Re: git checkout creates strange '(null)'-branch

[Erik Faye-Lund <kusmabite@gmail.com>]

On Tue, May 8, 2012 at 6:25 PM, Junio C Hamano <gitster@pobox.com> wrote:
> Erik Faye-Lund <kusmabite@gmail.com> writes:
>
>> But if I apply the following patch, the test passes. I'm not saying
>> it's the right thing to do, though.
>>
>> (Warning: white-space damaged because of copying diffs between terminals)
>>
>> diff --git a/builtin/checkout.c b/builtin/checkout.c
>> index 23fc56d..d70e819 100644
>> --- a/builtin/checkout.c
>> +++ b/builtin/checkout.c
>> @@ -1096,8 +1096,11 @@ int cmd_checkout(int argc, const char **argv,
>> const char *prefix)
>>                 int flag;
>>
>>                 if (!read_ref_full("HEAD", rev, 0, &flag) &&
>> -                   (flag & REF_ISSYMREF) && is_null_sha1(rev))
>> +                   (flag & REF_ISSYMREF) && is_null_sha1(rev)) {
>> +                       if (!opts.new_branch)
>> +                               return 0;
>>                         return switch_unborn_to_new_branch(&opts);
>> +               }
>>         }
>>         return switch_branches(&opts, &new);
>>  }
>
> This patch, if we ignore submodules for a while, actually makes sense to
> me.  If you have a commit, you are on that commit and you haven't done
> anything since you have checked out that commit, "git checkout" (no other
> parameters) would be a no-op.  If you "git init" a repository, and you
> haven't done anything since then, the above makes "git checkout" (no other
> parameters) a no-op.

I agree. It does seem to make sense.

> Am I missing some corner cases that we _should_ error out, perhaps for the
> sake of safety?

I don't know, and I was kind of hoping someone would jump in and
enlighten me of something, but that didn't happen (yet).

I have a slight preference for this operation, though I'm not
religious; it simply seem more consistent with other "git checkout"
operation.

Do you want me to prepare a proper patch including a variation of the
test-case from the other patch?