Re: [PATCH v6 3/5] difftool: avoid returning -1 to cmd_main() from run_dir_diff()

From: Junio C Hamano <gitster@pobox.com>
Date: Oct 1, 2021, 00:12 UTC
Message-ID: <xmqqmtnt8uxl.fsf@gitster.g>
In-Reply-To: <CAJDDKr41qzbUudAHgokRARK-kEG4araFwJ0UhLmJ4Pca=xFZRA@mail.gmail.com>

David Aguilar <davvid@gmail.com> writes:

Show 9 quoted lines

>> Among them, the return value from run_command_v_opt() eventually
>> come from wait_or_whine(), I think, so it may be generic -1 or
>> it may be WEXITSTATUS() of the child process.
>>
>> But I am not sure if this particular caller cares.  It is not
>
> The property I was trying to maintain is that we would forward the result
> from the child process in most situations, so we should try and forward
> the result from run_command_v_opt() whenever possible.

Oh, if it were the case, then I agree that !!ret would not be sufficient.

I just had an impression that this particular caller of the run_command_v_opt() did not care, as it did not special case -1 returned from the function as different from what came from WEXITSTATUS().

> That suggests that Ævar's sug is better...
>
>     return (ret < 0) ? 1 : ret;

Yup, that would work, as long as exit status 1 from the external diff tool is not all that special (in which case, we cannot tell it from 1 that came from ret = error()).

Thanks.