Re: [PATCHv4 3/3] Advices about 'git rm' during conflicts (unmerged paths) more relevant

[konglu@minatec.inpg.fr]

Junio C Hamano <gitster@pobox.com> a écrit :

>> +		if (d->stagemask == 1 && !both_deleted_flag)
>> +			both_deleted_flag = 1;
>> +		else if ((d->stagemask == 3 || d->stagemask == 5) &&  
>> !simple_deleted_flag)
>> +			simple_deleted_flag = 1;
>> +		else if ((d->stagemask == 2 || d->stagemask == 4 || d->stagemask == 6 ||
>> +				d->stagemask == 7) && !not_deleted_flag)
>> +			not_deleted_flag = 1;
>> +	}
>
> 	switch (d->stagemask) {
>         case 1:
>         	both_deleted = 1;
>                 break;
>         case 3:
>         case 5:
>         	simple_deleted = 1;
>                 break;
>         default:
>         	not_deleted = 1;
>                 break;
> 	}

Yes, using switch makes the code easier to read. In fact, I put explicitly the
condition when there are no delete (cases 2, 4, 6 and 7) because there  
are cases
when d->stagemask can take other number than 1..7. For example:

git init git &&
cd git &&
test_commit foo main.txt foo &&
git branch second_branch &&
git mv main.txt sub_master.txt &&
git commit -m "main.txt renamed in sub_master.txt" &&
git checkout second_branch &&
git mv main.txt sub_second.txt &&
git commit -m "main.txt renamed in sub_second.txt" &&
git merge master &&
git add sub_master.txt &&
git add sub_second.txt

At this point, the output of git status shows in unmerged paths the  
file main.txt
that is marked as 'both deleted' so the number of elements in  
s->change.items should
be 1. However, in this case, there are 2 elements. One is about the  
file main.txt and
its stagemask is '1' (as expected) but the stagemask of the other  
element (I don't
know its origin) is '0'.
Anyway, the new code becomes something like:

switch (d->stagemask) {
case 0:
	break;
case 1:
	both_deleted = 1;
	break;
case 3:
case 5:
	del_mod_conflict = 1;
	break;
default:
	not_deleted = 1;
	break;
}

Though I don't know if d->stagemask can take values such as 8, 9, etc.