The current behavior of git checkout/switch is that --guess is currently
enabled by default. However, some users may not wish for this to happen
automatically. Instead of forcing users to specify --no-guess manually
each time, teach these commands the checkout.guess configuration
variable that gives users the option to set a default behavior.

Teach the completion script to recognize the new config variable and
disable DWIM logic if it is set to false.
Signed-off-by: NDenton Liu <liu.denton@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

The completion tests used that name unnecessarily, and it is a
non-inclusive term, so let's avoid using it here.

Since three of the touched test cases make use of the fact that two of
the branch names (`master` and `maint`) start with the same letter (or
even with the same two letters), we choose to replace the use of
`master` by a name that also has that property: `main`.
Signed-off-by: NJohannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

As our test suite partially reflects how we work in the Git project, it
is natural that the branch name `pu` was used in a couple places.

Since that branch was renamed to `seen`, let's use the new name
consistently.
Signed-off-by: NJohannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

The --orphan option is used to create a local branch which is detached
from the current history. In git switch, it always resets to the empty
tree, and thus the only completion we can provide is a branch name.
Follow the same rules for -c/-C (and -b/-B) when completing the argument
to --orphan.

In the case of git switch, after we complete the argument, there is
nothing more we can complete for git switch, so do not even try. Nothing
else would be valid.

In the case of git checkout, --orphan takes a start point which it uses
to determine the checked out tree, even though it created orphaned
history.

Update the previously added test cases as they are now passing.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

A previous commit added several test cases highlighting the subpar
completion logic for -c/-C and -b/-B when completing git switch and git
checkout.

In order to distinguish completing the argument vs the start-point for
this option, we now use the wordlist to determine the previous full word
on the command line.

If it's -c or -C (-b/-B for checkout), then we know that we are
completing the argument for the branch name.

Given that a user who already knows the branch name they want to
complete will simply not use completion, it makes sense to complete the
small subset of local branches when completing the argument for -c/-C.

In all other cases, if -c/-C are on the command line but are not the
most recent option, then we must be completing a start-point, and should
allow completing against all references.

Update the -c/-C and -b/-B tests to indicate they now pass.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Current completion for the --track option of git switch and git checkout
is sub par. In addition to the DWIM logic of a bare branch name, --track
has DWIM logic to convert specified remote/branch names into a local
branch tracking that remote. For example

  $git switch --track origin/master

This will create a local branch name master, that tracks the master
branch of the origin remote.

In fact, git switch --track on its own will not accept other forms of
references. These must instead be specified manually via the -c/-C/-b/-B
options.

Introduce __git_remote_heads() and the "remote-heads" mode for
__git_complete_refs. Use this when the --track option is provided while
completing in _git_switch and _git_checkout. Just as in the --detach
case, we never enable DWIM mode for --track, because it doesn't make
sense.

It should be noted that completion support is still a bit sub par when
it comes to handling -c/-C and --orphan. This will be resolved in
a future change.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Just like git switch, we should not complete DWIM remote branch names
if --detach has been specified. To avoid this, refactor _git_checkout in
a similar way to _git_switch.

Note that we don't simply clear dwim_opt when we find -d or --detach, as
we will be adding other modes and checks, making this flow easier to
follow.

Update the previously failing tests to show that the breakage has been
resolved.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Add a new --mode option to __git_complete_refs, which allows changing
the behavior to call __git_heads instead of __git_refs.

By passing --mode=heads, __git_complete_refs will only output local
branches. This enables using "--mode=heads --dwim" to enable listing
local branches and the remote unique branch names for DWIM.

Refactor completion support to use the new mode option, rather than
calling __git_heads directly. This has the advantage that we can now
correctly allow local branches along with suitable DWIM refs, rather
than only allowing DWIM when we complete all references.

Choose what mode it uses when calling __git_complete_refs. If -d or
--detach have been provided, then simply complete all refs, but
*without* the DWIM option as these DWIM names won't work properly in
--detach mode.

Otherwise, call __git_complete_refs with the default dwim_opt value and
use the new "heads" mode.

In this way, the basic support for completing just "git switch <TAB>"
will result in only local branches and remote unique names for DWIM.

The basic no-options tests for git switch, as well as several of the
-c/-C tests now pass, so remove the known breakage tags.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

A new helper, __git_find_last_on_cmdline is introduced, similar to the
already existing __git_find_on_cmdline, but which operates in reverse,
finding the *last* matching word of the provided wordlist.

Use this in a new __git_checkout_default_dwim_mode() function that will
determine when to enable listing of DWIM remote branches.

The __git_find_last_on_cmdline() function is used to determine which
--guess or --no-guess is in effect. If either one is provided, then we
unconditionally enable or disable the DWIM mode based on the last
provided option.

If neither --guess nor --no-guess is provided, then we check for
--no-track, and finally for GIT_COMPLETION_CHECKOUT_NO_GUESS=1.

This function is then used in _git_switch and _git_checkout to improve
the handling for when we enable listing of these DWIM remote branches.

This new logic is more robust, as we will correctly identify superseded
options, and ensure that both _git_switch and _git_checkout enable DWIM
in similar ways.

We can now update a few tests to indicate they pass. A few of the tests
previously added to highlight issues with the old DWIM logic still fail.
This is because of a separate issue related to the default completion
behavior of git switch, which will be addressed in a future change.

Additionally, due to this change, a few tests for the -b/-B handling of
git checkout now fail. This is a minor regression, and will be fixed by
a following change that improves the overall handling of -b/-B. Mark
these tests as known breakages for now.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Similar to -c/-C, --orphan takes an argument which is the branch name to
use. We ought to complete this branch name using similar rules as to how
we complete new branch names for -c/-C and -b/-B. Namely, limit the
total number of options provided by completing to the local branches.

Additionally, git switch --orphan does not take any start point and will
always create using the empty-tree. Thus, after the branch name is
completed, git switch --orphan should not complete any references.

Add test cases showing the expected behavior of --orphan, for both the
argument and starting point.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

When using the branch creation argument for git switch or git checkout
(-c/-C or -b/-B), the commands switch to a different mode: `git switch
-c <branch> <some-referance>` means to create a branch named <branch> at
the commit referred to by <some-reference>.

When completing git switch or git checkout, it makes sense to complete
the branch name differently from the start point.

When completing a branch, one might consider that we do not have
anything worth completing. After all, a new branch must have an entirely
new name. Consider, however, that if a user names branches using some
similar scheme, they might wish to name a new branch by modifying the
name of an existing branch.

To avoid overloading completion for the argument, it seems reasonable to
complete only the local branch names and the valid "Do What I Mean"
remote branch names.

Add tests for the completion of the argument to -c/-C and -b/-B,
highlighting this preferred completion behavior.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

When using the branch creation argument for git switch or git checkout,
-c/-C or -b/-B, the commands operate in a different mode: `git switch -c
<branch> <some-reference>` means to create a branch named <branch> at
the commit referred to by <some-reference>.

When completing the start-point, we ought to always complete all valid
references.

Add tests for the completion of the start-point to -c/-C and -b/-B.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

When the --track option is provided to git switch or git checkout, and
no branch is specified by -c or -b, git will interpret the tracking
branch to determine the local branch name to use. This "Do What I Mean"
logic is similar but distinct from the default DWIM logic of
interpreting a unique remote branch name as a request to create and
track that branch.

For example, `git switch --track origin/master` is interpreted as
a request to create a local branch named master that is tracking
origin/master.

The current completion for git checkout in this regard is only somewhat
poor:

 $git checkout --track <TAB>
 HEAD
 master
 matching-branch
 matching-tag
 other/branch-in-other
 other/master-in-other

At least it still includes remote references. The clutter from including
all references isn't too bad.

However, git switch completion is terrible:

 $git switch --track <TAB>
 master
 matching-branch

It only shows local branches, not even allowing any form of completion
of the remote references!

Add tests which highlight the expected behavior of completing --track on
its own.

Note that when -c/-C or -b/-B are provided we do expect completing more
references, but this will be discussed in a future change that addresses
these options specifically.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

When completing words for git switch, the completion function correctly
disables the DWIM remote branch names when in the '--detach' mode. These
DWIM remote branch names will not work when the --detach option is
specified, so it does not make sense to complete them.

git checkout, however, does not disable the completion of DWIM remote
branch names in this case.

Add test cases for both git switch and git checkout showing the expected
behavior.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

When provided with a single argument that is the name of a remote branch
that does not yet exist locally, both git switch and git checkout can
interpret this as a request to create a local branch that tracks that
remote branch. We call this behavior "Do What I Mean", or DWIM for
short.

To aid in using this DWIM, it makes sense for completion to list these
unique remote branch names when completing possible arguments for git
switch and git checkout. Indeed, both _git_checkout and _git_switch
implement support for completing such DWIM branch names.

In other words, in addition to the usual completions provided for git
switch, this "DWIM" logic means completion will include the names of
branches on remotes that are unique and thus there can be no ambiguity
of which remote to track when creating the local branch.

However, the DWIM logic is not always active. Many options, such as
--no-guess, --no-track, and --track disable this DWIM logic, as they
cause git switch and git checkout to behave in different modes.

Additionally, some completion users do not wish to have tab completion
include these remote names by default, and thus introduced
GIT_COMPLETION_CHECKOUT_NO_GUESS as an optional way to configure the
completion support to disable this feature of completion support.

For this reason, _git_checkout and _git_switch have many rules about
when to enable or disable completing of these remote refs. The two
commands follow similar but not identical rules.

Set aside the question of command modes that do not accept this DWIM
logic (--track, -c, --orphan, --detach) for now. Thinking just about the
main mode of git checkout and git switch, the following guidelines will
help explain the basic rules we ought to support when deciding whether
to list the remote branches for DWIM in completion.

1.  if --guess is enabled, we should list DWIM remote branch names, even
    if something else would disable it
2.  if --no-guess, --no-track or GIT_COMPLETION_CHECKOUT_NO_GUESS=1,
    then we should disable listing DWIM remote branch names.
3.  Since the '--guess' option is a boolean option, a later --guess
    should override --no-guess, and a later --no-guess should override
    --guess.

Putting all of these together, add some tests that highlight the
expected behavior of this DWIM logic.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

When provided with no options, git switch only allows switching between
branches. The one exception to this is the "Do What I Mean" logic that
allows a unique remote branch name to be interpreted as a request to
create a branch of the same name that is tracking that remote branch.

Unfortunately, the logic for the completion of git switch results in
completing not just branch names, but also pseudorefs like HEAD, tags,
and fully specified <remote>/<branch> references.

For example, we currently complete the following:

 $git switch <TAB>
 HEAD
 branch-in-other
 master
 master-in-other
 matching-branch
 matching-tag
 other/branch-in-other
 other/master-in-other

Indeed, if one were to attempt to use git switch with some of these
provided options, git will reject the request:

 $git switch HEAD
 fatal: a branch is expected, got 'HEAD

 $git switch matching-tag
 fatal: a branch is expected, got tag 'matching-tag'

 $git switch other/branch-in-other
 fatal: a branch is expected, got remote branch 'other/branch-in-other'

Ideally, git switch without options ought to complete only words which
will be accepted. Without options, this means to list local branch names
and the unique remote branch names without their remote name pre-pended.

 $git switch <TAB>
 branch-in-other
 master
 master-in-other
 matching-branch

Add a test case that highlights this subpar completion. Also add
a similar test for git checkout completion that shows that due to the
complex nature of git checkout, it must complete all references.
Signed-off-by: NJacob Keller <jacob.keller@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

As reported on the git mailing list, since git-2.25,
    git add untracked-dir/
has been tab completing to
    git add untracked-dir/./

The cause for this was that with commit b9670c1f (dir: fix checks on
common prefix directory, 2019-12-19),
    git ls-files -o --directory untracked-dir/
(or the equivalent `git -C untracked-dir ls-files -o --directory`) began
reporting
    untracked-dir/
instead of listing paths underneath that directory.  It may also be
worth noting that the real command in question was
    git -C untracked-dir ls-files -o --directory '*'
which is equivalent to
    git ls-files -o --directory 'untracked-dir/*'
which behaves the same for the purposes of this issue (the '*' can match
the empty string), but becomes relevant for the proposed fix.

At first, based on the report, I decided to try to view this as a
regression and tried to find a way to recover the old behavior without
breaking other stuff, or at least breaking as little as possible.
However, in the end, I couldn't figure out a way to do it that wouldn't
just cause lots more problems than it solved.  The old behavior was a
bug:
  * Although older git would avoid cleaning anything with `git clean -f
    .git`, it would wipe out everything under that direcotry with `git
    clean -f .git/`.  Despite the difference in command used, this is
    relevant because the exact same change that fixed clean changed the
    behavior of ls-files.
  * Older git would report different results based solely on presence or
    absence of a trailing slash for $SUBDIR in the command `git ls-files
    -o --directory $SUBDIR`.
  * Older git violated the documented behavior of not recursing into
    directories that matched the pathspec when --directory was
    specified.
  * And, after all, commit b9670c1f (dir: fix checks on common prefix
    directory, 2019-12-19) didn't overlook this issue; it explicitly
    stated that the behavior of the command was being changed to bring
    it inline with the docs.

(Also, if it helps, despite that commit being merged during the 2.25
series, this bug was not reported during the 2.25 cycle, nor even during
most of the 2.26 cycle -- it was reported a day before 2.26 was
released.  So the impact of the change is at least somewhat small.)

Instead of relying on a bug of ls-files in reporting the wrong content,
change the invocation of ls-files used by git-completion to make it grab
paths one depth deeper.  Do this by changing '$DIR/*' (match $DIR/ plus
0 or more characters) into '$DIR/?*' (match $DIR/ plus 1 or more
characters).  Note that the '?' character should not be added when
trying to complete a filename (e.g. 'git ls-files -o --directory
"merge.c?*"' would not correctly return "merge.c" when such a file
exists), so we have to make sure to add the '?' character only in cases
where the path specified so far is a directory.
Signed-off-by: NElijah Newren <newren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

When using the __git_find_on_cmdline() helper function so far we've
only been interested in which one of a set of words appear on the
command line.  To complete options for some of 'git worktree's
subcommands in the following patches we'll need not only that, but the
index of that word on the command line as well.

Extend __git_find_on_cmdline() to optionally show the index of the
found word on the command line (IOW in the $words array) when the
'--show-idx' option is given.
Signed-off-by: NSZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

The following two patches will refactor and extend the
__git_find_on_cmdline() helper function, so let's add a few tests
first to make sure that its basic behavior doesn't change.
Signed-off-by: NSZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Decisions taken for simplicity:
1) For now, `--pathspec-from-file` is declared incompatible with
   `--patch`, even when <file> is not `stdin`. Such use case it not
   really expected.
2) It is not allowed to pass pathspec in both args and file.

`you must specify path(s) to restore` block was moved down to be able to
test for `pathspec.nr` instead, because testing for `argc` is no longer
correct.

`git switch` does not support the new options because it doesn't expect
`<pathspec>` arguments.
Signed-off-by: NAlexandr Miloslavskiy <alexandr.miloslavskiy@syntevo.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Reported-by: NJens Schleusener <Jens.Schleusener@fossies.org>
Signed-off-by: NElijah Newren <newren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Reported-by: NJens Schleusener <Jens.Schleusener@fossies.org>
Signed-off-by: NElijah Newren <newren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Before, when format-patch generated a cover letter, only the body would
be populated with a branch's description while the subject would be
populated with placeholder text. However, users may want to have the
subject of their cover letter automatically populated in the same way.

Teach format-patch to accept the `--cover-from-description` option and
corresponding `format.coverFromDescription` config, allowing users to
populate different parts of the cover letter (including the subject
now).
Signed-off-by: NDenton Liu <liu.denton@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

t9902 had a list of three random porcelain commands as a sanity check,
one of which was filter-branch.  Since we are recommending people not
use filter-branch, let's update this test to use rebase instead of
filter-branch.
Signed-off-by: NElijah Newren <newren@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Completing configuration sections and variable names for the stuck
argument of 'git clone --config=<TAB>' requires a bit of extra care
compared to doing the same for the unstuck argument of 'git clone
--config <TAB>', because we have to deal with that '--config=' being
part of the current word to be completed.

Add an option to the __git_complete_config_variable_name_and_value()
and in turn to the __git_complete_config_variable_name() helper
functions to specify the current section/variable name to be
completed, so they can be used even when completing the stuck argument
of '--config='.

__git_complete_config_variable_value() already has such an option, and
thus no further changes were necessary to complete possible values
after 'git clone --config=section.name=<TAB>'.
Signed-off-by: NSZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

'git config' expects a configuration variable's name and value in
separate options, so we complete values as they stand on their own on
the command line.  'git -c', however, expects them in a single option
joined by a '=' character, so we should be able to complete values
when they are following 'section.name=' in the same word.

Add new options to the __git_complete_config_variable_value() function
to allow callers to specify the current word to be completed and the
configuration variable whose value is to be completed, and use these
to complete possible values after 'git -c 'section.name=<TAB>'.
Signed-off-by: NSZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

'git config' expects a configuration variable's name and value in
separate arguments, so we let the __gitcomp() helper append a space
character to each variable name by default, like we do for most other
things (--options, refs, paths, etc.).  'git -c', however, expects
them in a single option joined by a '=' character, i.e.
'section.name=value', so we should append a '=' character to each
fully completed variable name, but no space, so the user can continue
typing the value right away.

Add an option to the __git_complete_config_variable_name() function to
allow callers to specify an alternate suffix to add, and use it to
append that '=' character to configuration variables.  Update the
__gitcomp() helper function to not append a trailing space to any
completion words ending with a '=', not just to those option with a
stuck argument.
Signed-off-by: NSZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

The number of configuration variables listed by the completion script
grew quite when we started to auto-generate it from the documentation
[1], so we now complete them in two steps: first we list only the
section names, then the rest [2].  To get the section names we simply
strip everything following the first dot in each variable name,
resulting in a lot of repeated section names, because most sections
contain more than one configuration variable.  This is not a
correctness issue in practice, because Bash's completion facilities
remove all repetitions anyway, but these repetitions make testing a
bit harder.

Replace the small 'sed' script removing subsections and variable names
with an 'awk' script that does the same, and in addition removes any
repeated configuration sections as well (by first creating and filling
an associative array indexed by all encountered configuration
sections, and then iterating over this array and printing the indices,
i.e. the unique section names).  This change makes the failing 'git
config - section' test in 't9902-completion.sh' pass.

Note that this changes the order of section names in the output, and
makes it downright undeterministic, but this is not an issue, because
Bash sorts them before presenting them to the user, and our completion
tests sort them as well before comparing with the expected output.

Yeah, it would be simpler and shorter to just append '| sort -u' to
that command, but that would incur the overhead of one more external
process and pipeline stage every time a user completes configuration
sections.

[1] e17ca926 (completion: drop the hard coded list of config vars,
    2018-05-26)
[2] f22f6826 (completion: complete general config vars in two steps,
    2018-05-27)
Signed-off-by: NSZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

The next patches will change/refactor the way we complete
configuration variable names and values, so add a few tests to cover
the basics, namely the completion of matching configuration sections,
full variable names, and their values.

Note that the test checking the completion of configuration sections
is currently failing, though it's not a sign of an actual bug.  If a
section contains multiple variables, then that section is currently
repeated as many times as the number of variables in there.  This is
not a correctness issue in practice, because Bash's completion
facilities remove all repetitions anyway.  Consequently, we could list
all those repeated sections in the expected output of this test as
well, but then it would have to be updated whenever a new
configuration variable is added to those sections.  Instead, list each
matching configuration section only once, mark the test as failing for
now, and the next patch will update the completion script to avoid
those repetitions.
Signed-off-by: NSZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Change the GIT_TEST_GETTEXT_POISON variable from being "non-empty?" to
being a more standard boolean variable.

Since it needed to be checked in both C code and shellscript (via test
-n) it was one of the remaining shellscript-like variables. Now that
we have "env--helper" we can change that.

There's a couple of tricky edge cases that arise because we're using
git_env_bool() early, and the config-reading "env--helper".

If GIT_TEST_GETTEXT_POISON is set to an invalid value die_bad_number()
will die, but to do so it would usually call gettext(). Let's detect
the special case of GIT_TEST_GETTEXT_POISON and always emit that
message in the C locale, lest we infinitely loop.

As seen in the updated tests in t0017-env-helper.sh there's also a
caveat related to "env--helper" needing to read the config for trace2
purposes.

Since the C_LOCALE_OUTPUT prerequisite is lazy and relies on
"env--helper" we could get invalid results if we failed to read the
config (e.g. because we'd loop on includes) when combined with
e.g. "test_i18ngrep" wanting to check with "env--helper" if
GIT_TEST_GETTEXT_POISON was true or not.

I'm crossing my fingers and hoping that a test similar to the one I
removed in the earlier "config tests: simplify include cycle test"
change in this series won't happen again, and testing for this
explicitly in "env--helper"'s own tests.

This change breaks existing uses of
e.g. GIT_TEST_GETTEXT_POISON=YesPlease, which we've documented in
po/README and other places. As noted in [1] we might want to consider
also accepting "YesPlease" in "env--helper" as a special-case.

But as the lack of uproar over 6cdccfce ("i18n: make GETTEXT_POISON
a runtime option", 2018-11-08) demonstrates the audience for this
option is a really narrow set of git developers, who shouldn't have
much trouble modifying their test scripts, so I think it's better to
deal with that minor headache now and make all the relevant GIT_TEST_*
variables boolean in the same way than carry the "YesPlease"
special-case forward.

1. https://public-inbox.org/git/xmqqtvckm3h8.fsf@gitster-ct.c.googlers.com/Signed-off-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Commit 6532f374 ("completion: allow to customize the completable
command list", 2018-05-20) tried to allow multiple space-separated
entries in completion.commands. To do this, it copies each parsed token
into a strbuf so that the result is NUL-terminated.

However, for tokens starting with "-", it accidentally passes the
original non-terminated string, meaning that only the final one worked.
Switch to using the strbuf.
Reported-by: NTodd Zullinger <tmz@pobox.com>
Signed-off-by: NJeff King <peff@peff.net>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

6532f374 ("completion: allow to customize the completable command
list", 2018-05-20) added the completion.commands config variable.
Multiple commands may be added or removed, separated by a space.

Demonstrate the failure of multiple removals.
Signed-off-by: NTodd Zullinger <tmz@pobox.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Currently 'git checkout' is defined as an overlay operation, which
means that if in 'git checkout <tree-ish> -- [<pathspec>]' we have an
entry in the index that matches <pathspec>, but that doesn't exist in
<tree-ish>, that entry will not be removed from the index or the
working tree.

Introduce a new --{,no-}overlay option, which allows using 'git
checkout' in non-overlay mode, thus removing files from the working
tree if they do not exist in <tree-ish> but match <pathspec>.

Note that 'git checkout -p <tree-ish> -- [<pathspec>]' already works
this way, so no changes are needed for the patch mode.  We disallow
'git checkout --overlay -p' to avoid confusing users who would expect
to be able to force overlay mode in 'git checkout -p' this way.

Untracked files are not affected by this change, so 'git checkout
--no-overlay HEAD -- untracked' will not remove untracked from the
working tree.  This is so e.g. 'git checkout --no-overlay HEAD -- dir/'
doesn't delete all untracked files in dir/, but rather just resets the
state of files that are known to git.
Suggested-by: NJunio C Hamano <gitster@pobox.com>
Signed-off-by: NThomas Gummerer <t.gummerer@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Let's say there are files named 'foo bar.txt', and 'abc def/test.txt' in
repository. When following commands trigger a completion:

    git show HEAD:fo<Tab>
    git show HEAD:ab<Tab>

The completion results in bash/zsh:

    git show HEAD:foo bar.txt
    git show HEAD:abc def/

Where the both of them have an unescaped space in paths, so they'll be
misread by git. All entries of git ls-tree either a filename or a
directory, so __gitcomp_file() is proper rather than __gitcomp_nl().

Note the commit f12785a3, which handles quoted paths properly. Like this
case, we should dequote $cur_ for ?*:* case. For example, let's say
there is untracked directory 'abc deg', then trigger a completion:

    git show HEAD:abc\ de<Tab>
    git show HEAD:'abc de<Tab>
    git show HEAD:"abc de<Tab>

should uniquely complete 'abc def', but bash completes 'abc def' and
'abc deg' instead. In zsh, triggering a completion:

    git show HEAD:abc\ def/<Tab>

should complete 'test.txt', but nothing comes. The both problems will be
resolved by dequoting paths.

__git_complete_revlist_file() passes arguments to __gitcomp_nl() where
the first one is a list something like:

    abc def/Z
    foo bar.txt Z

where Z is the mark of the EOL.

- The trailing space of blob in __git ls-tree | sed.
  It makes the completion results become:

      git show HEAD:foo\ bar.txt\ <CURSOR>

  So git will try to find a file named 'foo bar.txt ' instead.

- The trailing slash of tree in __git ls-tree | sed.
  It makes the completion results on zsh become:

      git show HEAD:abc\ def/ <CURSOR>

  So that the last space on command like should be removed on zsh to
  complete filenames under 'abc def/'.
Signed-off-by: NChayoung You <yousbe@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

The oneline notwithstanding, 13374987 (completion: use _gitcompbuiltin
for format-patch, 2018-11-03) changed also the way send-email options
are completed, by asking the git send-email command itself what options
it offers.

Necessarily, this must fail when built with NO_PERL because send-email
itself is a Perl script. Which means that we need the PERL prerequisite
for the send-email test case in t9902.
Signed-off-by: NJohannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Some of the functions in our test library check that they were invoked
properly with conditions like this:

  test "$#" = 2 ||
  error "bug in the test script: not 2 parameters to test-expect-success"

If this particular condition is triggered, then 'error' will abort the
whole test script with a bold red error message [1] right away.

However, under certain circumstances the test script will be aborted
completely silently, namely if:

  - a similar condition in a test helper function like
    'test_line_count' is triggered,
  - which is invoked from the test script's "main" shell [2],
  - and the test script is run manually (i.e. './t1234-foo.sh' as
    opposed to 'make t1234-foo.sh' or 'make test') [3]
  - and without the '--verbose' option,

because the error message is printed from within 'test_eval_', where
standard output is redirected either to /dev/null or to a log file.
The only indication that something is wrong is that not all tests in
the script are executed and at the end of the test script's output
there is no "# passed all N tests" message, which are subtle and can
easily go unnoticed, as I had to experience myself.

Send these "bug in the test script" error messages directly to the
test scripts standard error and thus to the terminal, so those bugs
will be much harder to overlook.  Instead of updating all ~20 such
'error' calls with a redirection, let's add a BUG() function to
'test-lib.sh', wrapping an 'error' call with the proper redirection
and also including the common prefix of those error messages, and
convert all those call sites [4] to use this new BUG() function
instead.

[1] That particular error message from 'test_expect_success' is
    printed in color only when running with or without '--verbose';
    with '--tee' or '--verbose-log' the error is printed without
    color, but it is printed to the terminal nonetheless.

[2] If such a condition is triggered in a subshell of a test, then
    'error' won't be able to abort the whole test script, but only the
    subshell, which in turn causes the test to fail in the usual way,
    indicating loudly and clearly that something is wrong.

[3] Well, 'error' aborts the test script the same way when run
    manually or by 'make' or 'prove', but both 'make' and 'prove' pay
    attention to the test script's exit status, and even a silently
    aborted test script would then trigger those tools' usual
    noticable error messages.

[4] Strictly speaking, not all those 'error' calls need that
    redirection to send their output to the terminal, see e.g.
    'test_expect_success' in the opening example, but I think it's
    better to be consistent.
Signed-off-by: NSZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

When checkout dwim is added in [1], it is restricted to only dwim when
certain conditions are met and fall back to default checkout behavior
otherwise. It turns out falling back could be confusing. One of the
conditions to turn

    git checkout frotz

to

    git checkout -b frotz origin/frotz

is that frotz must not exist as a file. But when the user comes to
expect "git checkout frotz" to create the branch "frotz" and there
happens to be a file named "frotz", git's silently reverting "frotz"
file content is not helping. This is reported in Git mailing list [2]
and even used as an example of "Git is bad" elsewhere [3].

We normally try to do the right thing, but when there are multiple
"right things" to do, it's best to leave it to the user to decide.
Check this case, ask the user to to disambiguate:

- "git checkout -- foo" will check out path "foo"
- "git checkout foo --" will dwim and create branch "foo" [4]

For users who do not want dwim, use --no-guess. It's useless in this
particular case because "git checkout --no-guess foo --" will just
fail. But it could be used by scripts.

[1] 70c9ac2f (DWIM "git checkout frotz" to "git checkout -b frotz
    origin/frotz" - 2009-10-18)
[2] https://public-inbox.org/git/CACsJy8B2TVr1g+k+eSQ=pBEO3WN4_LtgLo9gpur8X7Z9GOFL_A@mail.gmail.com/
[3] https://news.ycombinator.com/item?id=18230655
[4] a047fafc (checkout: allow dwim for branch creation for "git
    checkout $branch --" - 2013-10-18)
Signed-off-by: NNguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Change the GETTEXT_POISON compile-time + runtime GIT_GETTEXT_POISON
test parameter to only be a GIT_TEST_GETTEXT_POISON=<non-empty?>
runtime parameter, to be consistent with other parameters documented
in "Running tests with special setups" in t/README.

When I added GETTEXT_POISON in bb946bba ("i18n: add GETTEXT_POISON
to simulate unfriendly translator", 2011-02-22) I was concerned with
ensuring that the _() function would get constant folded if NO_GETTEXT
was defined, and likewise that GETTEXT_POISON would be compiled out
unless it was defined.

But as the benchmark in my [1] shows doing a one-off runtime
getenv("GIT_TEST_[...]") is trivial, and since GETTEXT_POISON was
originally added the GIT_TEST_* env variables have become the common
idiom for turning on special test setups.

So change GETTEXT_POISON to work the same way. Now the
GETTEXT_POISON=YesPlease compile-time option is gone, and running the
tests with GIT_TEST_GETTEXT_POISON=[YesPlease|] can be toggled on/off
without recompiling.

This allows for conditionally amending tests to test with/without
poison, similar to what 859fdc0c ("commit-graph: define
GIT_TEST_COMMIT_GRAPH", 2018-08-29) did for GIT_TEST_COMMIT_GRAPH. Do
some of that, now we e.g. always run the t0205-gettext-poison.sh test.

I did enough there to remove the GETTEXT_POISON prerequisite, but its
inverse C_LOCALE_OUTPUT is still around, and surely some tests using
it can be converted to e.g. always set GIT_TEST_GETTEXT_POISON=.

Notes on the implementation:

 * We still compile a dedicated GETTEXT_POISON build in Travis
   CI. Perhaps this should be revisited and integrated into the
   "linux-gcc" build, see ae59a4e4 ("travis: run tests with
   GIT_TEST_SPLIT_INDEX", 2018-01-07) for prior art in that area. Then
   again maybe not, see [2].

 * We now skip a test in t0000-basic.sh under
   GIT_TEST_GETTEXT_POISON=YesPlease that wasn't skipped before. This
   test relies on C locale output, but due to an edge case in how the
   previous implementation of GETTEXT_POISON worked (reading it from
   GIT-BUILD-OPTIONS) wasn't enabling poison correctly. Now it does,
   and needs to be skipped.

 * The getenv() function is not reentrant, so out of paranoia about
   code of the form:

       printf(_("%s"), getenv("some-env"));

   call use_gettext_poison() in our early setup in git_setup_gettext()
   so we populate the "poison_requested" variable in a codepath that's
   won't suffer from that race condition.

 * We error out in the Makefile if you're still saying
   GETTEXT_POISON=YesPlease to prompt users to change their
   invocation.

 * We should not print out poisoned messages during the test
   initialization itself to keep it more readable, so the test library
   hides the variable if set in $GIT_TEST_GETTEXT_POISON_ORIG during
   setup. See [3].

See also [4] for more on the motivation behind this patch, and the
history of the GETTEXT_POISON facility.

1. https://public-inbox.org/git/871s8gd32p.fsf@evledraar.gmail.com/
2. https://public-inbox.org/git/20181102163725.GY30222@szeder.dev/
3. https://public-inbox.org/git/20181022202241.18629-2-szeder.dev@gmail.com/
4. https://public-inbox.org/git/878t2pd6yu.fsf@evledraar.gmail.com/Signed-off-by: NÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

A fair number of tests need to check that the filesystem supports file
names including "funny" characters, like newline, tab, and double-quote.
Jonathan Nieder suggested that this be extracted into a lazy prereq in
the top-level `test-lib.sh`. This patch effects that change.

The FUNNYNAMES prereq now uniformly requires support for newlines, tabs,
and double-quotes in filenames. This very slightly decreases the power
of some tests, which might have run previously on a system that supports
(e.g.) newlines and tabs but not double-quotes, but now will not. This
seems to me like an acceptable tradeoff for consistency.

One test (`t/t9902-completion.sh`) defined FUNNYNAMES to further require
the separators \034 through \037, the test for which was implemented
using the Bash-specific $'\034' syntax. I've elected to leave this one
as is, renaming it to FUNNIERNAMES.

After this patch, `git grep 'test_\(set\|lazy\)_prereq.*FUNNYNAMES'` has
only one result.
Signed-off-by: NWilliam Chargin <wchargin@gmail.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Signed-off-by: NEric Sunshine <sunshine@sunshineco.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>