delta/git.git, branch ma/parse-maybe-bool

parse_decoration_style: drop unused argument `var`

2017-08-07T20:31:52+00:00

The previous commit left it unused.

Signed-off-by: Martin Ågren 
Signed-off-by: Junio C Hamano

treewide: deprecate git_config_maybe_bool, use git_parse_maybe_bool

2017-08-07T20:29:22+00:00

The only difference between these is that the former takes an argument
`name` which it ignores completely. Still, the callers are quite careful
to provide reasonable values for it.

Once in-flight topics have landed, we should be able to remove
git_config_maybe_bool. In the meantime, document it as deprecated in the
technical documentation. While at it, document git_parse_maybe_bool.

Signed-off-by: Martin Ågren 
Signed-off-by: Junio C Hamano

config: make git_{config,parse}_maybe_bool equivalent

2017-08-07T20:27:24+00:00

Both of these act on a string `value` which they parse as a boolean. The
"parse"-variant was introduced as a replacement for the "config"-variant
which for historical reasons takes an unused argument `name`. That it
was intended as a replacement is not obvious from commit 9a549d43
("config.c: rename git_config_maybe_bool_text and export it as
git_parse_maybe_bool", 2015-08-19), but that is what the background on
the mailing list suggests [1].

However, these two functions do not parse `value` in exactly the same
way. In particular, git_config_maybe_bool accepts integers (0 for false,
non-0 for true). This means there are two slightly different definitions
of "maybe_bool" in the code-base, and that every time a call to
git_config_maybe_bool is changed to use git_parse_maybe_bool, it risks
breaking someone's workflow.

Move the implementation of "config" into "parse" and make the latter a
trivial wrapper.

This also fixes the only user of git_parse_maybe_bool, `git push
--signed=..`.

[1] https://public-inbox.org/git/xmqq7fotd71o.fsf@gitster.dls.corp.google.com/

Signed-off-by: Martin Ågren 
Signed-off-by: Junio C Hamano

config: introduce git_parse_maybe_bool_text

2017-08-07T20:27:23+00:00

Commit 9a549d43 ("config.c: rename git_config_maybe_bool_text and export
it as git_parse_maybe_bool", 2015-08-19) intended git_parse_maybe_bool
to be a replacement for git_config_maybe_bool, which could then be
retired. That is not obvious from the commit message, but that is what
the background on the mailing list suggests [1].

However, git_{config,parse}_maybe_bool do not handle all input the same.
Before the rename, that was by design and there is a caller in config.c
which requires git_parse_maybe_bool to behave exactly as it does.

Prepare for the next patch by renaming git_parse_maybe_bool to ..._text
and reimplementing the first one as a simple call to the second one. Let
the existing users in config.c use ..._text, since it does what they
need.

[1] https://public-inbox.org/git/xmqq7fotd71o.fsf@gitster.dls.corp.google.com/

Signed-off-by: Martin Ågren 
Signed-off-by: Junio C Hamano

t5334: document that git push --signed=1 does not work

2017-08-07T20:27:21+00:00

When accepting booleans as command-line or config options throughout
Git, there are several documented synonyms for true and false.
However, one particular user is slightly broken: `git push --signed=..`
does not understand the integer synonyms for true and false.

This is hardly wanted. The --signed option has a different notion of
boolean than all other arguments and config options, including the
config option corresponding to it, push.gpgSign.

Add a test documenting the failure to handle --signed=1.

Signed-off-by: Martin Ågren 
Signed-off-by: Junio C Hamano

Doc/git-{push,send-pack}: correct --sign= to --signed=

2017-08-07T20:27:20+00:00

Since we're about to touch the behavior of --signed=, do this as a
preparatory step.

The documentation mentions --sign=, and it works. But that's just
because it's an unambiguous abbreviation of --signed, which is how it is
actually implemented. This was added in commit 30261094 ("push: support
signing pushes iff the server supports it", 2015-08-19). Back when that
series was developed [1] [2], there were suggestions about both --sign=
and --signed=. The final implementation settled on --signed=, but some
of the documentation and commit messages ended up using --sign=.

The option is referred to as --signed= in Documentation/config.txt
(under push.gpgSign).

One could argue that we have promised --sign for two years now, so we
should implement it as an alias for --signed. (Then we might also
deprecate the latter, something which was considered already then.) That
would be a slightly more intrusive change.

This minor issue would only be a problem once we want to implement some
other option --signfoo, but the earlier we do this step, the better.

[1] v1-thread:
https://public-inbox.org/git/1439492451-11233-1-git-send-email-dborowitz@google.com/T/#u

[2] v2-thread:
https://public-inbox.org/git/1439998007-28719-1-git-send-email-dborowitz@google.com/T/#m6533a6c4707a30b0d81e86169ff8559460cbf6eb

Signed-off-by: Martin Ågren 
Signed-off-by: Junio C Hamano

Git 2.12

2017-02-24T18:49:58+00:00

Signed-off-by: Junio C Hamano

Merge branch 'ps/doc-gc-aggressive-depth-update'

2017-02-24T18:48:10+00:00

Doc update.

* ps/doc-gc-aggressive-depth-update:
  docs/git-gc: fix default value for `--aggressiveDepth`

Merge branch 'bc/worktree-doc-fix-detached'

2017-02-24T18:48:10+00:00

Doc update.

* bc/worktree-doc-fix-detached:
  Documentation: correctly spell git worktree --detach

Merge branch 'dr/doc-check-ref-format-normalize'

2017-02-24T18:48:09+00:00

Doc update.

* dr/doc-check-ref-format-normalize:
  git-check-ref-format: clarify documentation for --normalize