<host>:]<directory>),
usage message is being displayed instead of performing
the actual transmission.

The reason for this issue is that destination and refspecs are being set
in the same conditional and are populated from argv. When a target
repository is passed, refspecs is being populated as well with its value.
This makes the check for refspecs not being NULL to always return true,
which, in conjunction with the check for --all or --mirror options,
is always true as well and returns usage message instead of proceeding.

This ensures that send-pack will stop execution only when --all
or --mirror switch is used in conjunction with any refspecs passed.

Signed-off-by: Stanislav Kolotinskiy <stanislav@assembla.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agofor-each-ref: fix description of '--contains' in manpage
SZEDER Gábor [Wed, 30 Mar 2016 13:48:30 +0000 (15:48 +0200)]
for-each-ref: fix description of '--contains' in manpage

'git for-each-ref's manpage says that '--contains' only lists tags,
but it lists all kinds of refs.

Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agodiffcore: fix iteration order of identical files during rename detection
SZEDER Gábor [Wed, 30 Mar 2016 08:35:07 +0000 (10:35 +0200)]
diffcore: fix iteration order of identical files during rename detection

If the two paths 'dir/A/file' and 'dir/B/file' have identical content
and the parent directory is renamed, e.g. 'git mv dir other-dir', then
diffcore reports the following exact renames:

    renamed:    dir/B/file -> other-dir/A/file
    renamed:    dir/A/file -> other-dir/B/file

While technically not wrong, this is confusing not only for the user,
but also for git commands that make decisions based on rename
information, e.g. 'git log --follow other-dir/A/file' follows
'dir/B/file' past the rename.

This behavior is a side effect of commit v2.0.0-rc4~8^2~14
(diffcore-rename.c: simplify finding exact renames, 2013-11-14): the
hashmap storing sources returns entries from the same bucket, i.e.
sources matching the current destination, in LIFO order.  Thus the
iteration first examines 'other-dir/A/file' and 'dir/B/file' and, upon
finding identical content and basename, reports an exact rename.

Other hashmap users are apparently happy with the current iteration
order over the entries of a bucket.  Changing the iteration order
would risk upsetting other hashmap users and would increase the memory
footprint of each bucket by a pointer to the tail element.

Fill the hashmap with source entries in reverse order to restore the
original exact rename detection behavior.

Reported-by: Bill Okara <billokara@gmail.com>
Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoMSVC: use shipped headers instead of fallback definitions
Sven Strickroth [Wed, 30 Mar 2016 11:37:36 +0000 (13:37 +0200)]
MSVC: use shipped headers instead of fallback definitions

VS2010 comes with stdint.h [1]
VS2013 comes with inttypes.h [2]

[1] https://stackoverflow.com/a/2628014/3906760
[2] https://blogs.msdn.microsoft.com/vcblog/2013/07/19/c99-library-support-in-visual-studio-2013/

Signed-off-by: Sven Strickroth <sven@cs-ware.de>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Acked-by: Sebastian Schuberth <sschuberth@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoMSVC: vsnprintf in Visual Studio 2015 doesn't need SNPRINTF_SIZE_CORR any more
Sven Strickroth [Tue, 29 Mar 2016 16:25:28 +0000 (18:25 +0200)]
MSVC: vsnprintf in Visual Studio 2015 doesn't need SNPRINTF_SIZE_CORR any more

In MSVC2015 the behavior of vsnprintf was changed.
W/o this fix there is one character missing at the end.

Signed-off-by: Sven Strickroth <sven@cs-ware.de>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Acked-by: Sebastian Schuberth <sschuberth@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoDocumentation: git diff --check detects conflict markers
Ori Avtalion [Tue, 29 Mar 2016 18:59:45 +0000 (21:59 +0300)]
Documentation: git diff --check detects conflict markers

Signed-off-by: Ori Avtalion <ori@avtalion.name>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agobranch -d: refuse deleting a branch which is currently checked out
Kazuki Yamaguchi [Tue, 29 Mar 2016 09:38:39 +0000 (18:38 +0900)]
branch -d: refuse deleting a branch which is currently checked out

When a branch is checked out by current working tree, deleting the
branch is forbidden. However when the branch is checked out only by
other working trees, deleting incorrectly succeeds.
Use find_shared_symref() to check if the branch is in use, not just
comparing with the current working tree's HEAD.

Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Kazuki Yamaguchi <k@rhe.jp>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoworktree: add: introduce --checkout option
Ray Zhang [Tue, 29 Mar 2016 10:11:01 +0000 (10:11 +0000)]
worktree: add: introduce --checkout option

By adding this option which defaults to true, we can use the
corresponding --no-checkout to make some customizations before
the checkout, like sparse checkout, etc.

Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Helped-by: Junio C Hamano <gitster@pobox.com>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Ray Zhang <zhanglei002@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoGit 2.8 v2.8.0
Junio C Hamano [Mon, 28 Mar 2016 19:19:45 +0000 (12:19 -0700)]
Git 2.8

Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agorebase-i: print an abbreviated hash when stop for editing
Ralf Thielow [Mon, 28 Mar 2016 16:00:00 +0000 (18:00 +0200)]
rebase-i: print an abbreviated hash when stop for editing

The message that is shown when rebase-i stops for editing prints
the full hash of the commit where it stopped which makes the message
overflow to the next line on smaller terminal windows.  Print an
abbreviated hash instead.

Signed-off-by: Ralf Thielow <ralf.thielow@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoapi-parse-options.txt: document OPT_CMDMODE()
Pranit Bauva [Fri, 25 Mar 2016 18:58:58 +0000 (18:58 +0000)]
api-parse-options.txt: document OPT_CMDMODE()

OPT_CMDMODE mechanism was introduced in the release of 1.8.5 to actively
notice when multiple "operation mode" options that specify mutually
incompatible operation modes are given.

Signed-off-by: Pranit Bauva <pranit.bauva@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agocompletion: add 'revisions' and 'everyday' to 'git help'
Ralf Thielow [Thu, 24 Mar 2016 18:27:14 +0000 (19:27 +0100)]
completion: add 'revisions' and 'everyday' to 'git help'

Signed-off-by: Ralf Thielow <ralf.thielow@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agocompletion: add option '--guides' to 'git help'
Ralf Thielow [Thu, 24 Mar 2016 18:27:13 +0000 (19:27 +0100)]
completion: add option '--guides' to 'git help'

Signed-off-by: Ralf Thielow <ralf.thielow@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoMerge branch 'ls/p4-doc-markup'
Junio C Hamano [Thu, 24 Mar 2016 19:28:06 +0000 (12:28 -0700)]
Merge branch 'ls/p4-doc-markup'

* ls/p4-doc-markup:
  Documentation: fix git-p4 AsciiDoc formatting
  Documentation: use ASCII quotation marks in git-p4

8 years agoMerge branch 'js/mingw-tests-2.8'
Junio C Hamano [Thu, 24 Mar 2016 19:27:58 +0000 (12:27 -0700)]
Merge branch 'js/mingw-tests-2.8'

* js/mingw-tests-2.8:
  mingw: skip some tests in t9115 due to file name issues
  t1300: fix the new --show-origin tests on Windows
  t1300-repo-config: make it resilient to being run via 'sh -x'
  config --show-origin: report paths with forward slashes

8 years agoMerge branch 'sb/submodule-module-list-pathspec-fix'
Junio C Hamano [Thu, 24 Mar 2016 19:27:12 +0000 (12:27 -0700)]
Merge branch 'sb/submodule-module-list-pathspec-fix'

A fix for a small regression in "module_list" helper that was
rewritten in C (also applies to 2.7.x).

* sb/submodule-module-list-pathspec-fix:
  submodule: fix regression for deinit without submodules

8 years agoapply: report patch skipping in verbose mode
Nguyễn Thái Ngọc Duy [Thu, 24 Mar 2016 11:56:19 +0000 (18:56 +0700)]
apply: report patch skipping in verbose mode

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agogit-apply.txt: mention the behavior inside a subdir
Nguyễn Thái Ngọc Duy [Thu, 24 Mar 2016 11:56:17 +0000 (18:56 +0700)]
git-apply.txt: mention the behavior inside a subdir

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agogit-apply.txt: remove a space
Nguyễn Thái Ngọc Duy [Thu, 24 Mar 2016 11:56:16 +0000 (18:56 +0700)]
git-apply.txt: remove a space

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agosubmodule-config: use hashmap_iter_first()
Alexander Kuleshov [Wed, 16 Mar 2016 07:46:31 +0000 (13:46 +0600)]
submodule-config: use hashmap_iter_first()

The hashmap API provides hashmap_iter_first() helper for initialion
and getting the first entry of a hashmap. Let's use it instead of
doing initialization manually and then get the first entry.

There are no functional changes, just cleanup.

Signed-off-by: Alexander Kuleshov <kuleshovmail@gmail.com>
Reviewed-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agoMerge branch 'master' of git://github.com/git-l10n/git-po
Junio C Hamano [Wed, 23 Mar 2016 19:22:42 +0000 (12:22 -0700)]
Merge branch 'master' of git://github.com/git-l10n/git-po

* 'master' of git://github.com/git-l10n/git-po:
  l10n: pt_PT: Update and add new translations
  l10n: ca.po: update translation
  l10n: vi.po (2530t): Update translation

8 years agomerge: refuse to create too cool a merge by default
Junio C Hamano [Fri, 18 Mar 2016 20:21:09 +0000 (13:21 -0700)]
merge: refuse to create too cool a merge by default

While it makes sense to allow merging unrelated histories of two
projects that started independently into one, in the way "gitk" was
merged to "git" itself aka "the coolest merge ever", such a merge is
still an unusual event.  Worse, if somebody creates an independent
history by starting from a tarball of an established project and
sends a pull request to the original project, "git merge" however
happily creates such a merge without any sign of something unusual
is happening.

Teach "git merge" to refuse to create such a merge by default,
unless the user passes a new "--allow-unrelated-histories" option to
tell it that the user is aware that two unrelated projects are
merged.

Because such a "two project merge" is a rare event, a configuration
option to always allow such a merge is not added.

We could add the same option to "git pull" and have it passed
through to underlying "git merge".  I do not have a fundamental
opposition against such a feature, but this commit does not do so
and instead leaves it as low-hanging fruit for others, because such
a "two project merge" would be done after fetching the other project
into some location in the working tree of an existing project and
making sure how well they fit together, it is sufficient to allow a
local merge without such an option pass-through from "git pull" to
"git merge".  Many tests that are updated by this patch does the
pass-through manually by turning:

git pull something

into its equivalent:

git fetch something &&
git merge --allow-unrelated-histories FETCH_HEAD

If somebody is inclined to add such an option, updated tests in this
change need to be adjusted back to:

git pull --allow-unrelated-histories something

Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 years agomerge: fix NULL pointer dereference when merging nothing into void
Junio C Hamano [Mon, 21 Mar 2016 19:01:43 +0000 (12:01 -0700)]
merge: fix NULL pointer dereference when merging nothing into void

When we are on an unborn branch and merging only one foreign parent,
we allow "git merge" to fast-forward to that foreign parent commit.

This codepath incorrectly attempted to dereference the list of
parents that the merge is going to record even when the list is
empty.  It must refuse to operate instead when there is no parent.

All other codepaths make sure the list is not empty before they
dereference it, and are safe.

Reported-by: Jose Ivan B. Vilarouca Filho
Signed-off-by: Junio C Hamano <gitster@pobox.com>