git
4 years agot7700: replace egrep with grep
Denton Liu [Wed, 4 Dec 2019 22:03:19 +0000 (14:03 -0800)] 
t7700: replace egrep with grep

The egrep expressions in this test suite were of the form `^$variable`.
Although egrep works just fine, it's overkill since we're not using any
extended regex. Replace egrep invocations with grep so that we aren't
swatting flies with a sledgehammer.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot7700: consolidate code into test_has_duplicate_object()
Denton Liu [Wed, 4 Dec 2019 22:03:14 +0000 (14:03 -0800)] 
t7700: consolidate code into test_has_duplicate_object()

The code to test that objects were not duplicated from the packfile was
duplicated many times. Extract the duplicated code into
test_has_duplicate_object() and use that instead.

Refactor the resulting extraction so that if the git command fails,
the return code is not silently lost.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot7700: consolidate code into test_no_missing_in_packs()
Denton Liu [Wed, 4 Dec 2019 22:03:09 +0000 (14:03 -0800)] 
t7700: consolidate code into test_no_missing_in_packs()

The code to test that objects were not missing from the packfile was
duplicated many times. Extract the duplicated code into
test_no_missing_in_packs() and use that instead.

Refactor the resulting extraction so that if any git commands fail,
their return codes are not silently lost.

Instead of verifying each file of `alt_objects/pack/*.idx` individually
in a for-loop, batch them together into one verification step.

The original testing construct was O(n^2): it used a grep in a loop to
test whether any objects were missing in the packfile. Rewrite this to
extract the hash using sed or cut, sort the files, then use `comm -23`
so that finding missing lines from the original file is done more
efficiently.

While we're at it, add a space to `commit_and_pack ()` for style.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot7700: s/test -f/test_path_is_file/
Denton Liu [Wed, 27 Nov 2019 19:53:52 +0000 (11:53 -0800)] 
t7700: s/test -f/test_path_is_file/

Since we have debugging-friendly alternatives to `test -f`, replace
instances of `test -f` with `test_path_is_file` so that if a command
ever fails, we get better debugging information.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot7700: move keywords onto their own line
Denton Liu [Wed, 27 Nov 2019 19:53:49 +0000 (11:53 -0800)] 
t7700: move keywords onto their own line

The code style for tests is to have statements on their own line if
possible. Move keywords onto their own line so that they conform with
the test style.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot7700: remove spaces after redirect operators
Denton Liu [Wed, 27 Nov 2019 19:53:47 +0000 (11:53 -0800)] 
t7700: remove spaces after redirect operators

For shell scripts, the usual convention is for there to be no space
after redirection operators, (e.g. `>file`, not `> file`). Remove these
spaces wherever they appear.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot7700: drop redirections to /dev/null
Denton Liu [Wed, 27 Nov 2019 19:53:45 +0000 (11:53 -0800)] 
t7700: drop redirections to /dev/null

Since output is silenced when running without `-v` and debugging output
is useful with `-v`, remove redirections to /dev/null as it is not
useful.

In one case where the output of stdout is consumed, redirect the output
of test_commit to stderr.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot7501: stop losing return codes of git commands
Denton Liu [Wed, 27 Nov 2019 19:53:43 +0000 (11:53 -0800)] 
t7501: stop losing return codes of git commands

In a pipe, only the return code of the last command is used. Thus, all
other commands will have their return codes masked. Rewrite pipes so
that there are no git commands upstream so that we will know if a
command fails.

In the 'interactive add' test case, we prepend a `test_must_fail` to
`git commit --interactive`. When there are no changes to commit,
`git commit` will exit with status code 1. Following along with the rest
of the file, we use `test_must_fail` to test for this case.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot7501: remove spaces after redirect operators
Denton Liu [Wed, 27 Nov 2019 19:53:40 +0000 (11:53 -0800)] 
t7501: remove spaces after redirect operators

For shell scripts, the usual convention is for there to be no space
after redirection operators, (e.g. `>file`, not `> file`). Remove these
spaces wherever they appear.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot5703: stop losing return codes of git commands
Denton Liu [Wed, 27 Nov 2019 19:53:38 +0000 (11:53 -0800)] 
t5703: stop losing return codes of git commands

Currently, there are two ways where the return codes of git commands are
lost. The first way is when a command is in the upstream of a pipe. In a
pipe, only the return code of the last command is used. Thus, all other
commands will have their return codes masked. Rewrite pipes so that
there are no git commands upstream.

The other way is when a command is in a non-assignment command
substitution. The return code will be lost in favour of the surrounding
command's. Rewrite instances of this such that git commands are in an
assignment-only command substitution.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot5703: simplify one-time-sed generation logic
Denton Liu [Wed, 27 Nov 2019 19:53:36 +0000 (11:53 -0800)] 
t5703: simplify one-time-sed generation logic

In inconsistency(), we had two `git rev-parse` invocations in the
upstream of a pipe within a command substitution. In case this
invocation ever failed, its exit code would be swallowed up and we would
not know about it.

Pull the command substitutions out into variable assignments so that
their return codes are not lost.

Drop the pipe into `tr` because the $(...) substitution already takes
care of stripping out newlines, so the `tr` invocations in the code are
superfluous.

Finally, given the way the tests actually employ "one-time-sed" via
$(cat one-time-sed) in t/lib-httpd/apply-one-time-sed.sh, convert the
`printf` into an `echo`. This makes it consistent with the final "server
loses a ref - ref in want" test, which does use `echo` rather than
`printf`.

Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot5317: use ! grep to check for no matching lines
Denton Liu [Wed, 27 Nov 2019 19:53:33 +0000 (11:53 -0800)] 
t5317: use ! grep to check for no matching lines

Several times in t5317, we would use `wc -l` to ensure that a grep
result is empty. However, grep already has a way to do that... Its
return code! Use `! grep` in the cases where we are ensuring that there
are no matching lines.

While at it, drop unnecessary invocations of `awk` and `sort` in each
affected test since those commands do not influence the outcome. It's
not clear why that extra work was being done in the first place, and the
code's history doesn't shed any light on the matter since these tests
were simply born this way[1], likely due to copy-paste programming. The
unnecessary work wasn't noticed even when the code was later touched for
various cleanups[2][3].

[1]: 9535ce7337 (pack-objects: add list-objects filtering, 2017-11-21)
[2]: bdbc17e86a (tests: standardize pipe placement, 2018-10-05)
[3]: 61de0ff695 (tests: don't swallow Git errors upstream of pipes, 2018-10-05)

Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot5317: stop losing return codes of git commands
Denton Liu [Wed, 27 Nov 2019 19:53:31 +0000 (11:53 -0800)] 
t5317: stop losing return codes of git commands

Currently, there are two ways where the return codes of git commands are
lost. The first way is when a command is in the upstream of a pipe. In a
pipe, only the return code of the last command is used. Thus, all other
commands will have their return codes masked. Rewrite pipes so that
there are no git commands upstream.

The other way is when a command is in a non-assignment command
substitution. The return code will be lost in favour of the surrounding
command's. Rewrite instances of this such that git commands output to a
file and surrounding commands only call command substitutions with
non-git commands.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot4138: stop losing return codes of git commands
Denton Liu [Wed, 27 Nov 2019 19:53:29 +0000 (11:53 -0800)] 
t4138: stop losing return codes of git commands

In a pipe, only the return code of the last command is used. Thus, all
other commands will have their return codes masked. Rewrite pipes so
that there are no git commands upstream so that we will know if a
command fails.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot4015: use test_write_lines()
Denton Liu [Wed, 27 Nov 2019 19:53:26 +0000 (11:53 -0800)] 
t4015: use test_write_lines()

Instead of rolling our own method to write out some lines into a file,
use the existing test_write_lines().

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot4015: stop losing return codes of git commands
Denton Liu [Wed, 27 Nov 2019 19:53:24 +0000 (11:53 -0800)] 
t4015: stop losing return codes of git commands

Currently, there are two ways where the return codes of git commands are
lost. The first way is when a command is in the upstream of a pipe. In a
pipe, only the return code of the last command is used. Thus, all other
commands will have their return codes masked. Rewrite pipes so that
there are no git commands upstream.

The other way is when a command is in a non-assignment command
substitution. The return code will be lost in favour of the surrounding
command's. Rewrite instances of this so that git commands are either run
on their own or in an assignment-only command substitution.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot3600: comment on inducing SIGPIPE in `git rm`
Denton Liu [Wed, 27 Nov 2019 19:53:20 +0000 (11:53 -0800)] 
t3600: comment on inducing SIGPIPE in `git rm`

Add a comment about intentionally inducing SIGPIPE since this is unusual
and future developers should be aware. Also, even though we are trying
to refactor git commands out of the upstream of pipes, we cannot do it
here since we rely on it being upstream to induce SIGPIPE. Comment on
that as well so that future developers do not try to change it.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot3600: stop losing return codes of git commands
Denton Liu [Wed, 27 Nov 2019 19:53:18 +0000 (11:53 -0800)] 
t3600: stop losing return codes of git commands

When a command is in a non-assignment command substitution, the return
code will be lost in favour of the surrounding command's. As a result,
if a git command fails, we won't know about it. Rewrite instances of
this so that git commands are either run in an assignment-only command
substitution so that their return codes aren't lost.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot3600: use test_line_count() where possible
Denton Liu [Wed, 27 Nov 2019 19:53:16 +0000 (11:53 -0800)] 
t3600: use test_line_count() where possible

Since we have a helper function that can test the number of lines in a
file that gives better debugging information on failure, use
test_line_count() to test the number of lines.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot3301: stop losing return codes of git commands
Denton Liu [Wed, 27 Nov 2019 19:53:13 +0000 (11:53 -0800)] 
t3301: stop losing return codes of git commands

Currently, there are two ways where the return codes of git commands are
lost. The first way is when a command is in the upstream of a pipe. In a
pipe, only the return code of the last command is used. Thus, all other
commands will have their return codes masked. Rewrite pipes so that
there are no git commands upstream.

The other way is when a command is in a non-assignment command
substitution. The return code will be lost in favour of the surrounding
command's. Rewrite instances of this so that git commands are either run
on their own or in an assignment-only command substitution.

This patch fixes a real buggy test: in 'copy note with "git notes
copy"', `git notes` was mistyped as `git note`.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot0090: stop losing return codes of git commands
Denton Liu [Wed, 27 Nov 2019 19:53:11 +0000 (11:53 -0800)] 
t0090: stop losing return codes of git commands

In generate_expected_cache_tree_rec(), there are currently two instances
of `git ls-files` in the upstream of a pipe. In the case where the
upstream git command fails, its return code will be lost. Extract the
`git ls-files` into its own call so that if it ever fails, its return
code is not lost.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agot0014: remove git command upstream of pipe
Denton Liu [Wed, 27 Nov 2019 19:53:08 +0000 (11:53 -0800)] 
t0014: remove git command upstream of pipe

Before, the `git frotz` command would fail but its return code was
hidden since it was in the upstream of a pipe. Break the pipeline into
two commands so that the return code is no longer lost. Also, mark
`git frotz` with test_must_fail since it's supposed to fail.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agoapply-one-time-sed.sh: modernize style
Denton Liu [Wed, 27 Nov 2019 19:53:06 +0000 (11:53 -0800)] 
apply-one-time-sed.sh: modernize style

Convert `[ ... ]` to use `test` and test for the existence of a regular
file (`-f`) instead of any file (`-e`).

Move the `then`s onto their own lines so that it conforms with the
general test style.

Instead of redirecting input into sed, allow it to open its own input.

Use `cmp -s` instead of `diff` since we only care about whether the two
files are equal and `diff` is overkill for this.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4 years agolib-bash.sh: move `then` onto its own line
Denton Liu [Thu, 21 Nov 2019 00:45:48 +0000 (16:45 -0800)] 
lib-bash.sh: move `then` onto its own line

The code style for tests is to have statements on their own line if
possible. Move the `then` onto its own line so that it conforms with the
test style.

Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agoThe first batch post 2.24 cycle
Junio C Hamano [Sun, 10 Nov 2019 09:00:59 +0000 (18:00 +0900)] 
The first batch post 2.24 cycle

Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agoMerge branch 'bc/hash-independent-tests-part-6'
Junio C Hamano [Sun, 10 Nov 2019 09:02:17 +0000 (18:02 +0900)] 
Merge branch 'bc/hash-independent-tests-part-6'

Test updates to prepare for SHA-2 transition continues.

* bc/hash-independent-tests-part-6:
  t4048: abstract away SHA-1-specific constants
  t4045: make hash-size independent
  t4044: update test to work with SHA-256
  t4039: abstract away SHA-1-specific constants
  t4038: abstract away SHA-1 specific constants
  t4034: abstract away SHA-1-specific constants
  t4027: make hash-size independent
  t4015: abstract away SHA-1-specific constants
  t4011: abstract away SHA-1-specific constants
  t4010: abstract away SHA-1-specific constants
  t3429: remove SHA1 annotation
  t1305: avoid comparing extensions
  rev-parse: add a --show-object-format option
  t/oid-info: add empty tree and empty blob values
  t/oid-info: allow looking up hash algorithm name

5 years agoMerge branch 'js/update-index-ignore-removal-for-skip-worktree'
Junio C Hamano [Sun, 10 Nov 2019 09:02:16 +0000 (18:02 +0900)] 
Merge branch 'js/update-index-ignore-removal-for-skip-worktree'

"git stash save" in a working tree that is sparsely checked out
mistakenly removed paths that are outside the area of interest.

* js/update-index-ignore-removal-for-skip-worktree:
  stash: handle staged changes in skip-worktree files correctly
  update-index: optionally leave skip-worktree entries alone

5 years agoMerge branch 'pb/pretty-email-without-domain-part'
Junio C Hamano [Sun, 10 Nov 2019 09:02:16 +0000 (18:02 +0900)] 
Merge branch 'pb/pretty-email-without-domain-part'

The custom format for "git log --format=<format>" learned the l/L
placeholder that is similar to e/E that fills in the e-mail
address, but only the local part on the left side of '@'.

* pb/pretty-email-without-domain-part:
  pretty: add "%aL" etc. to show local-part of email addresses
  t4203: use test-lib.sh definitions
  t6006: use test-lib.sh definitions

5 years agoMerge branch 'hw/remove-api-docs-placeholder'
Junio C Hamano [Sun, 10 Nov 2019 09:02:15 +0000 (18:02 +0900)] 
Merge branch 'hw/remove-api-docs-placeholder'

Docfix.

* hw/remove-api-docs-placeholder:
  documentation: remove empty doc files

5 years agoMerge branch 'sg/commit-graph-usage-fix'
Junio C Hamano [Sun, 10 Nov 2019 09:02:15 +0000 (18:02 +0900)] 
Merge branch 'sg/commit-graph-usage-fix'

Message fix.

* sg/commit-graph-usage-fix:
  builtin/commit-graph.c: remove subcommand-less usage string

5 years agoMerge branch 'dl/apply-3way-diff3'
Junio C Hamano [Sun, 10 Nov 2019 09:02:15 +0000 (18:02 +0900)] 
Merge branch 'dl/apply-3way-diff3'

"git apply --3way" learned to honor merge.conflictStyle
configuration variable, like merges would.

* dl/apply-3way-diff3:
  apply: respect merge.conflictStyle in --3way
  t4108: demonstrate bug in apply
  t4108: use `test_config` instead of `git config`
  t4108: remove git command upstream of pipe
  t4108: replace create_file with test_write_lines

5 years agoMerge branch 'sg/dir-trie-fixes'
Junio C Hamano [Sun, 10 Nov 2019 09:02:14 +0000 (18:02 +0900)] 
Merge branch 'sg/dir-trie-fixes'

Code clean-up and a bugfix in the logic used to tell worktree local
and repository global refs apart.

* sg/dir-trie-fixes:
  path.c: don't call the match function without value in trie_find()
  path.c: clarify two field names in 'struct common_dir'
  path.c: mark 'logs/HEAD' in 'common_list' as file
  path.c: clarify trie_find()'s in-code comment
  Documentation: mention more worktree-specific exceptions

5 years agoMerge branch 'jc/am-show-current-patch-docfix'
Junio C Hamano [Sun, 10 Nov 2019 09:02:14 +0000 (18:02 +0900)] 
Merge branch 'jc/am-show-current-patch-docfix'

Doc update.

* jc/am-show-current-patch-docfix:
  doc: am --show-current-patch gives an entire e-mail message

5 years agoMerge branch 'wb/midx-progress'
Junio C Hamano [Sun, 10 Nov 2019 09:02:14 +0000 (18:02 +0900)] 
Merge branch 'wb/midx-progress'

The code to generate multi-pack index learned to show (or not to
show) progress indicators.

* wb/midx-progress:
  multi-pack-index: add [--[no-]progress] option.
  midx: honor the MIDX_PROGRESS flag in midx_repack
  midx: honor the MIDX_PROGRESS flag in verify_midx_file
  midx: add progress to expire_midx_packs
  midx: add progress to write_midx_file
  midx: add MIDX_PROGRESS flag

5 years agoMerge branch 'en/merge-recursive-directory-rename-fixes'
Junio C Hamano [Sun, 10 Nov 2019 09:02:13 +0000 (18:02 +0900)] 
Merge branch 'en/merge-recursive-directory-rename-fixes'

When all files from some subdirectory were renamed to the root
directory, the directory rename heuristics would fail to detect that
as a rename/merge of the subdirectory to the root directory, which has
been corrected.

* en/merge-recursive-directory-rename-fixes:
  t604[236]: do not run setup in separate tests
  merge-recursive: fix merging a subdirectory into the root directory
  merge-recursive: clean up get_renamed_dir_portion()

5 years agoMerge branch 'js/rebase-deprecate-preserve-merges'
Junio C Hamano [Sun, 10 Nov 2019 09:02:13 +0000 (18:02 +0900)] 
Merge branch 'js/rebase-deprecate-preserve-merges'

"git rebase --preserve-merges" has been marked as deprecated; this
release stops advertising it in the "git rebase -h" output.

* js/rebase-deprecate-preserve-merges:
  rebase: hide --preserve-merges option

5 years agoMerge branch 'hv/bitshift-constants-in-blame'
Junio C Hamano [Sun, 10 Nov 2019 09:02:12 +0000 (18:02 +0900)] 
Merge branch 'hv/bitshift-constants-in-blame'

Move the definition of a set of bitmask constants from 0ctal
literal to (1U<<count) notation.

* hv/bitshift-constants-in-blame:
  builtin/blame.c: constants into bit shift format

5 years agoMerge branch 'dd/notes-copy-default-dst-to-head'
Junio C Hamano [Sun, 10 Nov 2019 09:02:12 +0000 (18:02 +0900)] 
Merge branch 'dd/notes-copy-default-dst-to-head'

"git notes copy $original" ought to copy the notes attached to the
original object to HEAD, but a mistaken tightening to command line
parameter validation made earlier disabled that feature by mistake.

* dd/notes-copy-default-dst-to-head:
  notes: fix minimum number of parameters to "copy" subcommand
  t3301: test diagnose messages for too few/many paramters

5 years agoMerge branch 'pw/post-commit-from-sequencer'
Junio C Hamano [Sun, 10 Nov 2019 09:02:12 +0000 (18:02 +0900)] 
Merge branch 'pw/post-commit-from-sequencer'

"rebase -i" ceased to run post-commit hook by mistake in an earlier
update, which has been corrected.

* pw/post-commit-from-sequencer:
  sequencer: run post-commit hook
  move run_commit_hook() to libgit and use it there
  sequencer.h fix placement of #endif
  t3404: remove uneeded calls to set_fake_editor
  t3404: set $EDITOR in subshell
  t3404: remove unnecessary subshell

5 years agoMerge branch 'dl/format-patch-cover-from-desc'
Junio C Hamano [Sun, 10 Nov 2019 09:02:11 +0000 (18:02 +0900)] 
Merge branch 'dl/format-patch-cover-from-desc'

The branch description ("git branch --edit-description") has been
used to fill the body of the cover letters by the format-patch
command; this has been enhanced so that the subject can also be
filled.

* dl/format-patch-cover-from-desc:
  format-patch: teach --cover-from-description option
  format-patch: use enum variables
  format-patch: replace erroneous and condition

5 years agoMerge branch 'es/walken-tutorial'
Junio C Hamano [Sun, 10 Nov 2019 09:02:11 +0000 (18:02 +0900)] 
Merge branch 'es/walken-tutorial'

A tutorial on object enumeration.

* es/walken-tutorial:
  documentation: add tutorial for object walking

5 years agoMerge branch 'jt/fetch-pack-record-refs-in-the-dot-promisor'
Junio C Hamano [Sun, 10 Nov 2019 09:02:10 +0000 (18:02 +0900)] 
Merge branch 'jt/fetch-pack-record-refs-in-the-dot-promisor'

Debugging support for lazy cloning has been a bit improved.

* jt/fetch-pack-record-refs-in-the-dot-promisor:
  fetch-pack: write fetched refs to .promisor

5 years agoGit 2.24 v2.24.0
Junio C Hamano [Mon, 4 Nov 2019 04:32:41 +0000 (13:32 +0900)] 
Git 2.24

Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agoMerge branch 'bc/doc-use-docbook-5'
Junio C Hamano [Mon, 4 Nov 2019 04:33:06 +0000 (13:33 +0900)] 
Merge branch 'bc/doc-use-docbook-5'

Finishing touches to the recent update to the build procedure for
the documentation.

* bc/doc-use-docbook-5:
  manpage-bold-literal.xsl: match for namespaced "d:literal" in template

5 years agoMerge branch 'ds/commit-graph-on-fetch'
Junio C Hamano [Mon, 4 Nov 2019 04:33:06 +0000 (13:33 +0900)] 
Merge branch 'ds/commit-graph-on-fetch'

Regression fix.

* ds/commit-graph-on-fetch:
  commit-graph: fix writing first commit-graph during fetch
  t5510-fetch.sh: demonstrate fetch.writeCommitGraph bug

5 years agoMerge branch 'jt/delay-fetch-if-missing'
Junio C Hamano [Mon, 4 Nov 2019 04:33:05 +0000 (13:33 +0900)] 
Merge branch 'jt/delay-fetch-if-missing'

Work-around a lazy fetch glitch.

* jt/delay-fetch-if-missing:
  fetch: delay fetch_if_missing=0 until after config

5 years agoMerge https://github.com/prati0100/git-gui
Junio C Hamano [Mon, 4 Nov 2019 04:29:38 +0000 (13:29 +0900)] 
Merge https://github.com/prati0100/git-gui

* https://github.com/prati0100/git-gui:
  git-gui: improve Japanese translation
  git-gui: add a readme
  git-gui: support for diff3 conflict style
  git-gui: use existing interface to query a path's attribute
  git-gui (Windows): use git-bash.exe if it is available
  treewide: correct several "up-to-date" to "up to date"
  Fix build with core.autocrlf=true

5 years agoMerge tag 'l10n-2.24.0-rnd2' of https://github.com/git-l10n/git-po
Junio C Hamano [Mon, 4 Nov 2019 04:25:13 +0000 (13:25 +0900)] 
Merge tag 'l10n-2.24.0-rnd2' of https://github.com/git-l10n/git-po

l10n-2.24.0-rnd2

* tag 'l10n-2.24.0-rnd2' of https://github.com/git-l10n/git-po:
  l10n: zh_CN: for git v2.24.0 l10n round 1~2
  l10n: de.po: Update German translation
  l10n: sv.po: Update Swedish translation (4695t0f0u)
  l10n: bg.po: Updated Bulgarian translation (4694)
  l10n: vi(4694t): Updated translation for v2.24.0
  l10n: es: 2.24.0 round 2
  l10n: it.po: update the Italian translation for Git 2.24.0 round #2
  l10n: fr v2.24.0 rnd2
  l10n: git.pot: v2.24.0 round 2 (1 new)
  l10n: it.po: update the Italian translation for Git 2.24.0
  l10n: fr 2.24.0 rnd 1
  l10n: git.pot: v2.24.0 round 1 (35 new, 16 removed)
  l10n: bg.po: Updated Bulgarian translation (4693)
  l10n: sv.po: Update Swedish translation (4674t0f0u)
  l10n: Update Catalan translation

5 years agol10n: zh_CN: for git v2.24.0 l10n round 1~2
Jiang Xin [Tue, 29 Oct 2019 02:50:30 +0000 (10:50 +0800)] 
l10n: zh_CN: for git v2.24.0 l10n round 1~2

Translate 36 new messages (4694t0f0u) for git 2.24.0.

Signed-off-by: Jiang Xin <worldhello.net@gmail.com>
5 years agostash: handle staged changes in skip-worktree files correctly
Johannes Schindelin [Wed, 30 Oct 2019 10:49:38 +0000 (10:49 +0000)] 
stash: handle staged changes in skip-worktree files correctly

When calling `git stash` while changes were staged for files that are
marked with the `skip-worktree` bit (e.g. files that are excluded in a
sparse checkout), the files are recorded as _deleted_ instead.

The reason is that `git stash` tries to construct the tree reflecting
the worktree essentially by copying the index to a temporary one and
then updating the files from the worktree. Crucially, it calls `git
diff-index` to update also those files that are in the HEAD but have
been unstaged in the index.

However, when the temporary index is updated via `git update-index --add
--remove`, skip-worktree entries mark the files as deleted by mistake.

Let's use the newly-introduced `--ignore-skip-worktree-entries` option
of `git update-index` to prevent exactly this from happening.

Note that the regression test case deliberately avoids replicating the
scenario described above and instead tries to recreate just the symptom.

Reported by Dan Thompson.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agoupdate-index: optionally leave skip-worktree entries alone
Johannes Schindelin [Wed, 30 Oct 2019 10:49:37 +0000 (10:49 +0000)] 
update-index: optionally leave skip-worktree entries alone

While `git update-index` mostly ignores paths referring to index entries
whose skip-worktree bit is set, in b4d1690df11 (Teach Git to respect
skip-worktree bit (reading part), 2009-08-20), for reasons that are not
entirely obvious, the `--remove` option was made special: it _does_
remove index entries even if their skip-worktree bit is set.

Seeing as this behavior has been in place for a decade now, it does not
make sense to change it.

However, in preparation for fixing a bug in `git stash` where it
pretends that skip-worktree entries have actually been removed, we need
a mode where `git update-index` leaves all skip-worktree entries alone,
even if the `--remove` option was passed.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agoRelNotes/2.24.0: fix self-contradictory note
Elijah Newren [Wed, 30 Oct 2019 20:21:18 +0000 (20:21 +0000)] 
RelNotes/2.24.0: fix self-contradictory note

As per Wikipedia, "In current technical usage, for one to state that a
feature is deprecated is merely a recommendation against using it."  It
is thus contradictory to claim that something is not "officially
deprecated" and then to immediately state that we are both discouraging
its use and pointing people elsewhere.

Signed-off-by: Elijah Newren <newren@gmail.com>
Reviewed-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agomanpage-bold-literal.xsl: match for namespaced "d:literal" in template
Martin Ågren [Thu, 31 Oct 2019 06:22:27 +0000 (07:22 +0100)] 
manpage-bold-literal.xsl: match for namespaced "d:literal" in template

We recently regressed our rendering with Asciidoctor of "literal"
elements in our manpages, i.e, stuff we have placed within `backticks`
in order to render as monospace. In particular, we lost the bold
rendering of such literal text.

The culprit is f6461b82b9 ("Documentation: fix build with Asciidoctor 2",
2019-09-15), where we switched from DocBook 4.5 to DocBook 5 with
Asciidoctor. As part of the switch, we started using the namespaced
DocBook XSLT stylesheets rather than the non-namespaced ones. (See
f6461b82b9 for more details on why we changed to the namespaced ones.)

The bold literals are implemented as an XSLT snippet <xsl:template
match="literal">...</xsl:template>. Now that we use namespaces, this
doesn't pick up our literals like it used to.

Match for "d:literal" in addition to just "literal", after defining the
d namespace. ("d" is what
http://docbook.sourceforge.net/release/xsl-ns/current/manpages/docbook.xsl
uses.) Note that we need to keep matching without the namespace for
AsciiDoc.

This boldness was introduced by 5121a6d993 ("Documentation: option to
render literal text as bold for manpages", 2009-03-27) and made the
default in 5945717009 ("Documentation: bold literals in man",
2016-05-31).

One reason this was not caught in review is that our doc-diff tool diffs
without any boldness, i.e., it "only" compares text. As pointed out by
Peff in review of this patch, one can use `MAN_KEEP_FORMATTING=1
./doc-diff <...>`

This has been optically tested with AsciiDoc 8.6.10, Asciidoctor 1.5.5
and Asciidoctor 2.0.10. I've also verified that doc-diff produces the
empty output for all three programs, as expected, and that with the
MAN_KEEP_FORMATTING trick, AsciiDoc yields no diff, whereas with
Asciidoctor, we get bold literals, just like we want.

Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Acked-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agoRelNotes/2.24.0: typofix
Elijah Newren [Wed, 30 Oct 2019 20:21:17 +0000 (20:21 +0000)] 
RelNotes/2.24.0: typofix

Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agol10n: de.po: Update German translation
Matthias Rüster [Wed, 23 Oct 2019 17:13:19 +0000 (19:13 +0200)] 
l10n: de.po: Update German translation

Signed-off-by: Matthias Rüster <matthias.ruester@gmail.com>
Reviewed-by: Ralf Thielow <ralf.thielow@gmail.com>
Reviewed-by: Phillip Szelat <phillip.szelat@gmail.com>
5 years agol10n: sv.po: Update Swedish translation (4695t0f0u)
Peter Krefting [Wed, 30 Oct 2019 22:22:13 +0000 (23:22 +0100)] 
l10n: sv.po: Update Swedish translation (4695t0f0u)

Signed-off-by: Peter Krefting <peter@softwolves.pp.se>
5 years agoGit 2.24-rc2 v2.24.0-rc2
Junio C Hamano [Wed, 30 Oct 2019 06:12:53 +0000 (15:12 +0900)] 
Git 2.24-rc2

Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agoMerge branch 'wb/fsmonitor-bitmap-fix'
Junio C Hamano [Wed, 30 Oct 2019 06:13:13 +0000 (15:13 +0900)] 
Merge branch 'wb/fsmonitor-bitmap-fix'

Comment update.

* wb/fsmonitor-bitmap-fix:
  t7519-status-fsmonitor: improve comments

5 years agoMerge branch 'rl/gitweb-blame-prev-fix'
Junio C Hamano [Wed, 30 Oct 2019 06:13:13 +0000 (15:13 +0900)] 
Merge branch 'rl/gitweb-blame-prev-fix'

Fix a rather old bug in gitweb---incremental blame output in
javascript actions mode never worked.

* rl/gitweb-blame-prev-fix:
  gitweb: correctly store previous rev in javascript-actions mode

5 years agoMerge branch 'js/mingw-needs-hiding-fix'
Junio C Hamano [Wed, 30 Oct 2019 06:13:13 +0000 (15:13 +0900)] 
Merge branch 'js/mingw-needs-hiding-fix'

Fix for a (rather old) buffer-overrun bug.

* js/mingw-needs-hiding-fix:
  mingw: avoid a buffer overrun in `needs_hiding()`

5 years agoMerge branch 'master' of github.com:vnwildman/git
Jiang Xin [Wed, 30 Oct 2019 06:07:58 +0000 (14:07 +0800)] 
Merge branch 'master' of github.com:vnwildman/git

* 'master' of github.com:vnwildman/git:
  l10n: vi(4694t): Updated translation for v2.24.0

5 years agoMerge branch 'next' of github.com:ChrisADR/git-po
Jiang Xin [Wed, 30 Oct 2019 06:02:22 +0000 (14:02 +0800)] 
Merge branch 'next' of github.com:ChrisADR/git-po

* 'next' of github.com:ChrisADR/git-po:
  l10n: es: 2.24.0 round 2

5 years agot7519-status-fsmonitor: improve comments
William Baker [Wed, 16 Oct 2019 19:35:47 +0000 (19:35 +0000)] 
t7519-status-fsmonitor: improve comments

The comments for the staging/unstaging test did not accurately
describe the scenario being tested.  It is not essential that
the test files being staged/unstaged appear at the end of the
index.  All that is required is that the test files are not
flagged with CE_FSMONITOR_VALID and have a position in the
index greater than the number of entries in the index after
unstaging.

The comment for this test has been updated to be more
accurate with respect to the scenario that's being tested.

Signed-off-by: William Baker <William.Baker@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agopretty: add "%aL" etc. to show local-part of email addresses
Prarit Bhargava [Tue, 29 Oct 2019 12:09:14 +0000 (08:09 -0400)] 
pretty: add "%aL" etc. to show local-part of email addresses

In many projects the number of contributors is low enough that users know
each other and the full email address doesn't need to be displayed.
Displaying only the author's username saves a lot of columns on the screen.

Existing 'e/E' (as in "%ae" and "%aE") placeholders would show the
author's address as "prarit@redhat.com", which would waste columns to show
the same domain-part for all contributors when used in a project internal
to redhat.  Introduce 'l/L' placeholders that strip '@' and domain part from
the e-mail address.

Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agol10n: bg.po: Updated Bulgarian translation (4694)
Alexander Shopov [Tue, 29 Oct 2019 10:37:45 +0000 (11:37 +0100)] 
l10n: bg.po: Updated Bulgarian translation (4694)

Signed-off-by: Alexander Shopov <ash@kambanaria.org>
5 years agol10n: vi(4694t): Updated translation for v2.24.0
Tran Ngoc Quan [Tue, 29 Oct 2019 07:38:42 +0000 (14:38 +0700)] 
l10n: vi(4694t): Updated translation for v2.24.0

Signed-off-by: Tran Ngoc Quan <vnwildman@gmail.com>
5 years agol10n: es: 2.24.0 round 2
Christopher Diaz Riveros [Sun, 15 Sep 2019 21:56:56 +0000 (23:56 +0200)] 
l10n: es: 2.24.0 round 2

Signed-off-by: Christopher Diaz Riveros <christopher.diaz.riv@gmail.com>
5 years agoMerge branch 'l10n/it/update-italian-translation'
Jiang Xin [Mon, 28 Oct 2019 23:27:51 +0000 (07:27 +0800)] 
Merge branch 'l10n/it/update-italian-translation'

* 'update-italian-translation' of github.com:AlessandroMenti/git-po:
  l10n: it.po: update the Italian translation for Git 2.24.0 round #2

5 years agol10n: it.po: update the Italian translation for Git 2.24.0 round #2
Alessandro Menti [Mon, 28 Oct 2019 19:44:05 +0000 (20:44 +0100)] 
l10n: it.po: update the Italian translation for Git 2.24.0 round #2

Signed-off-by: Alessandro Menti <alessandro.menti@alessandromenti.it>
5 years agol10n: fr v2.24.0 rnd2
Jean-Noël Avila [Mon, 28 Oct 2019 19:36:01 +0000 (20:36 +0100)] 
l10n: fr v2.24.0 rnd2

Signed-off-by: Jean-Noël Avila <jn.avila@free.fr>
5 years agol10n: git.pot: v2.24.0 round 2 (1 new)
Jiang Xin [Mon, 28 Oct 2019 05:16:31 +0000 (13:16 +0800)] 
l10n: git.pot: v2.24.0 round 2 (1 new)

Generate po/git.pot from v2.24.0-rc1 for git v2.24.0 l10n round 2.

Signed-off-by: Jiang Xin <worldhello.net@gmail.com>
5 years agoMerge tag 'v2.24.0-rc1' of github.com:git/git into master
Jiang Xin [Mon, 28 Oct 2019 05:18:55 +0000 (13:18 +0800)] 
Merge tag 'v2.24.0-rc1' of github.com:git/git into master

Git 2.24-rc1

* tag 'v2.24.0-rc1' of github.com:git/git:
  Git 2.24-rc1
  repo-settings: read an int for index.version
  ci: fix GCC install in the Travis CI GCC OSX job
  Eleventh batch
  ci(osx): use new location of the `perforce` cask
  t7419: change test_must_fail to ! for grep
  t4014: make output-directory tests self-contained
  ci(visual-studio): actually run the tests in parallel
  ci(visual-studio): use strict compile flags, and optimization
  userdiff: fix some corner cases in dts regex
  test-progress: fix test failures on big-endian systems
  completion: clarify installation instruction for zsh
  grep: avoid leak of chartables in PCRE2
  grep: make PCRE2 aware of custom allocator
  grep: make PCRE1 aware of custom allocator
  remote-curl: pass on atomic capability to remote side
  diff-highlight: fix a whitespace nit
  fsmonitor: don't fill bitmap with entries to be removed

5 years agomingw: avoid a buffer overrun in `needs_hiding()`
Johannes Schindelin [Fri, 25 Oct 2019 14:13:36 +0000 (14:13 +0000)] 
mingw: avoid a buffer overrun in `needs_hiding()`

When this function is passed a path with a trailing slash, it runs right
over the end of that path.

Let's fix this.

Co-authored-by: Alexandr Miloslavskiy <alexandr.miloslavskiy@syntevo.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agobuiltin/commit-graph.c: remove subcommand-less usage string
SZEDER Gábor [Fri, 25 Oct 2019 16:49:09 +0000 (18:49 +0200)] 
builtin/commit-graph.c: remove subcommand-less usage string

The first line in 'git commit-graph's usage string indicates that this
command can be invoked without specifying a subcommand.  However, this
is not the case:

  $ git commit-graph
  usage: git commit-graph [--object-dir <objdir>]
     or: git commit-graph read [--object-dir <objdir>]
  [...]
  $ echo $?
  129

Remove this line from the usage string.

The synopsis in the manpage doesn't contain this line.

Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4048: abstract away SHA-1-specific constants
brian m. carlson [Mon, 28 Oct 2019 00:59:07 +0000 (00:59 +0000)] 
t4048: abstract away SHA-1-specific constants

Adjust the test so that it computes variables for object IDs instead of
using hard-coded hashes.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4045: make hash-size independent
brian m. carlson [Mon, 28 Oct 2019 00:59:06 +0000 (00:59 +0000)] 
t4045: make hash-size independent

Replace a hard-coded all-zeros object ID with a use of $ZERO_OID.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4044: update test to work with SHA-256
brian m. carlson [Mon, 28 Oct 2019 00:59:05 +0000 (00:59 +0000)] 
t4044: update test to work with SHA-256

This test produces pseudo-collisions and tests git diff's behavior with
them, and is therefore sensitive to the hash in use. Update the test to
compute the collisions for both SHA-1 and SHA-256 using appropriate
constants. Move the heredocs inside the setup block so that all of the
setup code can be tested for failure.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4039: abstract away SHA-1-specific constants
brian m. carlson [Mon, 28 Oct 2019 00:59:04 +0000 (00:59 +0000)] 
t4039: abstract away SHA-1-specific constants

Adjust the test so that it computes variables for object IDs instead of
using hard-coded hashes.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4038: abstract away SHA-1 specific constants
brian m. carlson [Mon, 28 Oct 2019 00:59:03 +0000 (00:59 +0000)] 
t4038: abstract away SHA-1 specific constants

Compute several object IDs that exist in expected output, since we don't
care about the specific object IDs, only that the format of the output
is syntactically correct.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4034: abstract away SHA-1-specific constants
brian m. carlson [Mon, 28 Oct 2019 00:59:02 +0000 (00:59 +0000)] 
t4034: abstract away SHA-1-specific constants

Adjust the test so that it computes variables for object IDs instead of
using hard-coded hashes.  Move some expected result heredocs around so
that they can use computed variables.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4027: make hash-size independent
brian m. carlson [Mon, 28 Oct 2019 00:59:01 +0000 (00:59 +0000)] 
t4027: make hash-size independent

Instead of hard-coding the length of an object ID, look this value up
using the translation tables.  Similarly, compute input data for invalid
submodule entries using the tables as well.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4015: abstract away SHA-1-specific constants
brian m. carlson [Mon, 28 Oct 2019 00:59:00 +0000 (00:59 +0000)] 
t4015: abstract away SHA-1-specific constants

Adjust the test so that it computes variables for object IDs instead of
using hard-coded hashes.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4011: abstract away SHA-1-specific constants
brian m. carlson [Mon, 28 Oct 2019 00:58:59 +0000 (00:58 +0000)] 
t4011: abstract away SHA-1-specific constants

Adjust the test so that it computes variables for object IDs instead of
using hard-coded hashes.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4010: abstract away SHA-1-specific constants
brian m. carlson [Mon, 28 Oct 2019 00:58:58 +0000 (00:58 +0000)] 
t4010: abstract away SHA-1-specific constants

Adjust the test so that it computes variables for object IDs instead of
using hard-coded hashes.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot3429: remove SHA1 annotation
brian m. carlson [Mon, 28 Oct 2019 00:58:57 +0000 (00:58 +0000)] 
t3429: remove SHA1 annotation

This test passes successfully with SHA-256, so remove the annotation
which limits it to SHA-1.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot1305: avoid comparing extensions
brian m. carlson [Mon, 28 Oct 2019 00:58:56 +0000 (00:58 +0000)] 
t1305: avoid comparing extensions

A repository using a hash other than SHA-1 will need to have an
extension in the config file.  Ignore any extensions when comparing
config files, since they don't usefully contribute to the goal of the
test.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agorev-parse: add a --show-object-format option
brian m. carlson [Mon, 28 Oct 2019 00:58:55 +0000 (00:58 +0000)] 
rev-parse: add a --show-object-format option

Add an option to print the object format used for input, output, or
storage. This allows shell scripts to discover the hash algorithm in
use.

Since the transition plan allows for multiple input algorithms, document
that we may provide multiple results for input, and the format that the
results may take. While we don't support this now, documenting it early
means that script authors can future-proof their scripts for when we do.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agogitweb: correctly store previous rev in javascript-actions mode
Robert Luberda [Sun, 27 Oct 2019 09:14:26 +0000 (10:14 +0100)] 
gitweb: correctly store previous rev in javascript-actions mode

Without this change, the setting

 $feature{'javascript-actions'}{'default'} = [1];

in gitweb.conf breaks gitweb's blame page: clicking on line numbers
displayed in the second column on the page has no effect.

For comparison, with javascript-actions disabled, clicking on line
numbers loads the previous version of the line.

Addresses https://bugs.debian.org/741883.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Robert Luberda <robert@debian.org>
Acked-by: Jakub Narębski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot4203: use test-lib.sh definitions
Prarit Bhargava [Thu, 24 Oct 2019 23:36:16 +0000 (19:36 -0400)] 
t4203: use test-lib.sh definitions

Use name and email definitions from test-lib.sh.

Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot6006: use test-lib.sh definitions
Prarit Bhargava [Thu, 24 Oct 2019 23:36:15 +0000 (19:36 -0400)] 
t6006: use test-lib.sh definitions

Use name and email definitions from test-lib.sh.

Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agocommit-graph: fix writing first commit-graph during fetch
Derrick Stolee [Thu, 24 Oct 2019 13:40:42 +0000 (13:40 +0000)] 
commit-graph: fix writing first commit-graph during fetch

The previous commit includes a failing test for an issue around
fetch.writeCommitGraph and fetching in a repo with a submodule. Here, we
fix that bug and set the test to "test_expect_success".

The problem arises with this set of commands when the remote repo at
<url> has a submodule. Note that --recurse-submodules is not needed to
demonstrate the bug.

$ git clone <url> test
$ cd test
$ git -c fetch.writeCommitGraph=true fetch origin
Computing commit graph generation numbers: 100% (12/12), done.
BUG: commit-graph.c:886: missing parent <hash1> for commit <hash2>
Aborted (core dumped)

As an initial fix, I converted the code in builtin/fetch.c that calls
write_commit_graph_reachable() to instead launch a "git commit-graph
write --reachable --split" process. That code worked, but is not how we
want the feature to work long-term.

That test did demonstrate that the issue must be something to do with
internal state of the 'git fetch' process.

The write_commit_graph() method in commit-graph.c ensures the commits we
plan to write are "closed under reachability" using close_reachable().
This method walks from the input commits, and uses the UNINTERESTING
flag to mark which commits have already been visited. This allows the
walk to take O(N) time, where N is the number of commits, instead of
O(P) time, where P is the number of paths. (The number of paths can be
exponential in the number of commits.)

However, the UNINTERESTING flag is used in lots of places in the
codebase. This flag usually means some barrier to stop a commit walk,
such as in revision-walking to compare histories. It is not often
cleared after the walk completes because the starting points of those
walks do not have the UNINTERESTING flag, and clear_commit_marks() would
stop immediately.

This is happening during a 'git fetch' call with a remote. The fetch
negotiation is comparing the remote refs with the local refs and marking
some commits as UNINTERESTING.

I tested running clear_commit_marks_many() to clear the UNINTERESTING
flag inside close_reachable(), but the tips did not have the flag, so
that did nothing.

It turns out that the calculate_changed_submodule_paths() method is at
fault. Thanks, Peff, for pointing out this detail! More specifically,
for each submodule, the collect_changed_submodules() runs a revision
walk to essentially do file-history on the list of submodules. That
revision walk marks commits UNININTERESTING if they are simplified away
by not changing the submodule.

Instead, I finally arrived on the conclusion that I should use a flag
that is not used in any other part of the code. In commit-reach.c, a
number of flags were defined for commit walk algorithms. The REACHABLE
flag seemed like it made the most sense, and it seems it was not
actually used in the file. The REACHABLE flag was used in early versions
of commit-reach.c, but was removed by 4fbcca4 (commit-reach: make
can_all_from_reach... linear, 2018-07-20).

Add the REACHABLE flag to commit-graph.c and use it instead of
UNINTERESTING in close_reachable(). This fixes the bug in manual
testing.

Reported-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Helped-by: Jeff King <peff@peff.net>
Helped-by: Szeder Gábor <szeder.dev@gmail.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agot5510-fetch.sh: demonstrate fetch.writeCommitGraph bug
Derrick Stolee [Thu, 24 Oct 2019 13:40:41 +0000 (13:40 +0000)] 
t5510-fetch.sh: demonstrate fetch.writeCommitGraph bug

While dogfooding, Johannes found a bug in the fetch.writeCommitGraph
config behavior. His example initially happened during a clone with
--recurse-submodules, we found that this happens with the first fetch
after cloning a repository that contains a submodule:

$ git clone <url> test
$ cd test
$ git -c fetch.writeCommitGraph=true fetch origin
Computing commit graph generation numbers: 100% (12/12), done.
BUG: commit-graph.c:886: missing parent <hash1> for commit <hash2>
Aborted (core dumped)

In the repo I had cloned, there were really 60 commits to scan, but
only 12 were in the list to write when calling
compute_generation_numbers(). A commit in the list expects to see a
parent, but that parent is not in the list.

A follow-up will fix the bug, but first we create a test that
demonstrates the problem. This test must be careful about an existing
commit-graph file, since GIT_TEST_COMMIT_GRAPH=1 will cause the repo we
are cloning to already have one. This then prevents the incremtnal
commit-graph write during the first 'git fetch'.

Helped-by: Jeff King <peff@peff.net>
Helped-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Helped-by: Szeder Gábor <szeder.dev@gmail.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agodocumentation: remove empty doc files
Heba Waly [Thu, 24 Oct 2019 11:29:11 +0000 (11:29 +0000)] 
documentation: remove empty doc files

Remove empty and redundant documentation files from the
Documentation/technical/ directory.

The empty doc files included only TODO messages with no documentation for
years. Instead an approach is being taken to keep all doc beside the code
in the relevant header files.
Having empty doc files is confusing and disappointing to anybody looking
for information, besides having the documentation in header files makes it
easier for developers to find the information they are looking for.

Some of the content which could have gone here already exists elsewhere:
- api-object-access.txt -> sha1-file.c and object.h have some details.
- api-quote.txt -> quote.h has some details.
- api-xdiff-interface.txt -> xdiff-interface.h has some details.
- api-grep.txt -> grep.h does not have enough documentation at the moment.

Signed-off-by: Heba Waly <heba.waly@gmail.com>
Reviewed-by: Emily Shaffer <emilyshaffer@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agoGit 2.24-rc1 v2.24.0-rc1
Junio C Hamano [Thu, 24 Oct 2019 04:33:49 +0000 (13:33 +0900)] 
Git 2.24-rc1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
5 years agoMerge branch 'sg/ci-osx-gcc8-fix'
Junio C Hamano [Thu, 24 Oct 2019 04:34:03 +0000 (13:34 +0900)] 
Merge branch 'sg/ci-osx-gcc8-fix'

CI build fix.

* sg/ci-osx-gcc8-fix:
  ci: fix GCC install in the Travis CI GCC OSX job

5 years agoMerge branch 'ds/feature-macros'
Junio C Hamano [Thu, 24 Oct 2019 04:34:02 +0000 (13:34 +0900)] 
Merge branch 'ds/feature-macros'

The codepath that reads the index.version configuration was broken
with a recent update, which has been corrected.

* ds/feature-macros:
  repo-settings: read an int for index.version

5 years agoMerge branch 'js/azure-ci-osx-fix'
Junio C Hamano [Thu, 24 Oct 2019 04:34:02 +0000 (13:34 +0900)] 
Merge branch 'js/azure-ci-osx-fix'

Update installation procedure for Perforce on MacOS in the CI jobs
running on Azure pipelines, which was failing.

* js/azure-ci-osx-fix:
  ci(osx): use new location of the `perforce` cask

5 years agoMerge branch 'bw/format-patch-o-create-leading-dirs'
Junio C Hamano [Thu, 24 Oct 2019 04:34:02 +0000 (13:34 +0900)] 
Merge branch 'bw/format-patch-o-create-leading-dirs'

Test update.

* bw/format-patch-o-create-leading-dirs:
  t4014: make output-directory tests self-contained

5 years agoMerge branch 'dl/submodule-set-branch'
Junio C Hamano [Thu, 24 Oct 2019 04:34:02 +0000 (13:34 +0900)] 
Merge branch 'dl/submodule-set-branch'

Test update.

* dl/submodule-set-branch:
  t7419: change test_must_fail to ! for grep

5 years agofetch: delay fetch_if_missing=0 until after config
Jonathan Tan [Wed, 23 Oct 2019 23:34:03 +0000 (16:34 -0700)] 
fetch: delay fetch_if_missing=0 until after config

Suppose, from a repository that has ".gitmodules", we clone with
--filter=blob:none:

  git clone --filter=blob:none --no-checkout \
    https://kernel.googlesource.com/pub/scm/git/git

Then we fetch:

  git -C git fetch

This will cause a "unable to load config blob object", because the
fetch_config_from_gitmodules() invocation in cmd_fetch() will attempt to
load ".gitmodules" (which Git knows to exist because the client has the
tree of HEAD) while fetch_if_missing is set to 0.

fetch_if_missing is set to 0 too early - ".gitmodules" here should be
lazily fetched.  Git must set fetch_if_missing to 0 before the fetch
because as part of the fetch, packfile negotiation happens (and we do
not want to fetch any missing objects when checking existence of
objects), but we do not need to set it so early. Move the setting of
fetch_if_missing to the earliest possible point in cmd_fetch(), right
before any fetching happens.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>