git/git.git
3 years agoMerge branch 'ma/bisect-leakfix'
Junio C Hamano [Wed, 15 Nov 2017 03:14:28 +0000 (12:14 +0900)]
Merge branch 'ma/bisect-leakfix'

Leak fixes.

* ma/bisect-leakfix:
  bisect: fix memory leak when returning best element
  bisect: fix off-by-one error in `best_bisection_sorted()`
  bisect: fix memory leak in `find_bisection()`
  bisect: change calling-convention of `find_bisection()`

3 years agoMerge branch 'rs/sequencer-rewrite-file-cleanup'
Junio C Hamano [Wed, 15 Nov 2017 03:14:28 +0000 (12:14 +0900)]
Merge branch 'rs/sequencer-rewrite-file-cleanup'

Code cleanup.

* rs/sequencer-rewrite-file-cleanup:
  sequencer.c: check return value of close() in rewrite_file()
  sequencer: use O_TRUNC to truncate files
  sequencer: factor out rewrite_file()

3 years agoMerge branch 'ao/merge-verbosity-getenv-just-once'
Junio C Hamano [Wed, 15 Nov 2017 03:14:28 +0000 (12:14 +0900)]
Merge branch 'ao/merge-verbosity-getenv-just-once'

Code cleanup.

* ao/merge-verbosity-getenv-just-once:
  merge-recursive: check GIT_MERGE_VERBOSITY only once

3 years agoMerge branch 'mh/avoid-rewriting-packed-refs'
Junio C Hamano [Wed, 15 Nov 2017 03:14:27 +0000 (12:14 +0900)]
Merge branch 'mh/avoid-rewriting-packed-refs'

Recent update to the refs infrastructure implementation started
rewriting packed-refs file more often than before; this has been
optimized again for most trivial cases.

* mh/avoid-rewriting-packed-refs:
  files-backend: don't rewrite the `packed-refs` file unnecessarily
  t1409: check that `packed-refs` is not rewritten unnecessarily

3 years agoMerge branch 'rs/imap-send-next-arg-fix'
Junio C Hamano [Wed, 15 Nov 2017 03:14:26 +0000 (12:14 +0900)]
Merge branch 'rs/imap-send-next-arg-fix'

Error checking in "git imap-send" for empty response has been
improved.

* rs/imap-send-next-arg-fix:
  imap-send: handle missing response codes gracefully
  imap-send: handle NULL return of next_arg()

3 years agoMerge branch 'ab/mediawiki-name-truncation'
Junio C Hamano [Wed, 15 Nov 2017 03:14:25 +0000 (12:14 +0900)]
Merge branch 'ab/mediawiki-name-truncation'

The remote-helper for talking to MediaWiki has been updated to
truncate an overlong pagename so that ".mw" suffix can still be
added.

* ab/mediawiki-name-truncation:
  remote-mediawiki: limit filenames to legal

3 years agoMerge branch 'jm/status-ignored-files-list'
Junio C Hamano [Mon, 13 Nov 2017 05:44:59 +0000 (14:44 +0900)]
Merge branch 'jm/status-ignored-files-list'

The set of paths output from "git status --ignored" was tied
closely with its "--untracked=<mode>" option, but now it can be
controlled more flexibly.  Most notably, a directory that is
ignored because it is listed to be ignored in the ignore/exclude
mechanism can be handled differently from a directory that ends up
to be ignored only because all files in it are ignored.

* jm/status-ignored-files-list:
  status: test ignored modes
  status: document options to show matching ignored files
  status: report matching ignored and normal untracked
  status: add option to show ignored files differently

3 years agoRelNotes: the third batch for 2.16
Junio C Hamano [Thu, 9 Nov 2017 05:36:39 +0000 (14:36 +0900)]
RelNotes: the third batch for 2.16

Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoMerge branch 'js/mingw-redirect-std-handles'
Junio C Hamano [Thu, 9 Nov 2017 05:31:31 +0000 (14:31 +0900)]
Merge branch 'js/mingw-redirect-std-handles'

MinGW updates.

* js/mingw-redirect-std-handles:
  mingw: document the standard handle redirection
  mingw: optionally redirect stderr/stdout via the same handle
  mingw: add experimental feature to redirect standard handles

3 years agoMerge branch 'js/wincred-empty-cred'
Junio C Hamano [Thu, 9 Nov 2017 05:31:31 +0000 (14:31 +0900)]
Merge branch 'js/wincred-empty-cred'

MinGW updates.

* js/wincred-empty-cred:
  wincred: handle empty username/password correctly
  t0302: check helper can handle empty credentials

3 years agoMerge branch 'js/mingw-full-version-in-resources'
Junio C Hamano [Thu, 9 Nov 2017 05:31:31 +0000 (14:31 +0900)]
Merge branch 'js/mingw-full-version-in-resources'

MinGW updates.

* js/mingw-full-version-in-resources:
  mingw: include the full version information in the resources

3 years agoMerge branch 'dk/libsecret-unlock-to-load-fix'
Junio C Hamano [Thu, 9 Nov 2017 05:31:30 +0000 (14:31 +0900)]
Merge branch 'dk/libsecret-unlock-to-load-fix'

The credential helper for libsecret (in contrib/) has been improved
to allow possibly prompting the end user to unlock secrets that are
currently locked (otherwise the secrets may not be loaded).

* dk/libsecret-unlock-to-load-fix:
  credential-libsecret: unlock locked secrets

3 years agoMerge branch 'ks/mailmap'
Junio C Hamano [Thu, 9 Nov 2017 05:31:29 +0000 (14:31 +0900)]
Merge branch 'ks/mailmap'

* ks/mailmap:
  mailmap: use Kaartic Sivaraam's new address

3 years agoMerge branch 'js/early-config'
Junio C Hamano [Thu, 9 Nov 2017 05:31:29 +0000 (14:31 +0900)]
Merge branch 'js/early-config'

Correct start-up sequence so that a repository could be placed
immediately under the root directory again (which was broken at
around Git 2.13).

* js/early-config:
  setup: avoid double slashes when looking for HEAD

3 years agoMerge branch 'sg/travis-fixes'
Junio C Hamano [Thu, 9 Nov 2017 05:31:28 +0000 (14:31 +0900)]
Merge branch 'sg/travis-fixes'

TravisCI build updates.

* sg/travis-fixes:
  travis-ci: don't build Git for the static analysis job
  travis-ci: fix running P4 and Git LFS tests in Linux build jobs

3 years agoMerge branch 'bw/diff-opt-impl-to-bitfields'
Junio C Hamano [Thu, 9 Nov 2017 05:31:27 +0000 (14:31 +0900)]
Merge branch 'bw/diff-opt-impl-to-bitfields'

A single-word "unsigned flags" in the diff options is being split
into a structure with many bitfields.

* bw/diff-opt-impl-to-bitfields:
  diff: make struct diff_flags members lowercase
  diff: remove DIFF_OPT_CLR macro
  diff: remove DIFF_OPT_SET macro
  diff: remove DIFF_OPT_TST macro
  diff: remove touched flags
  diff: add flag to indicate textconv was set via cmdline
  diff: convert flags to be stored in bitfields
  add, reset: use DIFF_OPT_SET macro to set a diff flag

3 years agoMerge branch 'rs/hex-to-bytes-cleanup'
Junio C Hamano [Thu, 9 Nov 2017 05:31:27 +0000 (14:31 +0900)]
Merge branch 'rs/hex-to-bytes-cleanup'

Code cleanup.

* rs/hex-to-bytes-cleanup:
  sha1_file: use hex_to_bytes()
  http-push: use hex_to_bytes()
  notes: move hex_to_bytes() to hex.c and export it

3 years agoMerge branch 'ad/5580-unc-tests-on-cygwin'
Junio C Hamano [Thu, 9 Nov 2017 05:31:27 +0000 (14:31 +0900)]
Merge branch 'ad/5580-unc-tests-on-cygwin'

UNC paths are also relevant in Cygwin builds and they are now
tested just like Mingw builds.

* ad/5580-unc-tests-on-cygwin:
  t5580: add Cygwin support

3 years agoMerge branch 'ao/diff-populate-filespec-lstat-errorpath-fix'
Junio C Hamano [Thu, 9 Nov 2017 05:31:26 +0000 (14:31 +0900)]
Merge branch 'ao/diff-populate-filespec-lstat-errorpath-fix'

After an error from lstat(), diff_populate_filespec() function
sometimes still went ahead and used invalid data in struct stat,
which has been fixed.

* ao/diff-populate-filespec-lstat-errorpath-fix:
  diff: fix lstat() error handling in diff_populate_filespec()

3 years agoMerge branch 'sb/blame-config-doc'
Junio C Hamano [Thu, 9 Nov 2017 05:31:25 +0000 (14:31 +0900)]
Merge branch 'sb/blame-config-doc'

Description of blame.{showroot,blankboundary,showemail,date}
configuration variables have been added to "git config --help".

* sb/blame-config-doc:
  config: document blame configuration

3 years agoMerge branch 'jm/relnotes-2.15-typofix'
Junio C Hamano [Thu, 9 Nov 2017 05:31:25 +0000 (14:31 +0900)]
Merge branch 'jm/relnotes-2.15-typofix'

Typofix.

* jm/relnotes-2.15-typofix:
  fix typos in 2.15.0 release notes

3 years agoRelNotes: the second batch post 2.15 comes
Junio C Hamano [Mon, 6 Nov 2017 05:31:16 +0000 (14:31 +0900)]
RelNotes: the second batch post 2.15 comes

Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoMerge branch 'tg/deprecate-stash-save'
Junio C Hamano [Mon, 6 Nov 2017 05:24:32 +0000 (14:24 +0900)]
Merge branch 'tg/deprecate-stash-save'

"git stash save" has been deprecated in favour of "git stash push".

* tg/deprecate-stash-save:
  stash: remove now superfluos help for "stash push"
  stash: mark "git stash save" deprecated in the man page
  stash: replace "git stash save" with "git stash push" in the documentation

3 years agoMerge branch 'tb/complete-checkout'
Junio C Hamano [Mon, 6 Nov 2017 05:24:31 +0000 (14:24 +0900)]
Merge branch 'tb/complete-checkout'

Command line completion (in contrib/) update.

* tb/complete-checkout:
  completion: add remaining flags to checkout

3 years agoMerge branch 'gc/gitweb-filetest-acl'
Junio C Hamano [Mon, 6 Nov 2017 05:24:30 +0000 (14:24 +0900)]
Merge branch 'gc/gitweb-filetest-acl'

"gitweb" checks if a directory is searchable with Perl's "-x"
operator, which can be enhanced by using "filetest 'access'"
pragma, which now we do.

* gc/gitweb-filetest-acl:
  gitweb: use filetest to allow ACLs

3 years agoMerge branch 'mp/push-pushoption-config'
Junio C Hamano [Mon, 6 Nov 2017 05:24:30 +0000 (14:24 +0900)]
Merge branch 'mp/push-pushoption-config'

The "--push-option=<string>" option to "git push" now defaults to a
list of strings configured via push.pushOption variable.

* mp/push-pushoption-config:
  builtin/push.c: add push.pushOption config

3 years agoMerge branch 'hv/fetch-moved-submodules-on-demand'
Junio C Hamano [Mon, 6 Nov 2017 05:24:29 +0000 (14:24 +0900)]
Merge branch 'hv/fetch-moved-submodules-on-demand'

"git fetch --recurse-submodules" now knows that submodules can be
moved around in the superproject in addition to getting updated,
and finds the ones that need to be fetched accordingly.

* hv/fetch-moved-submodules-on-demand:
  submodule: simplify decision tree whether to or not to fetch
  implement fetching of moved submodules
  fetch: add test to make sure we stay backwards compatible

3 years agoMerge branch 'jc/check-ref-format-oor'
Junio C Hamano [Mon, 6 Nov 2017 05:24:28 +0000 (14:24 +0900)]
Merge branch 'jc/check-ref-format-oor'

"git check-ref-format --branch @{-1}" bit a "BUG()" when run
outside a repository for obvious reasons; clarify the documentation
and make sure we do not even try to expand the at-mark magic in
such a case, but still call the validation logic for branch names.

* jc/check-ref-format-oor:
  check-ref-format doc: --branch validates and expands <branch>
  check-ref-format --branch: strip refs/heads/ using skip_prefix
  check-ref-format --branch: do not expand @{...} outside repository

3 years agoMerge branch 'jc/t5601-copy-workaround'
Junio C Hamano [Mon, 6 Nov 2017 05:24:27 +0000 (14:24 +0900)]
Merge branch 'jc/t5601-copy-workaround'

A (possibly flakey) test fix.

* jc/t5601-copy-workaround:
  t5601: rm the target file of cp that could still be executing

3 years agoMerge branch 'bc/object-id'
Junio C Hamano [Mon, 6 Nov 2017 05:24:27 +0000 (14:24 +0900)]
Merge branch 'bc/object-id'

Conversion from uchar[20] to struct object_id continues.

* bc/object-id: (25 commits)
  refs/files-backend: convert static functions to object_id
  refs: convert read_raw_ref backends to struct object_id
  refs: convert peel_object to struct object_id
  refs: convert resolve_ref_unsafe to struct object_id
  worktree: convert struct worktree to object_id
  refs: convert resolve_gitlink_ref to struct object_id
  Convert remaining callers of resolve_gitlink_ref to object_id
  sha1_file: convert index_path and index_fd to struct object_id
  refs: convert reflog_expire parameter to struct object_id
  refs: convert read_ref_at to struct object_id
  refs: convert peel_ref to struct object_id
  builtin/pack-objects: convert to struct object_id
  pack-bitmap: convert traverse_bitmap_commit_list to object_id
  refs: convert dwim_log to struct object_id
  builtin/reflog: convert remaining unsigned char uses to object_id
  refs: convert dwim_ref and expand_ref to struct object_id
  refs: convert read_ref and read_ref_full to object_id
  refs: convert resolve_refdup and refs_resolve_refdup to struct object_id
  Convert check_connected to use struct object_id
  refs: update ref transactions to use struct object_id
  ...

3 years agoMerge branch 'jk/revision-pruning-optim'
Junio C Hamano [Mon, 6 Nov 2017 05:24:26 +0000 (14:24 +0900)]
Merge branch 'jk/revision-pruning-optim'

Pathspec-limited revision traversal was taught not to keep finding
unneeded differences once it knows two trees are different inside
given pathspec.

* jk/revision-pruning-optim:
  revision: quit pruning diff more quickly when possible

3 years agoMerge branch 'ds/find-unique-abbrev-optim'
Junio C Hamano [Mon, 6 Nov 2017 05:24:25 +0000 (14:24 +0900)]
Merge branch 'ds/find-unique-abbrev-optim'

Optimize the code to find shortest unique prefix of object names.

* ds/find-unique-abbrev-optim:
  sha1_name: minimize OID comparisons during disambiguation
  sha1_name: parse less while finding common prefix
  sha1_name: unroll len loop in find_unique_abbrev_r()
  p4211-line-log.sh: add log --online --raw --parents perf test

3 years agoMerge branch 'wk/pull-signoff'
Junio C Hamano [Mon, 6 Nov 2017 05:24:24 +0000 (14:24 +0900)]
Merge branch 'wk/pull-signoff'

"git pull" has been taught to accept "--[no-]signoff" option and
pass it down to "git merge".

* wk/pull-signoff:
  pull: pass --signoff/--no-signoff to "git merge"

3 years agoMerge branch 'pc/submodule-helper'
Junio C Hamano [Mon, 6 Nov 2017 05:24:23 +0000 (14:24 +0900)]
Merge branch 'pc/submodule-helper'

GSoC.

* pc/submodule-helper:
  submodule: port submodule subcommand 'status' from shell to C
  submodule--helper: introduce for_each_listed_submodule()
  submodule--helper: introduce get_submodule_displaypath()

3 years agoMerge branch 'pb/bisect-helper'
Junio C Hamano [Mon, 6 Nov 2017 05:24:23 +0000 (14:24 +0900)]
Merge branch 'pb/bisect-helper'

An early part of piece-by-piece rewrite of "git bisect".

* pb/bisect-helper:
  bisect--helper: `is_expected_rev` & `check_expected_revs` shell function in C
  t6030: explicitly test for bisection cleanup
  bisect--helper: `bisect_clean_state` shell function in C
  bisect--helper: `write_terms` shell function in C
  bisect--helper: rewrite `check_term_format` shell function in C
  bisect--helper: use OPT_CMDMODE instead of OPT_BOOL

3 years agoMerge branch 'dm/run-command-ignored-hook-advise'
Junio C Hamano [Mon, 6 Nov 2017 05:24:22 +0000 (14:24 +0900)]
Merge branch 'dm/run-command-ignored-hook-advise'

A hook script that is set unexecutable is simply ignored.  Git
notifies when such a file is ignored, unless the message is
squelched via advice.ignoredHook configuration.

* dm/run-command-ignored-hook-advise:
  run-command: add hint when a hook is ignored

3 years agoThe first batch for 2.16
Junio C Hamano [Mon, 6 Nov 2017 04:18:22 +0000 (13:18 +0900)]
The first batch for 2.16

The most notable change is that we no longer take "git add ''" and
add everything.  An empty string is now an error when used as a
pathspec element.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoMerge branch 'ex/deprecate-empty-pathspec-as-match-all'
Junio C Hamano [Mon, 6 Nov 2017 04:11:29 +0000 (13:11 +0900)]
Merge branch 'ex/deprecate-empty-pathspec-as-match-all'

The final step to make an empty string as a pathspec element
illegal.  We started this by first deprecating and warning a
pathspec that has such an element in 2.11 (Nov 2016).

Hopefully we can merge this down to the 'master' by the end of the
year?  A deprecation warning period that is about 1 year does not
sound too bad.

* ex/deprecate-empty-pathspec-as-match-all:
  pathspec: die on empty strings as pathspec
  t0027: do not use an empty string as a pathspec element

3 years agoMerge branch 'jk/rebase-i-exec-gitdir-fix'
Junio C Hamano [Mon, 6 Nov 2017 04:11:28 +0000 (13:11 +0900)]
Merge branch 'jk/rebase-i-exec-gitdir-fix'

A recent regression in "git rebase -i" that broke execution of git
commands from subdirectories via "exec" insn has been fixed.

* jk/rebase-i-exec-gitdir-fix:
  sequencer: pass absolute GIT_DIR to exec commands

3 years agoMerge branch 'cn/diff-indent-no-longer-is-experimental'
Junio C Hamano [Mon, 6 Nov 2017 04:11:27 +0000 (13:11 +0900)]
Merge branch 'cn/diff-indent-no-longer-is-experimental'

Doc update.

* cn/diff-indent-no-longer-is-experimental:
  diff: --indent-heuristic is no longer experimental

3 years agoMerge branch 'bw/grep-recurse-submodules'
Junio C Hamano [Mon, 6 Nov 2017 04:11:27 +0000 (13:11 +0900)]
Merge branch 'bw/grep-recurse-submodules'

A broken access to object databases in recent update to "git grep
--recurse-submodules" has been fixed.

* bw/grep-recurse-submodules:
  grep: take the read-lock when adding a submodule

3 years agoMerge branch 'mh/test-local-canary'
Junio C Hamano [Mon, 6 Nov 2017 04:11:26 +0000 (13:11 +0900)]
Merge branch 'mh/test-local-canary'

We try to see if somebody runs our test suite with a shell that
does not support "local" like bash/dash does.

* mh/test-local-canary:
  t0000: check whether the shell supports the "local" keyword

3 years agoMerge branch 'js/submodule-in-excluded'
Junio C Hamano [Mon, 6 Nov 2017 04:11:26 +0000 (13:11 +0900)]
Merge branch 'js/submodule-in-excluded'

"git status --ignored -u" did not stop at a working tree of a
separate project that is embedded in an ignored directory and
listed files in that other project, instead of just showing the
directory itself as ignored.

* js/submodule-in-excluded:
  status: do not get confused by submodules in excluded directories

3 years agoMerge branch 'ao/check-resolve-ref-unsafe-result'
Junio C Hamano [Mon, 6 Nov 2017 04:11:25 +0000 (13:11 +0900)]
Merge branch 'ao/check-resolve-ref-unsafe-result'

"git commit", after making a commit, did not check for errors when
asking on what branch it made the commit, which has been correted.

* ao/check-resolve-ref-unsafe-result:
  commit: check result of resolve_ref_unsafe

3 years agoMerge branch 'jk/misc-resolve-ref-unsafe-fixes'
Junio C Hamano [Mon, 6 Nov 2017 04:11:24 +0000 (13:11 +0900)]
Merge branch 'jk/misc-resolve-ref-unsafe-fixes'

Some codepaths did not check for errors when asking what branch the
HEAD points at, which have been fixed.

* jk/misc-resolve-ref-unsafe-fixes:
  worktree: handle broken symrefs in find_shared_symref()
  log: handle broken HEAD in decoration check
  remote: handle broken symrefs
  test-ref-store: avoid passing NULL to printf

3 years agoMerge branch 'sb/diff-color-moved-use-xdl-recmatch'
Junio C Hamano [Mon, 6 Nov 2017 04:11:24 +0000 (13:11 +0900)]
Merge branch 'sb/diff-color-moved-use-xdl-recmatch'

Instead of using custom line comparison and hashing functions to
implement "moved lines" coloring in the diff output, use the pair
of these functions from lower-layer xdiff/ code.

* sb/diff-color-moved-use-xdl-recmatch:
  diff.c: get rid of duplicate implementation
  xdiff-interface: export comparing and hashing strings

3 years agoMerge branch 'jk/diff-color-moved-fix'
Junio C Hamano [Mon, 6 Nov 2017 04:11:23 +0000 (13:11 +0900)]
Merge branch 'jk/diff-color-moved-fix'

The experimental "color moved lines differently in diff output"
feature was buggy around "ignore whitespace changes" edges, whihch
has been corrected.

* jk/diff-color-moved-fix:
  diff: handle NULs in get_string_hash()
  diff: fix whitespace-skipping with --color-moved
  t4015: test the output of "diff --color-moved -b"
  t4015: check "negative" case for "-w --color-moved"
  t4015: refactor --color-moved whitespace test

3 years agoMerge branch 'kd/auto-col-with-pager-fix'
Junio C Hamano [Mon, 6 Nov 2017 04:11:22 +0000 (13:11 +0900)]
Merge branch 'kd/auto-col-with-pager-fix'

"auto" as a value for the columnar output configuration ought to
judge "is the output consumed by humans?" with the same criteria as
"auto" for coloured output configuration, i.e. either the standard
output stream is going to tty, or a pager is in use.  We forgot the
latter, which has been fixed.

* kd/auto-col-with-pager-fix:
  column: do not include pager.c
  column: show auto columns when pager is active

3 years agoMerge branch 'jc/no-cmd-as-subroutine'
Junio C Hamano [Mon, 6 Nov 2017 04:11:21 +0000 (13:11 +0900)]
Merge branch 'jc/no-cmd-as-subroutine'

Calling cmd_foo() as if it is a general purpose helper function is
a no-no.  Correct two instances of such to set an example.

* jc/no-cmd-as-subroutine:
  merge-ours: do not use cmd_*() as a subroutine
  describe: do not use cmd_*() as a subroutine

3 years agoMerge branch 'ma/lockfile-fixes'
Junio C Hamano [Mon, 6 Nov 2017 04:11:21 +0000 (13:11 +0900)]
Merge branch 'ma/lockfile-fixes'

An earlier update made it possible to use an on-stack in-core
lockfile structure (as opposed to having to deliberately leak an
on-heap one).  Many codepaths have been updated to take advantage
of this new facility.

* ma/lockfile-fixes:
  read_cache: roll back lock in `update_index_if_able()`
  read-cache: leave lock in right state in `write_locked_index()`
  read-cache: drop explicit `CLOSE_LOCK`-flag
  cache.h: document `write_locked_index()`
  apply: remove `newfd` from `struct apply_state`
  apply: move lockfile into `apply_state`
  cache-tree: simplify locking logic
  checkout-index: simplify locking logic
  tempfile: fix documentation on `delete_tempfile()`
  lockfile: fix documentation on `close_lock_file_gently()`
  treewide: prefer lockfiles on the stack
  sha1_file: do not leak `lock_file`

3 years agofix typos in 2.15.0 release notes
Jean Carlo Machado [Sat, 4 Nov 2017 12:16:16 +0000 (10:16 -0200)]
fix typos in 2.15.0 release notes

Signed-off-by: Jean Carlo Machado <contato@jeancarlomachado.com.br>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agobisect: fix memory leak when returning best element
Martin Ågren [Sun, 5 Nov 2017 20:24:31 +0000 (21:24 +0100)]
bisect: fix memory leak when returning best element

When `find_bisection()` returns a single list entry, it leaks the other
entries. Move the to-be-returned item to the front and free the
remainder.

Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agobisect: fix off-by-one error in `best_bisection_sorted()`
Martin Ågren [Sun, 5 Nov 2017 20:24:30 +0000 (21:24 +0100)]
bisect: fix off-by-one error in `best_bisection_sorted()`

After we have sorted the `cnt`-many commits that we have selected, we
place them into the commit list. We then set `p->next` to NULL, but as
we do so, `p` is already pointing one beyond item number `cnt`. Indeed,
we check whether `p` is NULL before dereferencing it.

This only matters if there are TREESAME-commits. Since they should be
skipped, they are not included in `cnt` and we will hit the situation
where we set `p->next` to NULL. As a result, the list will be one longer
than it should be. The last commit in the list will be one which occurs
earlier, or which shouldn't be included.

Do not update `p` the very last round in the loop. This ensures that
after the loop, `p->next` points to the remainder of the list, and we
can set it to NULL. While we're here, free that remainder to fix a
memory leak.

Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agobisect: fix memory leak in `find_bisection()`
Martin Ågren [Sun, 5 Nov 2017 20:24:29 +0000 (21:24 +0100)]
bisect: fix memory leak in `find_bisection()`

`find_bisection()` rebuilds the commit list it is given by reversing it
and skipping uninteresting commits. The uninteresting list entries are
leaked. Free them to fix the leak.

Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agobisect: change calling-convention of `find_bisection()`
Martin Ågren [Sun, 5 Nov 2017 20:24:28 +0000 (21:24 +0100)]
bisect: change calling-convention of `find_bisection()`

This function takes a commit list and returns a commit list. The
returned list is built by modifying the original list. Thus the caller
should not use the original list again (and after the next commit fixes
a memory leak, it must not).

Change the function signature so that it takes a **list and has void
return type. That should make it harder to misuse this function.

While we're here, document this function.

Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoconfig: document blame configuration
Stefan Beller [Fri, 3 Nov 2017 19:21:58 +0000 (12:21 -0700)]
config: document blame configuration

The options are currently only referenced by the git-blame man page,
also explain them in git-config, which is the canonical page to
contain all config options.

Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agocredential-libsecret: unlock locked secrets
Dennis Kaarsemaker [Fri, 3 Nov 2017 20:44:49 +0000 (21:44 +0100)]
credential-libsecret: unlock locked secrets

Credentials exposed by the secret service DBUS interface may be locked.
Setting the SECRET_SEARCH_UNLOCK flag will make the secret service
unlock these secrets, possibly prompting the user for credentials to do
so. Without this flag, the secret is simply not loaded.

Signed-off-by: Dennis Kaarsemaker <dennis@kaarsemaker.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agosetup: avoid double slashes when looking for HEAD
Jeff King [Fri, 3 Nov 2017 12:58:02 +0000 (13:58 +0100)]
setup: avoid double slashes when looking for HEAD

Andrew Baumann reported that when called outside of any Git worktree,
`git rev-parse --is-inside-work-tree` eventually tries to access
`//HEAD`, i.e.  any `HEAD` file in the root directory, but with a double
slash.

This double slash is not only unintentional, but is allowed by the POSIX
standard to have a special meaning. And most notably on Windows, it
does, where it refers to a UNC path of the form `//server/share/`.

As a consequence, afore-mentioned `rev-parse` call not only looks for
the wrong thing, but it also causes serious delays, as Windows will try
to access a server called `HEAD`.  Let's simply avoid the unintended
double slash.

Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agomailmap: use Kaartic Sivaraam's new address
Kaartic Sivaraam [Fri, 3 Nov 2017 05:43:41 +0000 (11:13 +0530)]
mailmap: use Kaartic Sivaraam's new address

Map the old address to the new, hopefully more permanent one.

Signed-off-by: Kaartic Sivaraam <kaartic.sivaraam@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoimap-send: handle missing response codes gracefully
René Scharfe [Thu, 2 Nov 2017 17:27:05 +0000 (18:27 +0100)]
imap-send: handle missing response codes gracefully

Response codes are optional.  Exit parse_response_code() early if it's
passed a NULL string, indicating that we reached the end of the reply.
This avoids dereferencing said NULL pointer.

Noticed-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoimap-send: handle NULL return of next_arg()
René Scharfe [Wed, 1 Nov 2017 17:03:20 +0000 (18:03 +0100)]
imap-send: handle NULL return of next_arg()

next_arg() returns NULL if it runs out of arguments.  Most call sites
already handle that gracefully.  Check in the remaining cases as well.
Replace the NULL pointer with an empty string at the bottom of
get_cmd_result() -- it's nicely reported as an unexpected response a
few lines down.  Error out explicitly at the remaining sites.

Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agodiff: --indent-heuristic is no longer experimental
Carlos Martín Nieto [Sun, 29 Oct 2017 15:12:28 +0000 (16:12 +0100)]
diff: --indent-heuristic is no longer experimental

This heuristic has been the default since 2.14 so we should not confuse our
users by saying that it's experimental and off by default.

Signed-off-by: Carlos Martín Nieto <cmn@dwim.me>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agosequencer.c: check return value of close() in rewrite_file()
Simon Ruderich [Wed, 1 Nov 2017 14:45:42 +0000 (15:45 +0100)]
sequencer.c: check return value of close() in rewrite_file()

Not checking close(2) can hide errors as not all errors are reported
during the write(2).

Signed-off-by: Simon Ruderich <simon@ruderich.org>
Reviewed-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agomingw: document the standard handle redirection
Johannes Schindelin [Wed, 1 Nov 2017 17:10:33 +0000 (18:10 +0100)]
mingw: document the standard handle redirection

This feature has been in Git for Windows since v2.11.0(2), as an
experimental option. Now it is considered mature, and it is high time to
document it properly.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agomingw: optionally redirect stderr/stdout via the same handle
Johannes Schindelin [Wed, 1 Nov 2017 17:10:30 +0000 (18:10 +0100)]
mingw: optionally redirect stderr/stdout via the same handle

The "2>&1" notation in Powershell and in Unix shells implies that stderr
is redirected to the same handle into which stdout is already written.

Let's use this special value to allow the same trick with
GIT_REDIRECT_STDERR and GIT_REDIRECT_STDOUT: if the former's value is
`2>&1`, then stderr will simply be written to the same handle as stdout.

The functionality was suggested by Jeff Hostetler.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agomingw: add experimental feature to redirect standard handles
Johannes Schindelin [Wed, 1 Nov 2017 17:10:25 +0000 (18:10 +0100)]
mingw: add experimental feature to redirect standard handles

Particularly when calling Git from applications, such as Visual Studio's
Team Explorer, it is important that stdin/stdout/stderr are closed
properly. However, when spawning processes on Windows, those handles
must be marked as inheritable if we want to use them, but that flag is a
global flag and may very well be used by other spawned processes which
then do not know to close those handles.

Let's introduce a set of environment variables (GIT_REDIRECT_STDIN and
friends) that specify paths to files, or even better, named pipes (which
are similar to Unix sockets) and that are used by the spawned Git
process.  This helps work around above-mentioned issue: those named
pipes will be opened in a non-inheritable way upon startup, and no
handles are passed around (and therefore no inherited handles need to be
closed by any spawned child).

This feature shipped with Git for Windows (marked as experimental) since
v2.11.0(2), so it has seen some serious testing in the meantime.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agotravis-ci: don't build Git for the static analysis job
SZEDER Gábor [Wed, 1 Nov 2017 11:56:44 +0000 (12:56 +0100)]
travis-ci: don't build Git for the static analysis job

The static analysis job on Travis CI builds Git ever since it was
introduced in d8245bb3f (travis-ci: add static analysis build job to
run coccicheck, 2017-04-11).  However, Coccinelle, the only static
analysis tool in use, only needs Git's source code to work and it
doesn't care about built Git binaries at all.

Spare some of Travis CI's resources and don't build Git for the static
analysis job unnecessarily.

Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agotravis-ci: fix running P4 and Git LFS tests in Linux build jobs
SZEDER Gábor [Wed, 1 Nov 2017 11:55:35 +0000 (12:55 +0100)]
travis-ci: fix running P4 and Git LFS tests in Linux build jobs

Linux build jobs on Travis CI skip the P4 and Git LFS tests since
commit 657343a60 (travis-ci: move Travis CI code into dedicated
scripts, 2017-09-10), claiming there are no P4 or Git LFS installed.

The reason is that P4 and Git LFS binaries are not installed to a
directory in the default $PATH, but their directories are prepended to
$PATH.  This worked just fine before said commit, because $PATH was
set in a scriptlet embedded in our '.travis.yml', thus its new value
was visible during the rest of the build job.  However, after these
embedded scriptlets were moved into dedicated scripts executed in
separate shell processes, any variable set in one of those scripts is
only visible in that single script but not in any of the others.  In
this case, 'ci/install-dependencies.sh' downloads P4 and Git LFS and
modifies $PATH, but to no effect, because 'ci/run-tests.sh' only sees
Travis CI's default $PATH.

Move adjusting $PATH to 'ci/lib-travisci.sh', which is sourced in all
other 'ci/' scripts, so all those scripts will see the updated $PATH
value.

Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agogrep: take the read-lock when adding a submodule
Martin Ågren [Wed, 1 Nov 2017 20:45:06 +0000 (21:45 +0100)]
grep: take the read-lock when adding a submodule

With --recurse-submodules, we add each submodule that we encounter to
the list of alternate object databases. With threading, our changes to
the list are not protected against races. Indeed, ThreadSanitizer
reports a race when we call `add_to_alternates_memory()` around the same
time that another thread is reading in the list through
`read_sha1_file()`.

Take the grep read-lock while adding the submodule. The lock is used to
serialize uses of non-thread-safe parts of Git's API, including
`read_sha1_file()`.

Helped-by: Brandon Williams <bmwill@google.com>
Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Acked-by: Brandon Williams <bmwill@google.com>
Reviewed-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agosequencer: pass absolute GIT_DIR to exec commands
Jacob Keller [Tue, 31 Oct 2017 23:07:33 +0000 (16:07 -0700)]
sequencer: pass absolute GIT_DIR to exec commands

When we replaced the old shell script based interactive rebase in
commmit 18633e1a22a6 ("rebase -i: use the rebase--helper builtin",
2017-02-09) we introduced a regression of functionality in that the
GIT_DIR would be sent to the environment of the exec command as-is.

This generally meant that it would be passed as "GIT_DIR=.git", which
causes problems for any exec command that wants to run git commands in
a subdirectory.

This isn't a very large regression, since it is not that likely that the
exec command will run a git command, and even less likely that it will
need to do so in a subdir. This regression was discovered by a build
system which uses git-describe to find the current version of the build
system, and happened to do so from the src/ sub directory of the
project.

Fix this by passing in the absolute path of the git directory into the
child environment.

Signed-off-by: Jacob Keller <jacob.keller@gmail.com>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agowincred: handle empty username/password correctly
Jakub Bereżański [Mon, 30 Oct 2017 17:20:44 +0000 (18:20 +0100)]
wincred: handle empty username/password correctly

Empty (length 0) usernames and/or passwords, when saved in the Windows
Credential Manager, come back as null when reading the credential.

One use case for such empty credentials is with NTLM authentication, where
empty username and password instruct libcurl to authenticate using the
credentials of the currently logged-on user (single sign-on).

When locating the relevant credentials, make empty username match null.
When outputting the credentials, handle nulls correctly.

Signed-off-by: Jakub Bereżański <kuba@berezanscy.pl>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agot0302: check helper can handle empty credentials
Jakub Bereżański [Mon, 30 Oct 2017 17:20:12 +0000 (18:20 +0100)]
t0302: check helper can handle empty credentials

Make sure the helper does not crash when blank username and password is
provided. If the helper can save such credentials, it should be able to
read them back.

Signed-off-by: Jakub Bereżański <kuba@berezanscy.pl>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agomingw: include the full version information in the resources
Johannes Schindelin [Mon, 30 Oct 2017 17:19:42 +0000 (18:19 +0100)]
mingw: include the full version information in the resources

This fixes https://github.com/git-for-windows/git/issues/723

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoremote-mediawiki: limit filenames to legal
Antoine Beaupré [Sun, 29 Oct 2017 18:15:11 +0000 (14:15 -0400)]
remote-mediawiki: limit filenames to legal

mediawiki pages can have names longer than NAME_MAX (generally 255)
characters, which will fail on checkout. we simply strip out extra
characters, which may mean one page's content will overwrite another
(the last editing winning).

ideally, we would do a more clever system to find unique names, but
that would be more difficult and error prone for a situation that
should rarely happen in the first place.

Signed-off-by: Antoine Beaupré <anarcat@debian.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agodiff: make struct diff_flags members lowercase
Brandon Williams [Tue, 31 Oct 2017 18:19:11 +0000 (11:19 -0700)]
diff: make struct diff_flags members lowercase

Now that the flags stored in struct diff_flags are being accessed
directly and not through macros, change all struct members from being
uppercase to lowercase.
This conversion is done using the following semantic patch:

@@
expression E;
@@
- E.RECURSIVE
+ E.recursive

@@
expression E;
@@
- E.TREE_IN_RECURSIVE
+ E.tree_in_recursive

@@
expression E;
@@
- E.BINARY
+ E.binary

@@
expression E;
@@
- E.TEXT
+ E.text

@@
expression E;
@@
- E.FULL_INDEX
+ E.full_index

@@
expression E;
@@
- E.SILENT_ON_REMOVE
+ E.silent_on_remove

@@
expression E;
@@
- E.FIND_COPIES_HARDER
+ E.find_copies_harder

@@
expression E;
@@
- E.FOLLOW_RENAMES
+ E.follow_renames

@@
expression E;
@@
- E.RENAME_EMPTY
+ E.rename_empty

@@
expression E;
@@
- E.HAS_CHANGES
+ E.has_changes

@@
expression E;
@@
- E.QUICK
+ E.quick

@@
expression E;
@@
- E.NO_INDEX
+ E.no_index

@@
expression E;
@@
- E.ALLOW_EXTERNAL
+ E.allow_external

@@
expression E;
@@
- E.EXIT_WITH_STATUS
+ E.exit_with_status

@@
expression E;
@@
- E.REVERSE_DIFF
+ E.reverse_diff

@@
expression E;
@@
- E.CHECK_FAILED
+ E.check_failed

@@
expression E;
@@
- E.RELATIVE_NAME
+ E.relative_name

@@
expression E;
@@
- E.IGNORE_SUBMODULES
+ E.ignore_submodules

@@
expression E;
@@
- E.DIRSTAT_CUMULATIVE
+ E.dirstat_cumulative

@@
expression E;
@@
- E.DIRSTAT_BY_FILE
+ E.dirstat_by_file

@@
expression E;
@@
- E.ALLOW_TEXTCONV
+ E.allow_textconv

@@
expression E;
@@
- E.TEXTCONV_SET_VIA_CMDLINE
+ E.textconv_set_via_cmdline

@@
expression E;
@@
- E.DIFF_FROM_CONTENTS
+ E.diff_from_contents

@@
expression E;
@@
- E.DIRTY_SUBMODULES
+ E.dirty_submodules

@@
expression E;
@@
- E.IGNORE_UNTRACKED_IN_SUBMODULES
+ E.ignore_untracked_in_submodules

@@
expression E;
@@
- E.IGNORE_DIRTY_SUBMODULES
+ E.ignore_dirty_submodules

@@
expression E;
@@
- E.OVERRIDE_SUBMODULE_CONFIG
+ E.override_submodule_config

@@
expression E;
@@
- E.DIRSTAT_BY_LINE
+ E.dirstat_by_line

@@
expression E;
@@
- E.FUNCCONTEXT
+ E.funccontext

@@
expression E;
@@
- E.PICKAXE_IGNORE_CASE
+ E.pickaxe_ignore_case

@@
expression E;
@@
- E.DEFAULT_FOLLOW_RENAMES
+ E.default_follow_renames

Signed-off-by: Brandon Williams <bmwill@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agodiff: remove DIFF_OPT_CLR macro
Brandon Williams [Tue, 31 Oct 2017 18:19:10 +0000 (11:19 -0700)]
diff: remove DIFF_OPT_CLR macro

Remove the `DIFF_OPT_CLR` macro and instead set the flags directly.
This conversion is done using the following semantic patch:

@@
expression E;
identifier fld;
@@
- DIFF_OPT_CLR(&E, fld)
+ E.flags.fld = 0

@@
type T;
T *ptr;
identifier fld;
@@
- DIFF_OPT_CLR(ptr, fld)
+ ptr->flags.fld = 0

Signed-off-by: Brandon Williams <bmwill@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agodiff: remove DIFF_OPT_SET macro
Brandon Williams [Tue, 31 Oct 2017 18:19:09 +0000 (11:19 -0700)]
diff: remove DIFF_OPT_SET macro

Remove the `DIFF_OPT_SET` macro and instead set the flags directly.
This conversion is done using the following semantic patch:

@@
expression E;
identifier fld;
@@
- DIFF_OPT_SET(&E, fld)
+ E.flags.fld = 1

@@
type T;
T *ptr;
identifier fld;
@@
- DIFF_OPT_SET(ptr, fld)
+ ptr->flags.fld = 1

Signed-off-by: Brandon Williams <bmwill@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agodiff: remove DIFF_OPT_TST macro
Brandon Williams [Tue, 31 Oct 2017 18:19:08 +0000 (11:19 -0700)]
diff: remove DIFF_OPT_TST macro

Remove the `DIFF_OPT_TST` macro and instead access the flags directly.
This conversion is done using the following semantic patch:

@@
expression E;
identifier fld;
@@
- DIFF_OPT_TST(&E, fld)
+ E.flags.fld

@@
type T;
T *ptr;
identifier fld;
@@
- DIFF_OPT_TST(ptr, fld)
+ ptr->flags.fld

Signed-off-by: Brandon Williams <bmwill@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agodiff: remove touched flags
Brandon Williams [Tue, 31 Oct 2017 18:19:07 +0000 (11:19 -0700)]
diff: remove touched flags

Now that the set of parallel touched flags are no longer being used,
remove them.

Signed-off-by: Brandon Williams <bmwill@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agodiff: add flag to indicate textconv was set via cmdline
Brandon Williams [Tue, 31 Oct 2017 18:19:06 +0000 (11:19 -0700)]
diff: add flag to indicate textconv was set via cmdline

git-show is unique in that it wants to use textconv by default except
for when it is showing blobs.  When asked to show a blob, show doesn't
want to use textconv unless the user explicitly requested that it be
used by providing the command line flag '--textconv'.

Currently this is done by using a parallel set of 'touched' flags which
get set every time a particular flag is set or cleared.  In a future
patch we want to eliminate this parallel set of flags so instead of
relying on if the textconv flag has been touched, add a new flag
'TEXTCONV_SET_VIA_CMDLINE' which is only set if textconv is set to true
via the command line.

Signed-off-by: Brandon Williams <bmwill@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agodiff: convert flags to be stored in bitfields
Brandon Williams [Tue, 31 Oct 2017 18:19:05 +0000 (11:19 -0700)]
diff: convert flags to be stored in bitfields

We cannot add many more flags to the diff machinery due to the
limitations of the number of flags that can be stored in a single
unsigned int.  In order to allow for more flags to be added to the diff
machinery in the future this patch converts the flags to be stored in
bitfields in 'struct diff_flags'.

Signed-off-by: Brandon Williams <bmwill@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agosequencer: use O_TRUNC to truncate files
René Scharfe [Tue, 31 Oct 2017 09:58:16 +0000 (10:58 +0100)]
sequencer: use O_TRUNC to truncate files

Cut off any previous content of the file to be rewritten by passing the
flag O_TRUNC to open(2) instead of calling ftruncate(2) at the end.
That's easier and shorter.

Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agosequencer: factor out rewrite_file()
René Scharfe [Tue, 31 Oct 2017 09:54:21 +0000 (10:54 +0100)]
sequencer: factor out rewrite_file()

Reduce code duplication by extracting a function for rewriting an
existing file.

Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agot5580: add Cygwin support
Adam Dinwoodie [Tue, 31 Oct 2017 13:19:03 +0000 (13:19 +0000)]
t5580: add Cygwin support

t5580 tests that specifying Windows UNC paths works with Git.  Cygwin
supports UNC paths, albeit only using forward slashes, not backslashes,
so run the compatible tests on Cygwin as well as MinGW.

The only complication is Cygwin's `pwd`, which returns a *nix-style
path, and that's not suitable for calculating the UNC path to the
current directory.  Instead use Cygwin's `cygpath` utility to get the
Windows-style path.

Signed-off-by: Adam Dinwoodie <adam@dinwoodie.org>
Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agosha1_file: use hex_to_bytes()
René Scharfe [Tue, 31 Oct 2017 13:50:06 +0000 (14:50 +0100)]
sha1_file: use hex_to_bytes()

The path of a loose object contains its hash value encoded into two
substrings of 2 and 38 hexadecimal digits separated by a slash.  The
first part is handed to for_each_file_in_obj_subdir() in decoded form as
subdir_nr.  The current code builds a full hexadecimal representation of
the hash in a temporary buffer, then uses get_oid_hex() to decode it.

Avoid the intermediate step by taking subdir_nr as-is and using
hex_to_bytes() directly on the second substring.  That's shorter and
easier.

Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agohttp-push: use hex_to_bytes()
René Scharfe [Tue, 31 Oct 2017 13:49:56 +0000 (14:49 +0100)]
http-push: use hex_to_bytes()

The path of a loose object contains its hash value encoded into two
substrings of hexadecimal digits, separated by a slash.  The current
code copies the pieces into a temporary buffer to get rid of the slash
and then uses get_oid_hex() to decode the hash value.

Avoid the copy by using hex_to_bytes() directly on the substrings.
That's shorter and easier.

While at it correct the length of the second substring in a comment.

Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agonotes: move hex_to_bytes() to hex.c and export it
René Scharfe [Tue, 31 Oct 2017 13:46:49 +0000 (14:46 +0100)]
notes: move hex_to_bytes() to hex.c and export it

Make the function for converting pairs of hexadecimal digits to binary
available to other call sites.

Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agomerge-recursive: check GIT_MERGE_VERBOSITY only once
Andrey Okoshkin [Tue, 31 Oct 2017 09:09:13 +0000 (12:09 +0300)]
merge-recursive: check GIT_MERGE_VERBOSITY only once

Get rid of the duplicated getenv('GIT_MERGE_VERBOSITY') calls with the same
constant string argument. This makes code more readable and prevents typo in
the further development.

Signed-off-by: Andrey Okoshkin <a.okoshkin@samsung.com>
Reviewed-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoadd, reset: use DIFF_OPT_SET macro to set a diff flag
Brandon Williams [Mon, 30 Oct 2017 19:46:43 +0000 (12:46 -0700)]
add, reset: use DIFF_OPT_SET macro to set a diff flag

Instead of explicitly setting the 'DIFF_OPT_OVERRIDE_SUBMODULE_CONFIG'
flag, use the 'DIFF_OPT_SET' macro.

Signed-off-by: Brandon Williams <bmwill@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agostatus: test ignored modes
Jameson Miller [Mon, 30 Oct 2017 17:21:40 +0000 (13:21 -0400)]
status: test ignored modes

Add tests around status reporting ignord files that match an exclude
pattern for both --untracked-files=normal and --untracked-files=all.

Signed-off-by: Jameson Miller <jamill@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agostatus: document options to show matching ignored files
Jameson Miller [Mon, 30 Oct 2017 17:21:39 +0000 (13:21 -0400)]
status: document options to show matching ignored files

Signed-off-by: Jameson Miller <jamill@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agostatus: report matching ignored and normal untracked
Jameson Miller [Mon, 30 Oct 2017 17:21:38 +0000 (13:21 -0400)]
status: report matching ignored and normal untracked

Teach status command to handle `--ignored=matching` with
`--untracked-files=normal`

Signed-off-by: Jameson Miller <jamill@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agostatus: add option to show ignored files differently
Jameson Miller [Mon, 30 Oct 2017 17:21:37 +0000 (13:21 -0400)]
status: add option to show ignored files differently

Teach the status command more flexibility in how ignored files are
reported. Currently, the reporting of ignored files and untracked
files are linked. You cannot control how ignored files are reported
independently of how untracked files are reported (i.e. `all` vs
`normal`). This makes it impossible to show untracked files with the
`all` option, but show ignored files with the `normal` option.

This work 1) adds the ability to control the reporting of ignored
files independently of untracked files and 2) introduces the concept
of status reporting ignored paths that explicitly match an ignored
pattern. There are 2 benefits to these changes: 1) if a consumer needs
all untracked files but not all ignored files, there is a performance
benefit to not scanning all contents of an ignored directory and 2)
returning ignored files that explicitly match a path allow a consumer
to make more informed decisions about when a status result might be
stale.

This commit implements --ignored=matching with --untracked-files=all.
The following commit will implement --ignored=matching with
--untracked=files=normal.

As an example of where this flexibility could be useful is that our
application (Visual Studio) runs the status command and presents the
output. It shows all untracked files individually (e.g. using the
'--untracked-files==all' option), and would like to know about which
paths are ignored. It uses information about ignored paths to make
decisions about when the status result might have changed.
Additionally, many projects place build output into directories inside
a repository's working directory (e.g. in "bin/" and "obj/"
directories). Normal usage is to explicitly ignore these 2 directory
names in the .gitignore file (rather than or in addition to the *.obj
pattern).If an application could know that these directories are
explicitly ignored, it could infer that all contents are ignored as
well and make better informed decisions about files in these
directories. It could infer that any changes under these paths would
not affect the output of status. Additionally, there can be a
significant performance benefit by avoiding scanning through ignored
directories.

When status is set to report matching ignored files, it has the
following behavior. Ignored files and directories that explicitly
match an exclude pattern are reported. If an ignored directory matches
an exclude pattern, then the path of the directory is returned. If a
directory does not match an exclude pattern, but all of its contents
are ignored, then the contained files are reported instead of the
directory.

Signed-off-by: Jameson Miller <jamill@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agot0000: check whether the shell supports the "local" keyword
Michael Haggerty [Thu, 26 Oct 2017 08:18:53 +0000 (10:18 +0200)]
t0000: check whether the shell supports the "local" keyword

Add a test balloon to see if we get complaints from anybody who is
using a shell that doesn't support the "local" keyword. If so, this
test can be reverted. If not, we might want to consider using "local"
in shell code throughout the git code base.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoGit 2.15 v2.15.0
Junio C Hamano [Mon, 30 Oct 2017 05:00:44 +0000 (14:00 +0900)]
Git 2.15

Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agofiles-backend: don't rewrite the `packed-refs` file unnecessarily
Michael Haggerty [Sat, 28 Oct 2017 09:16:02 +0000 (11:16 +0200)]
files-backend: don't rewrite the `packed-refs` file unnecessarily

Even when we are deleting references, we needn't overwrite the
`packed-refs` file if the references that we are deleting only exist
as loose references. Implement this optimization as follows:

* Add a function `is_packed_transaction_needed()`, which checks
  whether a given packed-refs transaction actually needs to be carried
  out (i.e., it returns false if the transaction obviously wouldn't
  have any effect). This function must be called while holding the
  `packed-refs` lock to avoid races.

* Change `files_transaction_prepare()` to check whether the
  packed-refs transaction is actually needed. If not, squelch it, but
  continue holding the `packed-refs` lock until the end of the
  transaction to avoid races.

This fixes a mild regression caused by dc39e09942 (files_ref_store:
use a transaction to update packed refs, 2017-09-08). Before that
commit, unnecessary rewrites of `packed-refs` were suppressed by
`repack_without_refs()`. But the transaction-based writing introduced
by that commit didn't perform that optimization.

Note that the pre-dc39e09942 code still had to *read* the whole
`packed-refs` file to determine that the rewrite could be skipped, so
the performance for the cases that the write could be elided was
`O(N)` in the number of packed references both before and after
dc39e09942. But after that commit the constant factor increased.

This commit reimplements the optimization of eliding unnecessary
`packed-refs` rewrites. That, plus the fact that since
cfa2e29c34 (packed_ref_store: get rid of the `ref_cache` entirely,
2017-03-17) we don't necessarily have to read the whole `packed-refs`
file at all, means that deletes of one or a few loose references can
now be done with `O(n lg N)` effort, where `n` is the number of loose
references being deleted and `N` is the total number of packed
references.

This commit fixes two tests in t1409.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoMerge tag 'l10n-2.15.0-rnd2.1' of git://github.com/git-l10n/git-po
Junio C Hamano [Mon, 30 Oct 2017 00:32:54 +0000 (09:32 +0900)]
Merge tag 'l10n-2.15.0-rnd2.1' of git://github.com/git-l10n/git-po

l10n for Git 2.15.0 round 2 with Catalan updates

* tag 'l10n-2.15.0-rnd2.1' of git://github.com/git-l10n/git-po:
  l10n: Update Catalan translation

3 years agol10n: Update Catalan translation
Jordi Mas [Wed, 25 Oct 2017 17:50:59 +0000 (19:50 +0200)]
l10n: Update Catalan translation

Signed-off-by: Jordi Mas <jmas@softcatala.org>
3 years agodiff: fix lstat() error handling in diff_populate_filespec()
Andrey Okoshkin [Fri, 27 Oct 2017 09:33:25 +0000 (12:33 +0300)]
diff: fix lstat() error handling in diff_populate_filespec()

Add lstat() error handling not only for ENOENT case.
Otherwise uninitialised 'struct stat st' variable is used later in case of
lstat() non-ENOENT failure which leads to processing of rubbish values of
file mode ('S_ISLNK(st.st_mode)' check) or size ('xsize_t(st.st_size)').

Signed-off-by: Andrey Okoshkin <a.okoshkin@samsung.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years agoHopefully final batch before 2.15
Junio C Hamano [Sat, 28 Oct 2017 01:20:30 +0000 (10:20 +0900)]
Hopefully final batch before 2.15

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