Re: [PATCH v4] commit: add a commit.allowEmpty config variable

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Nov 13 2018, Johannes Schindelin wrote:

[Comments on the v4 patch also inline, found it easier to reply just to
this one]

> On Tue, 13 Nov 2018, Tanushree Tumane wrote:
>
>> From: tanushree27 <tanushreetumane@xxxxxxxxx>
>>
>> when we cherrypick an existing commit it doesn't change anything and
>> therefore it fails prompting to reset (skip commit) or commit using
>> --allow-empty attribute and then continue.
>
> This is a nice paragraph, but it might make sense to connect it to the
> commit's oneline somehow. I, for one, was surprised to see the oneline
> talk about `git commit` and the commit message about `git cherry-pick`.
>
> I could imagine that an introductory paragraph, talking about why one
> might want to commit empty commits, might be the best lead into the
> subject, and the paragraph about `cherry-pick` could follow (and be
> introduced by saying something along the lines that this config setting
> has more reach than just `git commit`; it also affects `git cherry-pick`)?

Agreed. I'm happy to see the test for-loop gone as I noted in
https://public-inbox.org/git/87d0rm7zeo.fsf@xxxxxxxxxxxxxxxxxxx/ but as
noted in that v3 feedback the whole "why would anyone want this?"
explanation is still missing, and this still smells like a workaround
for a bug we should be fixing elsewhere in the sequencing code.

[The rest of this for Tanushree]

>>
>> Add commit.allowEmpty configuration variable as a convenience to skip
>> this process.
>>
>> Add tests to check the behavior introduced by this commit.
>>
>> This closes https://github.com/git-for-windows/git/issues/1854
>>
>> Signed-off-by: tanushree27 <tanushreetumane@xxxxxxxxx>
>> Signed-off-by: Tanushree Tumane <tanushreetumane@xxxxxxxxx>
>> ---
>>  Documentation/config.txt     |  5 +++++
>>  Documentation/git-commit.txt |  3 ++-
>>  builtin/commit.c             |  8 ++++++++
>>  t/t3500-cherry.sh            | 10 ++++++++++
>>  4 files changed, 25 insertions(+), 1 deletion(-)
>>
>> diff --git a/Documentation/config.txt b/Documentation/config.txt
>> index c0727b7866..f3828518a5 100644
>> --- a/Documentation/config.txt
>> +++ b/Documentation/config.txt
>> @@ -1467,6 +1467,11 @@ commit.verbose::
>>  	A boolean or int to specify the level of verbose with `git commit`.
>>  	See linkgit:git-commit[1].
>>
>> +commit.allowEmpty::
>> +	A boolean to specify whether empty commits are allowed with `git
>> +	commit`. See linkgit:git-commit[1].
>> +	Defaults to false.
>> +
>>  credential.helper::
>>  	Specify an external helper to be called when a username or
>>  	password credential is needed; the helper may consult external
>> diff --git a/Documentation/git-commit.txt b/Documentation/git-commit.txt
>> index f970a43422..5d3bbf017a 100644
>> --- a/Documentation/git-commit.txt
>> +++ b/Documentation/git-commit.txt
>> @@ -176,7 +176,8 @@ The `-m` option is mutually exclusive with `-c`, `-C`, and `-F`.
>>  	Usually recording a commit that has the exact same tree as its
>>  	sole parent commit is a mistake, and the command prevents you
>>  	from making such a commit.  This option bypasses the safety, and
>> -	is primarily for use by foreign SCM interface scripts.
>> +	is primarily for use by foreign SCM interface scripts. See
>> +	`commit.allowEmpty` in linkgit:git-config[1].
>>
>>  --allow-empty-message::
>>         Like --allow-empty this command is primarily for use by foreign
>> diff --git a/builtin/commit.c b/builtin/commit.c
>> index 67fa949204..4516309ac2 100644
>> --- a/builtin/commit.c
>> +++ b/builtin/commit.c
>> @@ -101,6 +101,7 @@ static int all, also, interactive, patch_interactive, only, amend, signoff;
>>  static int edit_flag = -1; /* unspecified */
>>  static int quiet, verbose, no_verify, allow_empty, dry_run, renew_authorship;
>>  static int config_commit_verbose = -1; /* unspecified */
>> +static int config_commit_allow_empty = -1; /* unspecified */
>>  static int no_post_rewrite, allow_empty_message;
>>  static char *untracked_files_arg, *force_date, *ignore_submodule_arg, *ignored_arg;
>>  static char *sign_commit;
>> @@ -1435,6 +1436,10 @@ static int git_commit_config(const char *k, const char *v, void *cb)
>>  		config_commit_verbose = git_config_bool_or_int(k, v, &is_bool);
>>  		return 0;
>>  	}
>> +	if (!strcmp(k, "commit.allowempty")) {
>> +		config_commit_allow_empty = git_config_bool(k, v);
>> +		return 0;
>> +	}
>>
>>  	status = git_gpg_config(k, v, NULL);
>>  	if (status)
>> @@ -1556,6 +1561,9 @@ int cmd_commit(int argc, const char **argv, const char *prefix)
>>  	if (verbose == -1)
>>  		verbose = (config_commit_verbose < 0) ? 0 : config_commit_verbose;
>>
>> +	if (config_commit_allow_empty >= 0)  /* if allowEmpty is allowed in config*/
>> +		allow_empty = config_commit_allow_empty;
>> +

I had two comments on this hunk in my v3 feedback. It's fine to go for
something different (although I still think you should change it), but
for others following along you should at least say "I had such-and-such
feedback suggesting X, but decided to go for Y anyway because...".

>>  	if (dry_run)
>>  		return dry_run_commit(argc, argv, prefix, current_head, &s);
>>  	index_file = prepare_index(argc, argv, prefix, current_head, 0);
>> diff --git a/t/t3500-cherry.sh b/t/t3500-cherry.sh
>> index f038f34b7c..11504e2d9f 100755
>> --- a/t/t3500-cherry.sh
>> +++ b/t/t3500-cherry.sh
>> @@ -55,4 +55,14 @@ test_expect_success \
>>       expr "$(echo $(git cherry master my-topic-branch) )" : "+ [^ ]* - .*"
>>  '
>>
>> +
>> +# Tests for commit.allowEmpty config
>> +

Let's drop this comment. It's redundant to the test description.

>> +test_expect_success 'cherry-pick existing commit with commit.allowEmpty' '
>> +    test_tick &&
>> +	test_commit "first" &&
>> +	test_commit "second" &&
>> +	git -c commit.allowEmpty=true cherry-pick HEAD~1
>> +'

So now you've dropped any tests of "git commit" (even though you're
changing commit.c, and just testing revert.c. So again, if that's all we
want isn't this whole thing just a simple bugfix of:

diff --git a/builtin/commit.c b/builtin/commit.c
index 96d336ec3d..1a12cc559e 100644
--- a/builtin/commit.c
+++ b/builtin/commit.c
@@ -1546,6 +1546,9 @@ int cmd_commit(int argc, const char **argv, const char *prefix)
        if (verbose == -1)
                verbose = (config_commit_verbose < 0) ? 0 : config_commit_verbose;

+       if (whence == FROM_CHERRY_PICK)
+               allow_empty = allow_empty_message = 1;
+
        if (dry_run)
                return dry_run_commit(argc, argv, prefix, current_head, &s);
        index_file = prepare_index(argc, argv, prefix, current_head, 0);

Possibly dropping the allow_empty_message part, but it seems reasonable
that whether you're re-picking an empty commit or one with an empty
message cherry-pick should always work.

I see that fails various existing tests, and I'm going to stop digging
now, but that brings me back to the "let's explain this better" part of
the feedback. I.e. if we can't just fix it let's explain why it can't be
made a default because it breaks such-and-such, so we need the config
option.



[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]

  Powered by Linux