"git push --recurse-submodules $there HEAD:$target" was not
propagated down to the submodules, but now it is.

* bw/push-options-recursively-to-submodules:
  submodule--helper: teach push-check to handle HEAD

The "tag.pager" configuration variable was useless for those who
actually create tag objects, as it interfered with the use of an
editor.  A new mechanism has been introduced for commands to enable
pager depending on what operation is being carried out to fix this,
and then "git tag -l" is made to run pager by default.

If this works out OK, I think there are low-hanging fruits in
other commands like "git branch" that outputs long list in one mode
while taking input in another.

* ma/pager-per-subcommand-action:
  git.c: ignore pager.* when launching builtin as dashed external
  tag: change default of `pager.tag` to "on"
  tag: respect `pager.tag` in list-mode only
  t7006: add tests for how git tag paginates
  git.c: provide setup_auto_pager()
  git.c: let builtins opt for handling `pager.foo` themselves
  builtin.h: take over documentation from api-builtin.txt

"git log --tag=no-such-tag" showed log starting from HEAD, which
has been fixed---it now shows nothing.

* jk/rev-list-empty-input:
  revision: do not fallback to default when rev_input_given is set
  rev-list: don't show usage when we see empty ref patterns
  revision: add rev_input_given flag
  t6018: flesh out empty input/output rev-list tests

Some versions of GnuPG fails to kill gpg-agent it auto-spawned
and such a left-over agent can interfere with a test.  Work it
around by attempting to kill one before starting a new test.

* st/lib-gpg-kill-stray-agent:
  t: lib-gpg: flush gpg agent on startup

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

A test update.

* jt/t1450-fsck-corrupt-packfile:
  tests: ensure fsck fails on corrupt packfiles

Code clean-up.

* jb/t8008-cleanup:
  t8008: rely on rev-parse'd HEAD instead of sha1 value

Code cleanup.

* jt/subprocess-handshake:
  sub-process: refactor handshake to common function
  Documentation: migrate sub-process docs to header
  convert: add "status=delayed" to filter process protocol
  convert: refactor capabilities negotiation
  convert: move multiple file filter error handling to separate function
  convert: put the flags field before the flag itself for consistent style
  t0021: write "OUT <size>" only on success
  t0021: make debug log file name configurable
  t0021: keep filter log files on comparison

Code cleanup.

* dc/fmt-merge-msg-microcleanup:
  fmt-merge-msg: fix coding style

Doc update.

* ah/doc-wserrorhighlight:
  doc: add missing values "none" and "default" for diff.wsErrorHighlight

Code cleanup.

* cc/ref-is-hidden-microcleanup:
  refs: use skip_prefix() in ref_is_hidden()

API fix.

* js/run-process-parallel-api-fix:
  run_processes_parallel: change confusing task_cb convention

Code clean-up.

* rs/pack-objects-pbase-cleanup:
  pack-objects: remove unnecessary NULL check

Code clean-up.

* jt/fsck-code-cleanup:
  fsck: cleanup unused variable
  object: remove "used" field from struct object
  fsck: remove redundant parse_tree() invocation

Code clean-up.

* rs/stat-data-unaligned-reads-fix:
  dir: support platforms that require aligned reads

Code clean-up.

* rs/move-array:
  ls-files: don't try to prune an empty index
  apply: use COPY_ARRAY and MOVE_ARRAY in update_image()
  use MOVE_ARRAY
  add MOVE_ARRAY

Code clean-up.

* rs/bswap-ubsan-fix:
  bswap: convert get_be16, get_be32 and put_be32 to inline functions
  bswap: convert to unsigned before shifting in get_be32

A recently added test for the "credential-cache" helper revealed
that EOF detection done around the time the connection to the cache
daemon is torn down were flaky.  This was fixed by reacting to
ECONNRESET and behaving as if we got an EOF.

* dl/credential-cache-socket-in-xdg-cache:
  credential-cache: interpret an ECONNRESET as an EOF

When a directory is not readable, "gitweb" fails to build the
project list.  Work this around by skipping such a directory.

It might end up hiding a problem under the rug and a better
solution might be to loudly complain to the administrator pointing
out the problematic directory, but this will at least make it
"work".

* hb/gitweb-project-list:
  gitweb: skip unreadable subdirectories

"git commit" when seeing an totally empty message said "you did not
edit the message", which is clearly wrong.  The message has been
corrected.

* ks/commit-abort-on-empty-message-fix:
  commit: check for empty message before the check for untouched template

Numerous bugs in walking of reflogs via "log -g" and friends have
been fixed.

* jk/reflog-walk:
  reflog-walk: apply --since/--until to reflog dates
  reflog-walk: stop using fake parents
  rev-list: check reflog_info before showing usage
  get_revision_1(): replace do-while with an early return
  log: do not free parents when walking reflog
  log: clarify comment about reflog cycles
  revision: disallow reflog walking with revs->limited
  t1414: document some reflog-walk oddities

The http.{sslkey,sslCert} configuration variables are to be
interpreted as a pathname that honors "~[username]/" prefix, but
weren't, which has been fixed.

* jc/http-sslkey-and-ssl-cert-are-paths:
  http.c: http.sslcert and http.sslkey are both pathnames

"%C(color name)" in the pretty print format always produced ANSI
color escape codes, which was an early design mistake.  They now
honor the configuration (e.g. "color.ui = never") and also tty-ness
of the output medium.

* jk/ref-filter-colors:
  ref-filter: consult want_color() before emitting colors
  pretty: respect color settings for %C placeholders
  rev-list: pass diffopt->use_colors through to pretty-print
  for-each-ref: load config earlier
  color: check color.ui in git_default_config()
  ref-filter: pass ref_format struct to atom parsers
  ref-filter: factor out the parsing of sorting atoms
  ref-filter: make parse_ref_filter_atom a private function
  ref-filter: provide a function for parsing sort options
  ref-filter: move need_color_reset_at_eol into ref_format
  ref-filter: abstract ref format into its own struct
  ref-filter: simplify automatic color reset
  t: use test_decode_color rather than literal ANSI codes
  docs/for-each-ref: update pointer to color syntax
  check return value of verify_ref_format()

Because recent Git for Windows do come with a real msgfmt, the
build procedure for git-gui has been updated to use it instead of a
hand-rolled substitute.

* js/git-gui-msgfmt-on-windows:
  git-gui (MinGW): make use of MSys2's msgfmt
  git gui: allow for a long recentrepo list
  git gui: de-dup selected repo from recentrepo history
  git gui: cope with duplicates in _get_recentrepo
  git-gui: remove duplicate entries from .gitconfig's gui.recentrepo

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

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

Docfix.

* ah/patch-id-doc:
  doc: remove unsupported parameter from patch-id

A grammofix.

* as/diff-options-grammofix:
  diff-options doc: grammar fix

When running, e.g., `git -c alias.bar=foo bar`, we expand the alias and
execute `git-foo` as a dashed external. This is true even if git foo is
a builtin. That is on purpose, and is motivated in a comment which was
added in commit 441981bc ("git: simplify environment save/restore
logic", 2016-01-26).

Shortly before we launch a dashed external, and unless we have already
found out whether we should use a pager, we check `pager.foo`. This was
added in commit 92058e4d ("support pager.* for external commands",
2011-08-18). If the dashed external is a builtin, this does not match
that commit's intention and is arguably wrong, since it would be cleaner
if we let the "dashed external builtin" handle `pager.foo`.

This has not mattered in practice, but a recent patch taught `git-tag`
to ignore `pager.tag` under certain circumstances. But, when started
using an alias, it doesn't get the chance to do so, as outlined above.
That recent patch added a test to document this breakage.

Do not check `pager.foo` before launching a builtin as a dashed
external, i.e., if we recognize the name of the external as a builtin.
Change the test to use `test_expect_success`.
Signed-off-by: NMartin Ågren <martin.agren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

The previous patch taught `git tag` to only respect `pager.tag` in
list-mode. That patch left the default value of `pager.tag` at "off".

After that patch, it makes sense to let the default value be "on"
instead, since it will help with listing many tags, but will not hurt
users of `git tag -a` as it would have before. Make that change. Update
documentation and tests.
Signed-off-by: NMartin Ågren <martin.agren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Using, e.g., `git -c pager.tag tag -a new-tag` results in errors such as
"Vim: Warning: Output is not to a terminal" and a garbled terminal.
Someone who makes use of both `git tag -a` and `git tag -l` will
probably not set `pager.tag`, so that `git tag -a` will actually work,
at the cost of not paging output of `git tag -l`.

Use the mechanisms introduced in two earlier patches to ignore
`pager.tag` in git.c and let the `git tag` builtin handle it on its own.
Only respect `pager.tag` when running in list-mode.

There is a window between where the pager is started before and after
this patch. This means that early errors can behave slightly different
before and after this patch. Since operation-parsing has to happen
inside this window, this can be seen with `git -c pager.tag="echo pager
is used" tag -l --unknown-option`. This change in paging-behavior should
be acceptable since it only affects erroneous usages.

Update the documentation and update tests.

If an alias is used to run `git tag -a`, then `pager.tag` will still be
respected. Document this known breakage. It will be fixed in a later
commit. Add a similar test for `-l`, which works.
Noticed-by: NAnatoly Borodin <anatoly.borodin@gmail.com>
Suggested-by: NJeff King <peff@peff.net>
Signed-off-by: NMartin Ågren <martin.agren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Using, e.g., `git -c pager.tag tag -a new-tag` results in errors such as
"Vim: Warning: Output is not to a terminal" and a garbled terminal.
Someone who makes use of both `git tag -a` and `git tag -l` will
probably not set `pager.tag`, so that `git tag -a` will actually work,
at the cost of not paging output of `git tag -l`.

Since we're about to change how `git tag` respects `pager.tag`, add tests
around this, including how the configuration is ignored if --no-pager or
--paginate are used.

Construct tests with a few different subcommands. First, use -l. Second,
use "no arguments" and --contains, since those imply -l. (There are
more arguments which imply -l, but using these two should be enough.)

Third, use -a as a representative for "not -l". Actually, the tests use
`git tag -am` so no editor is launched, but that is irrelevant, since we
just want to see whether the pager is used or not. Make one of the tests
demonstrate the broken behavior mentioned above, where `git tag -a`
respects `pager.tag`.
Signed-off-by: NMartin Ågren <martin.agren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

The previous patch introduced a way for builtins to declare that they
will take responsibility for handling the `pager.foo`-config item. (See
the commit message of that patch for why that could be useful.)

Provide setup_auto_pager(), which builtins can call in order to handle
`pager.<cmd>`, including possibly starting the pager. Make this function
don't do anything if a pager has already been started, as indicated by
use_pager or pager_in_use().

Whenever this function is called from a builtin, git.c will already have
called commit_pager_choice(). Since commit_pager_choice() treats the
special value -1 as "punt" or "not yet decided", it is not a problem
that we might end up calling commit_pager_choice() once in git.c and
once (or more) in the builtin. Make the new function use -1 in the same
way and document it as "punt".

Don't add any users of setup_auto_pager just yet, one will follow in
a later patch.
Suggested-by: NJeff King <peff@peff.net>
Signed-off-by: NMartin Ågren <martin.agren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Before launching a builtin git foo and unless mechanisms with precedence
are in use, we check for and handle the `pager.foo` config. This is done
without considering exactly how git foo is being used, and indeed, git.c
cannot (and should not) know what the arguments to git foo are supposed
to achieve.

In practice this means that, e.g., `git -c pager.tag tag -a new-tag`
results in errors such as "Vim: Warning: Output is not to a terminal"
and a garbled terminal. Someone who makes use of both `git tag -a` and
`git tag -l` will probably not set `pager.tag`, so that `git tag -a`
will actually work, at the cost of not paging output of `git tag -l`.

To allow individual builtins to make more informed decisions about when
to respect `pager.foo`, introduce a flag DELAY_PAGER_CONFIG. If the flag
is set, do not check `pager.foo`.

Do not check for DELAY_PAGER_CONFIG in `execv_dashed_external()`. That
call site is arguably wrong, although in a way that is not yet visible,
and will be changed in a slightly different direction in a later patch.

Don't add any users of DELAY_PAGER_CONFIG just yet, one will follow in a
later patch.
Suggested-by: NJeff King <peff@peff.net>
Signed-off-by: NMartin Ågren <martin.agren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Delete Documentation/technical/api-builtin.txt and move its content
into builtin.h. Format it as a comment. Remove a '+' which was needed
when the information was formatted for AsciiDoc. Similarly, change
"::" to ":".

Document SUPPORT_SUPER_PREFIX, thereby bringing the documentation up to
date with the available flags.

While at it, correct '3 more things to do' to '4 more things to do'.
Signed-off-by: NMartin Ågren <martin.agren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

If revs->def is set (as it is in "git log") and there are no
pending objects after parsing the user's input, then we show
whatever is in "def". But if the user _did_ ask for some
input that just happened to be empty (e.g., "--glob" that
does not match anything), showing the default revision is
confusing. We should just show nothing, as that is what the
user's request yielded.
Signed-off-by: NJeff King <peff@peff.net>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

If the user gives us no starting point for a traversal, we
want to complain with our normal usage message. But if they
tried to do so with "--all" or "--glob", but that happened
not to match any refs, the usage message isn't helpful. We
should just give them the empty output they asked for
instead.
Signed-off-by: NJeff King <peff@peff.net>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Normally a caller that invokes setup_revisions() has to
check rev.pending to see if anything was actually queued for
the traversal. But they can't tell the difference between
two cases:

  1. The user gave us no tip from which to start a
     traversal.

  2. The user tried to give us tips via --glob, --all, etc,
     but their patterns ended up being empty.

Let's set a flag in the rev_info struct that callers can use
to tell the difference.  We can set this from the
init_all_refs_cb() function.  That's a little funny because
it's not exactly about initializing the "cb" struct itself.
But that function is the common setup place for doing
pattern traversals that is used by --glob, --all, etc.
Signed-off-by: NJeff King <peff@peff.net>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

In 751a2ac6 (rev-list --exclude: tests, 2013-11-01), we
added a few tests for handling "empty" inputs with rev-list
(i.e., where the user gave us some pattern but it turned out
not to queue any objects for traversal), all of which were
marked as failing.

In preparation for working on this area of the code, let's
give each test a more descriptive name. Let's also include
one more case which we should cover: feeding a --glob
pattern that doesn't match anything.

We can also drop the explanatory comment; we'll be
converting these to expect_success in the next few patches,
so the discussion isn't necessary.
Signed-off-by: NJeff King <peff@peff.net>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>