git/git.git
2 years agoMerge branch 'jc/no-grepping-for-strerror-in-tests'
Junio C Hamano [Thu, 14 Feb 2019 22:28:21 +0000 (14:28 -0800)]
Merge branch 'jc/no-grepping-for-strerror-in-tests'

* jc/no-grepping-for-strerror-in-tests:
  t1404: do not rely on the exact phrasing of strerror()

2 years agoMerge branch 'jt/fetch-v2-sideband'
Junio C Hamano [Thu, 14 Feb 2019 22:28:20 +0000 (14:28 -0800)]
Merge branch 'jt/fetch-v2-sideband'

"git fetch" and "git upload-pack" learned to send all exchange over
the sideband channel while talking the v2 protocol.

* jt/fetch-v2-sideband:
  t/lib-httpd: pass GIT_TEST_SIDEBAND_ALL through Apache

2 years agoMerge branch 'en/rebase-merge-on-sequencer'
Junio C Hamano [Thu, 14 Feb 2019 22:28:20 +0000 (14:28 -0800)]
Merge branch 'en/rebase-merge-on-sequencer'

"git rebase --merge" as been reimplemented by reusing the internal
machinery used for "git rebase -i".

* en/rebase-merge-on-sequencer:
  git-rebase.txt: update to reflect merge now implemented on sequencer

2 years agogit-rebase.txt: update to reflect merge now implemented on sequencer
Elijah Newren [Thu, 14 Feb 2019 20:25:41 +0000 (12:25 -0800)]
git-rebase.txt: update to reflect merge now implemented on sequencer

Since commit 8fe9c3f21dff (Merge branch 'en/rebase-merge-on-sequencer',
2019-02-06), --merge now uses the interactive backend (and matches its
behavior) so there is no separate merge backend anymore.  Fix an
oversight in the docs that should have been updated with the previous
change.

Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agot/lib-httpd: pass GIT_TEST_SIDEBAND_ALL through Apache
Todd Zullinger [Thu, 14 Feb 2019 06:35:13 +0000 (01:35 -0500)]
t/lib-httpd: pass GIT_TEST_SIDEBAND_ALL through Apache

07c3c2aa16 ("tests: define GIT_TEST_SIDEBAND_ALL", 2019-01-16) added
GIT_TEST_SIDEBAND_ALL to the apache.conf PassEnv list.  Avoid warnings
from Apache when the variable is unset, as we do for GIT_VALGRIND* and
GIT_TRACE, from f628825481 ("t/lib-httpd: handle running under
--valgrind", 2012-07-24) and 89c57ab3f0 ("t: pass GIT_TRACE through
Apache", 2015-03-13), respectively.

Signed-off-by: Todd Zullinger <tmz@pobox.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agot1404: do not rely on the exact phrasing of strerror()
Junio C Hamano [Thu, 14 Feb 2019 20:16:20 +0000 (12:16 -0800)]
t1404: do not rely on the exact phrasing of strerror()

Not even in C locale, it is wrong to expect that the exact phrasing
"File exists" is used to show EEXIST.

Reported-by: Randall S. Becker <rsbecker@nexbridge.com>
Helped-by: Duy Nguyen <pclouds@gmail.com>
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoGit 2.21-rc1 v2.21.0-rc1
Junio C Hamano [Thu, 14 Feb 2019 02:18:11 +0000 (18:18 -0800)]
Git 2.21-rc1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoMerge branch 'ab/rebase-test-fix'
Junio C Hamano [Thu, 14 Feb 2019 02:18:43 +0000 (18:18 -0800)]
Merge branch 'ab/rebase-test-fix'

* ab/rebase-test-fix:
  rebase: fix regression in rebase.useBuiltin=false test mode

2 years agoMerge branch 'rb/no-dev-zero-in-test'
Junio C Hamano [Thu, 14 Feb 2019 02:18:43 +0000 (18:18 -0800)]
Merge branch 'rb/no-dev-zero-in-test'

* rb/no-dev-zero-in-test:
  t5562: replace /dev/zero with a pipe from generate_zero_bytes
  t5318: replace use of /dev/zero with generate_zero_bytes
  test-lib-functions.sh: add generate_zero_bytes function

2 years agoMerge branch 'rs/bash-is-in-coreutils-on-nonstop'
Junio C Hamano [Thu, 14 Feb 2019 02:18:43 +0000 (18:18 -0800)]
Merge branch 'rs/bash-is-in-coreutils-on-nonstop'

* rs/bash-is-in-coreutils-on-nonstop:
  config.mak.uname: move location of bash on NonStop to CoreUtils

2 years agoMerge branch 'js/mingw-host-cpu'
Junio C Hamano [Thu, 14 Feb 2019 02:18:43 +0000 (18:18 -0800)]
Merge branch 'js/mingw-host-cpu'

Windows update.

* js/mingw-host-cpu:
  mingw: use a more canonical method to fix the CPU reporting

2 years agoMerge branch 'sg/stress-test'
Junio C Hamano [Thu, 14 Feb 2019 02:18:42 +0000 (18:18 -0800)]
Merge branch 'sg/stress-test'

Test improvement.

* sg/stress-test:
  test-lib: fix non-portable pattern bracket expressions
  test-lib: make '--stress' more bisect-friendly

2 years agoMerge branch 'kd/t0028-octal-del-is-377-not-777'
Junio C Hamano [Thu, 14 Feb 2019 02:18:42 +0000 (18:18 -0800)]
Merge branch 'kd/t0028-octal-del-is-377-not-777'

Test fix.

* kd/t0028-octal-del-is-377-not-777:
  t0028: fix wrong octal values for BOM in setup

2 years agoMerge branch 'bc/utf16-portability-fix'
Junio C Hamano [Thu, 14 Feb 2019 02:18:41 +0000 (18:18 -0800)]
Merge branch 'bc/utf16-portability-fix'

The code and tests assume that the system supplied iconv() would
always use BOM in its output when asked to encode to UTF-16 (or
UTF-32), but apparently some implementations output big-endian
without BOM.  A compile-time knob has been added to help such
systems (e.g. NonStop) to add BOM to the output to increase
portability.

* bc/utf16-portability-fix:
  utf8: handle systems that don't write BOM for UTF-16

2 years agoMerge branch 'nd/fileno-may-be-macro'
Junio C Hamano [Thu, 14 Feb 2019 02:18:41 +0000 (18:18 -0800)]
Merge branch 'nd/fileno-may-be-macro'

* nd/fileno-may-be-macro:
  git-compat-util: work around fileno(fp) that is a macro

2 years agoMerge branch 'nd/get-oid-with-context-returns-an-enum'
Junio C Hamano [Thu, 14 Feb 2019 02:18:41 +0000 (18:18 -0800)]
Merge branch 'nd/get-oid-with-context-returns-an-enum'

* nd/get-oid-with-context-returns-an-enum:
  get_oid_with_context(): match prototype and implementation

2 years agoMerge branch 'rj/sequencer-sign-off-header-static'
Junio C Hamano [Thu, 14 Feb 2019 02:18:41 +0000 (18:18 -0800)]
Merge branch 'rj/sequencer-sign-off-header-static'

Code clean-up.

* rj/sequencer-sign-off-header-static:
  sequencer: make sign_off_header a file local symbol

2 years agorebase: fix regression in rebase.useBuiltin=false test mode
Ævar Arnfjörð Bjarmason [Wed, 13 Feb 2019 21:49:08 +0000 (22:49 +0100)]
rebase: fix regression in rebase.useBuiltin=false test mode

Fix a recently introduced regression in c762aada1a ("rebase -x: sanity
check command", 2019-01-29) triggered when running the tests with
GIT_TEST_REBASE_USE_BUILTIN=false. See 62c23938fa ("tests: add a
special setup where rebase.useBuiltin is off", 2018-11-14) for how
that test mode works.

As discussed on-list[1] it's not worth it to implement the sanity
check in the legacy rebase code, we plan to remove it after the 2.21
release. So let's do the bare minimum to make the tests pass under the
GIT_TEST_REBASE_USE_BUILTIN=false special setup.

1. https://public-inbox.org/git/xmqqva1nbeno.fsf@gitster-ct.c.googlers.com/

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agomingw: use a more canonical method to fix the CPU reporting
Johannes Schindelin [Wed, 13 Feb 2019 10:19:49 +0000 (02:19 -0800)]
mingw: use a more canonical method to fix the CPU reporting

In `git version --build-options`, we report also the CPU, but in Git for
Windows we actually cross-compile the 32-bit version in a 64-bit Git for
Windows, so we cannot rely on the auto-detected value.

In 3815f64b0dd9 (mingw: fix CPU reporting in `git version
--build-options`, 2019-02-07), we fixed this by a Windows-only
workaround, making use of magic pre-processor constants, which works in
GCC, but most likely not all C compilers.

As pointed out by Eric Sunshine, there is a better way, anyway: to set
the Makefile variable HOST_CPU explicitly for cross-compiled Git. So
let's do that!

This reverts commit 3815f64b0dd983bdbf9242a0547706d5d81cb3e6 partially.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoconfig.mak.uname: move location of bash on NonStop to CoreUtils
Randall S. Becker [Sat, 9 Feb 2019 17:26:11 +0000 (12:26 -0500)]
config.mak.uname: move location of bash on NonStop to CoreUtils

The default bash is now officially in /usr/coreutils/bin instead
of in /usr/local/bin. This version of bash is more stable and
recommended for all use as of the J06.22 and L18.02 operating
system revision levels. This new version provides more stability
of test results.

Signed-off-by: Randall S. Becker <rsbecker@nexbridge.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agot5562: replace /dev/zero with a pipe from generate_zero_bytes
Randall S. Becker [Sat, 9 Feb 2019 18:59:30 +0000 (13:59 -0500)]
t5562: replace /dev/zero with a pipe from generate_zero_bytes

To help platforms that lack /dev/zero (e.g. NonStop), replace use
of /dev/zero to feed "git http-backend" with a pipe of output from
the generate_zero_bytes helper.

Signed-off-by: Randall S. Becker <rsbecker@nexbridge.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agot5318: replace use of /dev/zero with generate_zero_bytes
Randall S. Becker [Sat, 9 Feb 2019 18:59:29 +0000 (13:59 -0500)]
t5318: replace use of /dev/zero with generate_zero_bytes

There are platforms (e.g. NonStop) that lack /dev/zero; use the
generate_zero_bytes helper we just introduced to append stream
of NULs at the end of the file.

The original, even though it uses "dd seek=... count=..." to make it
look like it is overwriting the middle part of an existing file, has
truncated the file before this step with another use of "dd", which
may make it tricky to see why this rewrite is a correct one.

Signed-off-by: Randall S. Becker <rsbecker@nexbridge.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoget_oid_with_context(): match prototype and implementation
Duy Nguyen [Tue, 12 Feb 2019 12:43:23 +0000 (19:43 +0700)]
get_oid_with_context(): match prototype and implementation

The get_oid_with_context() function is declared to return an enum in
cache.h, but defined to return an int in sha1-name.c.  The compiler
notices this on AIX and rejects the build, since d1dd94b308 (Do not
print 'dangling' for cat-file in case of ambiguity - 2019-01-17) was
merged.

Return the correct type from the implementation to fix this.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agogit-compat-util: work around fileno(fp) that is a macro
Duy Nguyen [Tue, 12 Feb 2019 14:14:41 +0000 (21:14 +0700)]
git-compat-util: work around fileno(fp) that is a macro

On various BSD's, fileno(fp) is implemented as a macro that directly
accesses the fields in the FILE * object, which breaks a function that
accepts a "void *fp" parameter and calls fileno(fp) and expect it to
work.

Work it around by adding a compile-time knob FILENO_IS_A_MACRO that
inserts a real helper function in the middle of the callchain.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agotest-lib-functions.sh: add generate_zero_bytes function
Randall S. Becker [Sat, 9 Feb 2019 18:59:28 +0000 (13:59 -0500)]
test-lib-functions.sh: add generate_zero_bytes function

t5318 and t5562 used /dev/zero, which is not portable. This function
provides both a fixed block of NUL bytes and an infinite stream of NULs.

Signed-off-by: Randall S. Becker <rsbecker@nexbridge.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoRelNotes/2.21: misc typo/English fixups
Jeff King [Tue, 12 Feb 2019 06:33:04 +0000 (01:33 -0500)]
RelNotes/2.21: misc typo/English fixups

These are just some small fixes I noticed doing a complete read-through
(there are a few cases I left that are incomplete or abbreviated
sentences, but I think those are OK in this sort of bullet-list style).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoRelNotes/2.21: tweak "--date=auto" mention
Jeff King [Tue, 12 Feb 2019 06:32:48 +0000 (01:32 -0500)]
RelNotes/2.21: tweak "--date=auto" mention

In the feature that was eventually committed, "--date=auto" doesn't do
anything. It was generalized to "--date=auto:<format>".

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoMerge branch 'nd/imap-send-typofix'
Junio C Hamano [Tue, 12 Feb 2019 17:00:25 +0000 (09:00 -0800)]
Merge branch 'nd/imap-send-typofix'

* nd/imap-send-typofix:
  imap-send.c: add a missing space in error message

2 years agoimap-send.c: add a missing space in error message
Nguyễn Thái Ngọc Duy [Mon, 11 Feb 2019 09:40:11 +0000 (16:40 +0700)]
imap-send.c: add a missing space in error message

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoutf8: handle systems that don't write BOM for UTF-16
brian m. carlson [Tue, 12 Feb 2019 00:52:06 +0000 (00:52 +0000)]
utf8: handle systems that don't write BOM for UTF-16

When serializing UTF-16 (and UTF-32), there are three possible ways to
write the stream. One can write the data with a BOM in either big-endian
or little-endian format, or one can write the data without a BOM in
big-endian format.

Most systems' iconv implementations choose to write it with a BOM in
some endianness, since this is the most foolproof, and it is resistant
to misinterpretation on Windows, where UTF-16 and the little-endian
serialization are very common. For compatibility with Windows and to
avoid accidental misuse there, Git always wants to write UTF-16 with a
BOM, and will refuse to read UTF-16 without it.

However, musl's iconv implementation writes UTF-16 without a BOM,
relying on the user to interpret it as big-endian. This causes t0028 and
the related functionality to fail, since Git won't read the file without
a BOM.

Add a Makefile and #define knob, ICONV_OMITS_BOM, that can be set if the
iconv implementation has this behavior. When set, Git will write a BOM
manually for UTF-16 and UTF-32 and then force the data to be written in
UTF-16BE or UTF-32BE. We choose big-endian behavior here because the
tests use the raw "UTF-16" encoding, which will be big-endian when the
implementation requires this knob to be set.

Update the tests to detect this case and write test data with an added
BOM if necessary. Always write the BOM in the tests in big-endian
format, since all iconv implementations that omit a BOM must use
big-endian serialization according to the Unicode standard.

Preserve the existing behavior for systems which do not have this knob
enabled, since they may use optimized implementations, including
defaulting to the native endianness, which may improve performance.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agot0028: fix wrong octal values for BOM in setup
Kevin Daudt [Mon, 11 Feb 2019 21:38:18 +0000 (22:38 +0100)]
t0028: fix wrong octal values for BOM in setup

The setup code uses octal values with printf to generate a BOM for
UTF-16/32 BE/LE. It specifically uses '\777' to emit a 0xff byte. This
relies on the fact that most shells truncate the value above 0o377.

Ash however interprets '\777' as '\77' + a literal '7', resulting in an
invalid BOM.

Fix this by using the proper value of 0xff: '\377'.

Signed-off-by: Kevin Daudt <me@ikke.info>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agotest-lib: fix non-portable pattern bracket expressions
SZEDER Gábor [Mon, 11 Feb 2019 19:58:03 +0000 (20:58 +0100)]
test-lib: fix non-portable pattern bracket expressions

Use a '!' character to start a non-matching pattern bracket
expression, as specified by POSIX in Shell Command Language section
2.13.1 Patterns Matching a Single Character [1].

I used '^' instead in three places in the previous three commits, to
verify that the arguments of the '--stress=' and '--stress-limit='
options and the values of various '*_PORT' environment variables are
valid numbers.  With certain shells, at least with dash (upstream and
in Ubuntu 14.04) and mksh, this led to various undesired behaviors:

  # error message in case of a valid number
  $ ~/src/dash/src/dash ./t3903-stash.sh --stress=8
  error: --stress=<N> requires the number of jobs to run

  # not the expected error message
  $ ~/src/dash/src/dash ./t3903-stash.sh --stress=foo
  ./t3903-stash.sh: 238: test: Illegal number: foo

  # no error message at all?!
  $ mksh ./t3903-stash.sh --stress=foo
  $ echo $?
  0

Some other shells, e.g. Bash (even in posix mode), ksh, dash in Ubuntu
16.04 or later, are apparently happy to accept '^' just as well.

[1] http://pubs.opengroup.org/onlinepubs/009695399/utilities/xcu_chap02.html#tag_02_13

Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agosequencer: make sign_off_header a file local symbol
Ramsay Jones [Mon, 11 Feb 2019 17:16:58 +0000 (17:16 +0000)]
sequencer: make sign_off_header a file local symbol

Commit d0aaa46fd3 ("commit: move empty message checks to libgit",
2017-11-10) removes the last use of 'sign_off_header' outside of
the "sequencer.c" source file. Remove the extern declaration from
the header file and mark the definition of the symbol with the
static keyword.

Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoconfig.mak.uname: add FREAD_READS_DIRECTORIES for NonStop platform
Randall S. Becker [Sun, 10 Feb 2019 00:20:16 +0000 (19:20 -0500)]
config.mak.uname: add FREAD_READS_DIRECTORIES for NonStop platform

The NonStop platform needs this configuration item specified as
UnfortunatelyYes so that config directory files are correctly processed.

Signed-off-by: Randall S. Becker <rsbecker@nexbridge.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoFix typos in translatable strings for v2.21.0
Jean-Noël Avila [Mon, 11 Feb 2019 06:44:53 +0000 (07:44 +0100)]
Fix typos in translatable strings for v2.21.0

Signed-off-by: Jean-Noël Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoSeventh batch for 2.21
Junio C Hamano [Sat, 9 Feb 2019 04:45:48 +0000 (20:45 -0800)]
Seventh batch for 2.21

Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoMerge branch 'js/mingw-host-cpu'
Junio C Hamano [Sat, 9 Feb 2019 04:44:53 +0000 (20:44 -0800)]
Merge branch 'js/mingw-host-cpu'

Windows update.

* js/mingw-host-cpu:
  mingw: fix CPU reporting in `git version --build-options`

2 years agoMerge branch 'js/fuzz-commit-graph-update'
Junio C Hamano [Sat, 9 Feb 2019 04:44:53 +0000 (20:44 -0800)]
Merge branch 'js/fuzz-commit-graph-update'

Update to the fuzzer.

* js/fuzz-commit-graph-update:
  object: fix leak of shallow_stat
  fuzz-commit-graph: initialize repo object

2 years agoMerge branch 'kl/pretty-doc-markup-fix'
Junio C Hamano [Sat, 9 Feb 2019 04:44:52 +0000 (20:44 -0800)]
Merge branch 'kl/pretty-doc-markup-fix'

Doc update.

* kl/pretty-doc-markup-fix:
  doc: prevent overflowing <code> tag in rendered HTML

2 years agoMerge branch 'sg/ci-parallel-build'
Junio C Hamano [Sat, 9 Feb 2019 04:44:52 +0000 (20:44 -0800)]
Merge branch 'sg/ci-parallel-build'

Build update.

* sg/ci-parallel-build:
  ci: clear and mark MAKEFLAGS exported just once
  ci: make sure we build Git parallel

2 years agoMerge branch 'ld/git-p4-remove-flakey-test'
Junio C Hamano [Sat, 9 Feb 2019 04:44:52 +0000 (20:44 -0800)]
Merge branch 'ld/git-p4-remove-flakey-test'

A flakey "p4" test has been removed.

* ld/git-p4-remove-flakey-test:
  git-p4: remove ticket expiry test

2 years agoMerge branch 'js/rebase-i-redo-exec-fix'
Junio C Hamano [Sat, 9 Feb 2019 04:44:52 +0000 (20:44 -0800)]
Merge branch 'js/rebase-i-redo-exec-fix'

For "rebase -i --reschedule-failed-exec", we do not want the "-y"
shortcut after all.

* js/rebase-i-redo-exec-fix:
  Revert "rebase: introduce a shortcut for --reschedule-failed-exec"

2 years agoMerge branch 'nd/checkout-noisy-unmerge'
Junio C Hamano [Sat, 9 Feb 2019 04:44:51 +0000 (20:44 -0800)]
Merge branch 'nd/checkout-noisy-unmerge'

"git checkout [<tree-ish>] <pathspec>" started reporting the number
of paths that have got updated recently, but the same messages were
given when "git checkout -m <pathspec>" to unresolve conflicts that
have just been resolved.  The message now reports these unresolved
paths separately from the paths that are checked out from the index.

* nd/checkout-noisy-unmerge:
  checkout: count and print -m paths separately
  checkout: update count-checkouts messages

2 years agoMerge branch 'js/smart-http-detect-remote-error'
Junio C Hamano [Sat, 9 Feb 2019 04:44:51 +0000 (20:44 -0800)]
Merge branch 'js/smart-http-detect-remote-error'

Some errors from the other side coming over smart HTTP transport
were not noticed, which has been corrected.

* js/smart-http-detect-remote-error:
  t5551: test server-side ERR packet
  remote-curl: tighten "version 2" check for smart-http
  remote-curl: refactor smart-http discovery

2 years agoMerge branch 'ds/coverage-prove'
Junio C Hamano [Sat, 9 Feb 2019 04:44:51 +0000 (20:44 -0800)]
Merge branch 'ds/coverage-prove'

A new target "coverage-prove" to run the coverage test under
"prove" has been added.

* ds/coverage-prove:
  Makefile: add coverage-prove target

2 years agoMerge branch 'tz/gpg-test-fix'
Junio C Hamano [Sat, 9 Feb 2019 04:44:50 +0000 (20:44 -0800)]
Merge branch 'tz/gpg-test-fix'

Test fix.

* tz/gpg-test-fix:
  t/lib-gpg: drop redundant killing of gpg-agent
  t/lib-gpg: quote path to ${GNUPGHOME}/trustlist.txt

2 years agoMerge branch 'os/rebase-runs-post-checkout-hook'
Junio C Hamano [Sat, 9 Feb 2019 04:44:50 +0000 (20:44 -0800)]
Merge branch 'os/rebase-runs-post-checkout-hook'

Test fix.

* os/rebase-runs-post-checkout-hook:
  t5403: correct bash ambiguous redirect error in subtest 8 by quoting $GIT_DIR

2 years agotest-lib: make '--stress' more bisect-friendly
SZEDER Gábor [Fri, 8 Feb 2019 11:50:45 +0000 (12:50 +0100)]
test-lib: make '--stress' more bisect-friendly

Let's suppose that a test somehow becomes flaky between 'master' and
'pu', and tends to fail within the first 50 repetitions when run with
'--stress'.  In such a case we could use 'git bisect' to find the
culprit: if the test script fails with '--stress', then the commit is
definitely bad, but if it survives, say, 300 repetitions, then we could
consider it good with reasonable confidence.

Unfortunately, all this could only be done manually, because
'--stress' would run the test script repeatedly for all eternity on a
good commit, and it would exit with success even when it found a
failure on a bad commit.

So let's make '--stress' usable with 'git bisect run':

  - Make it exit with failure if a failure is found.

  - Add the '--stress-limit=<N>' option to repeat the test script
    at most N times in each of the parallel jobs, and exit with
    success when the limit is reached.

And then we could simply run something like:

  $ git bisect start origin/pu master
  $ git bisect run sh -c 'make && cd t &&
                          ./t1234-foo.sh --stress --stress-limit=300'

Sure, as a brand new feature it won't be any useful right now, but in
a release or three most cooking topics will already contain this, so
we could automatically bisect at least newly introduced flakiness.

Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agot5403: correct bash ambiguous redirect error in subtest 8 by quoting $GIT_DIR
Randall S. Becker [Fri, 8 Feb 2019 11:32:33 +0000 (06:32 -0500)]
t5403: correct bash ambiguous redirect error in subtest 8 by quoting $GIT_DIR

The embedded blanks in the full path of the test git repository cased bash
to generate an ambugious redirect error.

Signed-off-by: Randall S. Becker <rsbecker@nexbridge.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years ago.mailmap: map Clemens Buchacher's mail addresses
Johannes Schindelin [Fri, 8 Feb 2019 10:01:20 +0000 (02:01 -0800)]
.mailmap: map Clemens Buchacher's mail addresses

We have three email addresses for Clemens in our commit history, two of
them bouncing. Let's map the latter to the only one that still works.

Pointed out by Gábor Szeder.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agot/lib-gpg: drop redundant killing of gpg-agent
Todd Zullinger [Fri, 8 Feb 2019 03:17:46 +0000 (22:17 -0500)]
t/lib-gpg: drop redundant killing of gpg-agent

In 53fc999306 ("gpg-interface t: extend the existing GPG tests with
GPGSM", 2018-07-20), the gpgconf call which kills gpg-agent was copied
from the existing gpg setup code.

The reason for killing gpg-agent is given in 29ff1f8f74 ("t: lib-gpg:
flush gpg agent on startup", 2017-07-20):

  When running gpg-relevant tests, a gpg-daemon is spawned for each
  GNUPGHOME used. This daemon may stay running after the test and cache
  file descriptors for the trash directories, even after the trash
  directory is removed. This leads to ENOENT errors when attempting to
  create files if tests are run multiple times.

  Add a cleanup script to force flushing the gpg-agent for that GNUPGHOME
  (if any) before setting up the GPG relevant-environment.

Killing gpg-agent once per test is sufficient.

Signed-off-by: Todd Zullinger <tmz@pobox.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agot/lib-gpg: quote path to ${GNUPGHOME}/trustlist.txt
Todd Zullinger [Fri, 8 Feb 2019 03:17:45 +0000 (22:17 -0500)]
t/lib-gpg: quote path to ${GNUPGHOME}/trustlist.txt

When gpgsm is installed, lib-gpg.sh attempts to update trustlist.txt to
relax the checking of some root certificate requirements.  The path to
"${GNUPGHOME}" contains spaces which cause an "ambiguous redirect"
warning when bash is used to run the tests:

  $ bash t7030-verify-tag.sh
  /git/t/lib-gpg.sh: line 66: ${GNUPGHOME}/trustlist.txt: ambiguous redirect
  ok 1 - create signed tags
  ok 2 # skip create signed tags x509  (missing GPGSM)
  ...

No warning is issued when using bash called as /bin/sh, dash, or mksh.

Quote the path to ensure the redirect works as intended and sets the
GPGSM prereq.  While we're here, drop the space after ">>".

Signed-off-by: Todd Zullinger <tmz@pobox.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agomingw: fix CPU reporting in `git version --build-options`
Johannes Schindelin [Thu, 7 Feb 2019 10:46:06 +0000 (02:46 -0800)]
mingw: fix CPU reporting in `git version --build-options`

We cannot rely on `uname -m` in Git for Windows' SDK to tell us what
architecture we are compiling for, as we can compile both 32-bit and
64-bit `git.exe` from a 64-bit SDK, but the `uname -m` in that SDK will
always report `x86_64`.

So let's go back to our original design. And make it explicitly
Windows-specific.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoobject: fix leak of shallow_stat
Josh Steadmon [Thu, 7 Feb 2019 20:05:54 +0000 (12:05 -0800)]
object: fix leak of shallow_stat

In eee4502baaf ("shallow: migrate shallow information into the object
parser", 2018-05-17), we added a stat_validity pointer into the
parsed_object_pool struct, but did not add code to free this in
parsed_object_pool_clear(). This leak was found by fuzz-commit-graph.

Clear the struct and then free it in parsed_object_pool_clear() to
prevent the leak.

Signed-off-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agofuzz-commit-graph: initialize repo object
Josh Steadmon [Thu, 7 Feb 2019 20:05:53 +0000 (12:05 -0800)]
fuzz-commit-graph: initialize repo object

Various #DEFINE "constants" in commit-graph.c now depend on
the_hash_algo->rawsz, but this object must be initialized before it can
be used.

Signed-off-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agodoc: prevent overflowing <code> tag in rendered HTML
Katrin Leinweber [Thu, 7 Feb 2019 19:16:37 +0000 (20:16 +0100)]
doc: prevent overflowing <code> tag in rendered HTML

Add an apparently missing back-tick to fix a multi-line <code> section
on https://git-scm.com/docs/git-log which seems to have been caused by
commit 18fb7ffc ("pretty: respect color settings [...]", 2017-07-13).

Signed-off-by: Katrin Leinweber <katrin.leinweber@uni-konstanz.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoci: clear and mark MAKEFLAGS exported just once
Junio C Hamano [Thu, 7 Feb 2019 19:36:28 +0000 (11:36 -0800)]
ci: clear and mark MAKEFLAGS exported just once

Clearing it once upfront, and turning all the assignment into
appending, would future-proof the code even more, to prevent
mistakes the previous one fixed from happening again.

Also, mark the variable exported just once at the beginning.  There
is no point in marking it exported repeatedly.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoci: make sure we build Git parallel
SZEDER Gábor [Thu, 7 Feb 2019 18:37:36 +0000 (19:37 +0100)]
ci: make sure we build Git parallel

Commit 2c8921db2b (travis-ci: build with the right compiler,
2019-01-17) started to use MAKEFLAGS to specify which compiler to use
to build Git.  A bit later, and in a different topic branch commit
eaa62291ff (ci: inherit --jobs via MAKEFLAGS in run-build-and-tests,
2019-01-27) started to use MAKEFLAGS as well.  Unfortunately, there is
a semantic conflict between these two commits: both of them set
MAKEFLAGS, and since the line adding CC from 2c8921db2b comes later in
'ci/lib.sh', it overwrites the number of parallel jobs added in
eaa62291ff.

Consequently, since both commits have been merged all our CI jobs have
been building Git, building its documentation, and applying semantic
patches sequentially, making all build jobs a bit slower.  Running
the test suite is unaffected, because the number of test jobs comes
from GIT_PROVE_OPTS.

Append to MAKEFLAGS when setting the compiler to use, to ensure that
the number of parallel jobs to use is preserved.

Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoGit 2.21-rc0 v2.21.0-rc0
Junio C Hamano [Thu, 7 Feb 2019 05:30:09 +0000 (21:30 -0800)]
Git 2.21-rc0

Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoMerge branch 'jk/add-ignore-errors-bit-assignment-fix'
Junio C Hamano [Thu, 7 Feb 2019 06:05:30 +0000 (22:05 -0800)]
Merge branch 'jk/add-ignore-errors-bit-assignment-fix'

A hotfix to an incomplete fix made earlier.

* jk/add-ignore-errors-bit-assignment-fix:
  add_to_index(): convert forgotten HASH_RENORMALIZE check

2 years agoMerge branch 'rj/sparse-flags'
Junio C Hamano [Thu, 7 Feb 2019 06:05:30 +0000 (22:05 -0800)]
Merge branch 'rj/sparse-flags'

Use of the sparse tool got easier to customize from the command
line to help developers.

* rj/sparse-flags:
  Makefile: improve SPARSE_FLAGS customisation
  config.mak.uname: remove obsolete SPARSE_FLAGS setting

2 years agoMerge branch 'bc/fetch-pack-clear-alternate-shallow'
Junio C Hamano [Thu, 7 Feb 2019 06:05:30 +0000 (22:05 -0800)]
Merge branch 'bc/fetch-pack-clear-alternate-shallow'

"git fetch" over protocol v2 that needs to make a second connection
to backfill tags did not clear a variable that holds shallow
repository information correctly, leading to an access of freed
piece of memory.

* bc/fetch-pack-clear-alternate-shallow:
  fetch-pack: clear alternate shallow in one more place
  fetch-pack: clear alternate shallow when complete

2 years agoMerge branch 'ma/doc-diff-usage-fix'
Junio C Hamano [Thu, 7 Feb 2019 06:05:29 +0000 (22:05 -0800)]
Merge branch 'ma/doc-diff-usage-fix'

Running "Documentation/doc-diff x" from anywhere other than the
top-level of the working tree did not show the usage string
correctly, which has been fixed.

* ma/doc-diff-usage-fix:
  doc-diff: don't `cd_to_toplevel`

2 years agoMerge branch 'ab/diff-tree-doc-fix'
Junio C Hamano [Thu, 7 Feb 2019 06:05:29 +0000 (22:05 -0800)]
Merge branch 'ab/diff-tree-doc-fix'

Doc fix.

* ab/diff-tree-doc-fix:
  diff-tree doc: correct & remove wrong documentation

2 years agoMerge branch 'ss/describe-dirty-in-the-right-directory'
Junio C Hamano [Thu, 7 Feb 2019 06:05:29 +0000 (22:05 -0800)]
Merge branch 'ss/describe-dirty-in-the-right-directory'

"git --work-tree=$there --git-dir=$here describe --dirty" did not
work correctly as it did not pay attention to the location of the
worktree specified by the user by mistake, which has been
corrected.

* ss/describe-dirty-in-the-right-directory:
  t6120: test for describe with a bare repository
  describe: setup working tree for --dirty

2 years agoMerge branch 'sg/travis-osx-brew-breakage-workaround'
Junio C Hamano [Thu, 7 Feb 2019 06:05:28 +0000 (22:05 -0800)]
Merge branch 'sg/travis-osx-brew-breakage-workaround'

The way the OSX build jobs updates its build environment used the
"--quiet" option to "brew update" command, but it wasn't all that
quiet to be useful.  The use of the option has been replaced with
an explicit redirection to the /dev/null (which incidentally would
have worked around a breakage by recent updates to homebrew, which
has fixed itself already).

* sg/travis-osx-brew-breakage-workaround:
  travis-ci: make the OSX build jobs' 'brew update' more quiet

2 years agoMerge branch 'nd/commit-doc'
Junio C Hamano [Thu, 7 Feb 2019 06:05:28 +0000 (22:05 -0800)]
Merge branch 'nd/commit-doc'

Doc update.

* nd/commit-doc:
  git-commit.txt: better description what it does

2 years agoMerge branch 'sl/const'
Junio C Hamano [Thu, 7 Feb 2019 06:05:27 +0000 (22:05 -0800)]
Merge branch 'sl/const'

Code cleanup.

* sl/const:
  various: tighten constness of some local variables

2 years agoMerge branch 'sh/submodule-summary-abbrev-fix'
Junio C Hamano [Thu, 7 Feb 2019 06:05:27 +0000 (22:05 -0800)]
Merge branch 'sh/submodule-summary-abbrev-fix'

The "git submodule summary" subcommand showed shortened commit
object names by mechanically truncating them at 7-hexdigit, which
has been improved to let "rev-parse --short" scale the length of
the abbreviation with the size of the repository.

* sh/submodule-summary-abbrev-fix:
  git-submodule.sh: shorten submodule SHA-1s using rev-parse

2 years agoMerge branch 'nd/help-align-command-desc'
Junio C Hamano [Thu, 7 Feb 2019 06:05:27 +0000 (22:05 -0800)]
Merge branch 'nd/help-align-command-desc'

Output from "git help" was not correctly aligned, which has been
fixed.

* nd/help-align-command-desc:
  help: align the longest command in the command listing

2 years agoMerge branch 'jk/loose-object-cache-oid'
Junio C Hamano [Thu, 7 Feb 2019 06:05:27 +0000 (22:05 -0800)]
Merge branch 'jk/loose-object-cache-oid'

Code clean-up.

* jk/loose-object-cache-oid:
  prefer "hash mismatch" to "sha1 mismatch"
  sha1-file: avoid "sha1 file" for generic use in messages
  sha1-file: prefer "loose object file" to "sha1 file" in messages
  sha1-file: drop has_sha1_file()
  convert has_sha1_file() callers to has_object_file()
  sha1-file: convert pass-through functions to object_id
  sha1-file: modernize loose header/stream functions
  sha1-file: modernize loose object file functions
  http: use struct object_id instead of bare sha1
  update comment references to sha1_object_info()
  sha1-file: fix outdated sha1 comment references

2 years agoMerge branch 'pw/rebase-x-sanity-check'
Junio C Hamano [Thu, 7 Feb 2019 06:05:26 +0000 (22:05 -0800)]
Merge branch 'pw/rebase-x-sanity-check'

"git rebase -x $cmd" did not reject multi-line command, even though
the command is incapable of handling such a command.  It now is
rejected upfront.

* pw/rebase-x-sanity-check:
  rebase -x: sanity check command

2 years agoMerge branch 'js/vsts-ci'
Junio C Hamano [Thu, 7 Feb 2019 06:05:26 +0000 (22:05 -0800)]
Merge branch 'js/vsts-ci'

Prepare to run test suite on Azure Pipeline.

* js/vsts-ci: (22 commits)
  test-date: drop unused parameter to getnanos()
  ci: parallelize testing on Windows
  ci: speed up Windows phase
  tests: optionally skip bin-wrappers/
  t0061: workaround issues with --with-dashes and RUNTIME_PREFIX
  tests: add t/helper/ to the PATH with --with-dashes
  mingw: try to work around issues with the test cleanup
  tests: include detailed trace logs with --write-junit-xml upon failure
  tests: avoid calling Perl just to determine file sizes
  README: add a build badge (status of the Azure Pipelines build)
  mingw: be more generous when wrapping up the setitimer() emulation
  ci: use git-sdk-64-minimal build artifact
  ci: add a Windows job to the Azure Pipelines definition
  Add a build definition for Azure DevOps
  ci/lib.sh: add support for Azure Pipelines
  tests: optionally write results as JUnit-style .xml
  test-date: add a subcommand to measure times in shell scripts
  ci: use a junction on Windows instead of a symlink
  ci: inherit --jobs via MAKEFLAGS in run-build-and-tests
  ci/lib.sh: encapsulate Travis-specific things
  ...

2 years agoMerge branch 'br/commit-tree-fully-spelled-gpg-sign-option'
Junio C Hamano [Thu, 7 Feb 2019 06:05:25 +0000 (22:05 -0800)]
Merge branch 'br/commit-tree-fully-spelled-gpg-sign-option'

The documentation of "git commit-tree" said that the command
understands "--gpg-sign" in addition to "-S", but the command line
parser did not know about the longhand, which has been corrected.

* br/commit-tree-fully-spelled-gpg-sign-option:
  commit-tree: add missing --gpg-sign flag
  t7510: invoke git as part of &&-chain

2 years agoMerge branch 'sb/submodule-abort-update-upon-config-failure'
Junio C Hamano [Thu, 7 Feb 2019 06:05:25 +0000 (22:05 -0800)]
Merge branch 'sb/submodule-abort-update-upon-config-failure'

"git submodule update" learned to abort early when core.worktree
for the submodule is not set correctly to prevent spreading damage.

* sb/submodule-abort-update-upon-config-failure:
  git-submodule: abort if core.worktree could not be set correctly

2 years agoMerge branch 'sg/travis-specific-cc'
Junio C Hamano [Thu, 7 Feb 2019 06:05:25 +0000 (22:05 -0800)]
Merge branch 'sg/travis-specific-cc'

The travis CI scripts have been corrected to build Git with the
compiler(s) of our choice.

* sg/travis-specific-cc:
  travis-ci: build with the right compiler
  travis-ci: switch to Xcode 10.1 macOS image
  travis-ci: don't be '--quiet' when running the tests
  .gitignore: ignore external debug symbols from GCC on macOS

2 years agoMerge branch 'ds/push-sparse-tree-walk'
Junio C Hamano [Thu, 7 Feb 2019 06:05:24 +0000 (22:05 -0800)]
Merge branch 'ds/push-sparse-tree-walk'

"git pack-objects" learned another algorithm to compute the set of
objects to send, that trades the resulting packfile off to save
traversal cost to favor small pushes.

* ds/push-sparse-tree-walk:
  pack-objects: create GIT_TEST_PACK_SPARSE
  pack-objects: create pack.useSparse setting
  revision: implement sparse algorithm
  list-objects: consume sparse tree walk
  revision: add mark_tree_uninteresting_sparse

2 years agoMerge branch 'tb/test-lint-sed-options'
Junio C Hamano [Thu, 7 Feb 2019 06:05:24 +0000 (22:05 -0800)]
Merge branch 'tb/test-lint-sed-options'

The test lint learned to catch non-portable "sed" options.

* tb/test-lint-sed-options:
  test-lint: only use only sed [-n] [-e command] [-f command_file]

2 years agoMerge branch 'lt/date-human'
Junio C Hamano [Thu, 7 Feb 2019 06:05:24 +0000 (22:05 -0800)]
Merge branch 'lt/date-human'

A new date format "--date=human" that morphs its output depending
on how far the time is from the current time has been introduced.
"--date=auto" can be used to use this new format when the output is
going to the pager or to the terminal and otherwise the default
format.

* lt/date-human:
  Add `human` date format tests.
  Add `human` format to test-tool
  Add 'human' date format documentation
  Replace the proposed 'auto' mode with 'auto:'
  Add 'human' date format

2 years agoMerge branch 'jk/autocrlf-overrides-eol-doc'
Junio C Hamano [Thu, 7 Feb 2019 06:05:23 +0000 (22:05 -0800)]
Merge branch 'jk/autocrlf-overrides-eol-doc'

Documentation around core.crlf has been updated.

* jk/autocrlf-overrides-eol-doc:
  docs/config: clarify "text property" in core.eol
  doc/gitattributes: clarify "autocrlf overrides eol"

2 years agoMerge branch 'jk/unused-parameter-cleanup'
Junio C Hamano [Thu, 7 Feb 2019 06:05:23 +0000 (22:05 -0800)]
Merge branch 'jk/unused-parameter-cleanup'

Code cleanup.

* jk/unused-parameter-cleanup:
  convert: drop path parameter from actual conversion functions
  convert: drop len parameter from conversion checks
  config: drop unused parameter from maybe_remove_section()
  show_date_relative(): drop unused "tz" parameter
  column: drop unused "opts" parameter in item_length()
  create_bundle(): drop unused "header" parameter
  apply: drop unused "def" parameter from find_name_gnu()
  match-trees: drop unused path parameter from score functions

2 years agoMerge branch 'nd/the-index-final'
Junio C Hamano [Thu, 7 Feb 2019 06:05:22 +0000 (22:05 -0800)]
Merge branch 'nd/the-index-final'

The assumption to work on the single "in-core index" instance has
been reduced from the library-ish part of the codebase.

* nd/the-index-final:
  cache.h: flip NO_THE_INDEX_COMPATIBILITY_MACROS switch
  read-cache.c: remove the_* from index_has_changes()
  merge-recursive.c: remove implicit dependency on the_repository
  merge-recursive.c: remove implicit dependency on the_index
  sha1-name.c: remove implicit dependency on the_index
  read-cache.c: replace update_index_if_able with repo_&
  read-cache.c: kill read_index()
  checkout: avoid the_index when possible
  repository.c: replace hold_locked_index() with repo_hold_locked_index()
  notes-utils.c: remove the_repository references
  grep: use grep_opt->repo instead of explict repo argument

2 years agoMerge branch 'js/rebase-am'
Junio C Hamano [Thu, 7 Feb 2019 06:05:22 +0000 (22:05 -0800)]
Merge branch 'js/rebase-am'

Instead of going through "git-rebase--am" scriptlet to use the "am"
backend, the built-in version of "git rebase" learned to drive the
"am" backend directly.

* js/rebase-am:
  built-in rebase: call `git am` directly
  rebase: teach `reset_head()` to optionally skip the worktree
  rebase: avoid double reflog entry when switching branches
  rebase: move `reset_head()` into a better spot

2 years agoMerge branch 'tt/bisect-in-c'
Junio C Hamano [Thu, 7 Feb 2019 06:05:22 +0000 (22:05 -0800)]
Merge branch 'tt/bisect-in-c'

More code in "git bisect" has been rewritten in C.

* tt/bisect-in-c:
  bisect--helper: `bisect_start` shell function partially in C
  bisect--helper: `get_terms` & `bisect_terms` shell function in C
  bisect--helper: `bisect_next_check` shell function in C
  bisect--helper: `check_and_set_terms` shell function in C
  wrapper: move is_empty_file() and rename it as is_empty_or_missing_file()
  bisect--helper: `bisect_write` shell function in C
  bisect--helper: `bisect_reset` shell function in C

2 years agoMerge branch 'tb/utf-16-le-with-explicit-bom'
Junio C Hamano [Thu, 7 Feb 2019 06:05:21 +0000 (22:05 -0800)]
Merge branch 'tb/utf-16-le-with-explicit-bom'

A new encoding UTF-16LE-BOM has been invented to force encoding to
UTF-16 with BOM in little endian byte order, which cannot be directly
generated by using iconv.

* tb/utf-16-le-with-explicit-bom:
  Support working-tree-encoding "UTF-16LE-BOM"

2 years agoMerge branch 'dt/cat-file-batch-ambiguous'
Junio C Hamano [Thu, 7 Feb 2019 06:05:21 +0000 (22:05 -0800)]
Merge branch 'dt/cat-file-batch-ambiguous'

"git cat-file --batch" reported a dangling symbolic link by
mistake, when it wanted to report that a given name is ambiguous.

* dt/cat-file-batch-ambiguous:
  t1512: test ambiguous cat-file --batch and --batch-output
  Do not print 'dangling' for cat-file in case of ambiguity

2 years agoMerge branch 'km/init-doc-typofix'
Junio C Hamano [Thu, 7 Feb 2019 06:05:21 +0000 (22:05 -0800)]
Merge branch 'km/init-doc-typofix'

Docfix.

* km/init-doc-typofix:
  init docs: correct a punctuation typo

2 years agoMerge branch 'en/rebase-merge-on-sequencer'
Junio C Hamano [Thu, 7 Feb 2019 06:05:20 +0000 (22:05 -0800)]
Merge branch 'en/rebase-merge-on-sequencer'

"git rebase --merge" as been reimplemented by reusing the internal
machinery used for "git rebase -i".

* en/rebase-merge-on-sequencer:
  rebase: implement --merge via the interactive machinery
  rebase: define linearization ordering and enforce it
  git-legacy-rebase: simplify unnecessary triply-nested if
  git-rebase, sequencer: extend --quiet option for the interactive machinery
  am, rebase--merge: do not overlook --skip'ed commits with post-rewrite
  t5407: add a test demonstrating how interactive handles --skip differently
  rebase: fix incompatible options error message
  rebase: make builtin and legacy script error messages the same

2 years agoadd_to_index(): convert forgotten HASH_RENORMALIZE check
Jeff King [Thu, 7 Feb 2019 02:00:22 +0000 (21:00 -0500)]
add_to_index(): convert forgotten HASH_RENORMALIZE check

Commit 9e5da3d055 (add: use separate ADD_CACHE_RENORMALIZE flag,
2019-01-17) switched out using HASH_RENORMALIZE in our flags field for a
new ADD_CACHE_RENORMALIZE flag. However, it forgot to convert one of the
checks for HASH_RENORMALIZE into the new flag, which totally broke "git
add --renormalize".

To make matters even more confusing, the resulting code would racily
pass the tests!  The forgotten check was responsible for defeating the
up-to-date check of the index entry. That meant that "git add
--renormalize" would refuse to renormalize things that appeared
stat-clean. But most of the time the test commands run fast enough that
the file mtime is the same as the index mtime. And thus we err on the
conservative side and re-hash the file, which is what "--renormalize"
would have wanted.

But if you're unlucky and cross that one-second boundary between writing
the file and writing the index (which is more likely to happen on a slow
or heavily-loaded system), then the file appears stat-clean. And
"--renormalize" would effectively do nothing.

The fix is straightforward: convert this check to use the right flag.

Noticed-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agofetch-pack: clear alternate shallow in one more place
brian m. carlson [Wed, 6 Feb 2019 23:59:37 +0000 (23:59 +0000)]
fetch-pack: clear alternate shallow in one more place

The previous one did not clear the variable in one codepath,
but we should aim to be complete.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
[jc: made a reroll into incremental, as the previous one already is
 in the next branch]
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agogit-p4: remove ticket expiry test
Luke Diamand [Wed, 6 Feb 2019 15:11:53 +0000 (15:11 +0000)]
git-p4: remove ticket expiry test

The git-p4 login ticket expiry test causes unreliable test
runs. Since the handling of ticket expiry in git-p4 is far
from polished anyway, let's remove it for now.

A better way to actually run the test is to create a python
"fake" version of "p4" which returns whatever expiry results
the test requires.

Ideally git-p4 would look at the expiry time before starting
any long operations, and cleanup gracefully if there is not
enough time left. But that's quite hard to do.

Signed-off-by: Luke Diamand <luke@diamand.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agocheckout: count and print -m paths separately
Nguyễn Thái Ngọc Duy [Wed, 6 Feb 2019 02:51:15 +0000 (09:51 +0700)]
checkout: count and print -m paths separately

Since 0f086e6dca (checkout: print something when checking out paths -
2018-11-13), this command reports how many paths have been updated
from what source (either from a tree, or from the index). I forget
that there's a third source: when -m is used, the merge conflict is
re-created (granted, also from the index, but it's not a straight copy
from the index).

Count and report unmerged paths separately. There's a bit more update
to avoid reporting:

   Recreated X merge conflicts
   Updated 0 paths from the index

The second line is unnecessary. Though if there's no conflict
recreation, we still report

   Updated 0 paths from the index

to make it clear we're not really doing anything.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agocheckout: update count-checkouts messages
Nguyễn Thái Ngọc Duy [Wed, 6 Feb 2019 02:51:14 +0000 (09:51 +0700)]
checkout: update count-checkouts messages

Commit 0f086e6dca [1] counts the number of files updated by "git
checkout -- <paths>" command and prints it. Later on 536ec1839d [2]
adds the ability to remove files in "git checkout -- <paths>". This is
still an update on worktree and should be reported to the user.

To prepare for such an update since that commit is on track to
'master' now, the messages are rephrased to avoid "checked out" which
does not imply file deletion.

[1] 0f086e6dca (checkout: print something when checking out paths -
    2018-11-13)
[2] 536ec1839d (entry: support CE_WT_REMOVE flag in checkout_entry -
    2018-12-20)

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agot5551: test server-side ERR packet
Josh Steadmon [Wed, 6 Feb 2019 19:19:10 +0000 (14:19 -0500)]
t5551: test server-side ERR packet

When a smart HTTP server sends an error message via pkt-line, we detect
the error due to using PACKET_READ_DIE_ON_ERR_PACKET. This case was
added by 2d103c31c2 (pack-protocol.txt: accept error packets in any
context, 2018-12-29), but not covered by tests.

Signed-off-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoremote-curl: tighten "version 2" check for smart-http
Jeff King [Wed, 6 Feb 2019 19:18:58 +0000 (14:18 -0500)]
remote-curl: tighten "version 2" check for smart-http

In a v2 smart-http conversation, the server should reply to our initial
request with a pkt-line saying "version 2". We check that with
starts_with(), but really that should be the only thing in that packet.
A response of "version 20" should not match.

Let's tighten this check to use strcmp(). Note that we don't need to
worry about a trailing newline here, because the ptk-line code will have
chomped it for us already.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoremote-curl: refactor smart-http discovery
Jeff King [Wed, 6 Feb 2019 19:18:48 +0000 (14:18 -0500)]
remote-curl: refactor smart-http discovery

After making initial contact with an http server, we have to decide if
the server supports smart-http, and if so, which version. Our rules are
a bit inconsistent:

  1. For v0, we require that the content-type indicates a smart-http
     response. We also require the response to look vaguely like a
     pkt-line starting with "#". If one of those does not match, we fall
     back to dumb-http.

     But according to our http protocol spec[1]:

       Dumb servers MUST NOT return a return type starting with
       `application/x-git-`.

     If we see the expected content-type, we should consider it
     smart-http. At that point we can parse the pkt-line for real, and
     complain if it is not syntactically valid.

  2. For v2, we do not actually check the content-type. Our v2 protocol
     spec says[2]:

       When using the http:// or https:// transport a client makes a
       "smart" info/refs request as described in `http-protocol.txt`[...]

     and the http spec is clear that for a smart-http response[3]:

       The Content-Type MUST be `application/x-$servicename-advertisement`.

     So it is required according to the spec.

These inconsistencies were easy to miss because of the way the original
code was written as an inline conditional. Let's pull it out into its
own function for readability, and improve a few things:

 - we now predicate the smart/dumb decision entirely on the presence of
   the correct content-type

 - we do a real pkt-line parse before deciding how to proceed (and die
   if it isn't valid)

 - use skip_prefix() for comparing service strings, instead of
   constructing expected output in a strbuf; this avoids dealing with
   memory cleanup

Note that this _is_ tightening what the client will allow. It's all
according to the spec, but it's possible that other implementations
might violate these. However, violating these particular rules seems
like an odd choice for a server to make.

[1] Documentation/technical/http-protocol.txt, l. 166-167
[2] Documentation/technical/protocol-v2.txt, l. 63-64
[3] Documentation/technical/http-protocol.txt, l. 247

Helped-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agotest-date: drop unused parameter to getnanos()
Jeff King [Wed, 6 Feb 2019 19:35:52 +0000 (14:35 -0500)]
test-date: drop unused parameter to getnanos()

The getnanos() helper always gets the current time from our
getnanotime() facility. The caller cannot override it via TEST_DATE_NOW,
and hence we simply ignore the "now" parameter to the function. Let's
remove it, as it may mislead callers into thinking it does something.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoRevert "rebase: introduce a shortcut for --reschedule-failed-exec"
Johannes Schindelin [Wed, 6 Feb 2019 18:45:17 +0000 (10:45 -0800)]
Revert "rebase: introduce a shortcut for --reschedule-failed-exec"

This patch was contributed only as a tentative "we could introduce a
convenient short option if we do not want to change the default behavior
in the long run" patch, opening the discussion whether other people
agree with deprecating the current behavior in favor of the rescheduling
behavior.

But the consensus on the Git mailing list was that it would make sense
to show a warning in the near future, and flip the default
rebase.rescheduleFailedExec to reschedule failed `exec` commands by
default. See e.g.
<CAGZ79kZL5CRqCDRb6B-EedUm8Z_i4JuSF2=UtwwdRXMitrrOBw@mail.gmail.com>

So let's back out that patch that added the `-y` short option that we
agreed was not necessary or desirable.

This reverts commit 81ef8ee75d5f348d3c71ff633d13d302124e1a5e.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoFifth batch for 2.21
Junio C Hamano [Tue, 5 Feb 2019 22:27:34 +0000 (14:27 -0800)]
Fifth batch for 2.21

Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years agoMerge branch 'sg/object-as-type-commit-graph-fix'
Junio C Hamano [Tue, 5 Feb 2019 22:26:19 +0000 (14:26 -0800)]
Merge branch 'sg/object-as-type-commit-graph-fix'

The commit-graph facility did not work when in-core objects that
are promoted from unknown type to commit (e.g. a commit that is
accessed via a tag that refers to it) were involved, which has been
corrected.

* sg/object-as-type-commit-graph-fix:
  object_as_type: initialize commit-graph-related fields of 'struct commit'