Re: [PATCH 3/4] bisect: simplify the add of new bisect terms

[Matthieu Moy <Matthieu.Moy@grenoble-inp.fr>]

Junio C Hamano <gitster@pobox.com> writes:

> Matthieu Moy <Matthieu.Moy@grenoble-inp.fr> writes:
>
>> Louis-Alexandre Stuber <stuberl@ensimag.grenoble-inp.fr> writes:
>>
>>>> That is very different from ENOENT, which is an expected error when
>>>> you are not using a customized terms.
>>>
>>> But in the current state, we are going to create bisect_terms even if
>>> the bisection is in bad/good mode.
>>
>> Which means that in normal cases, you'll either succeed to open it, or
>> get ENOENT. We're talking about unexcepted cases (you don't have
>> permission to read it because it's not your file, because you messed up
>> with a chmod, or whatever reason).
>
> I think both I and you misunderstood what they wanted to do, which
> is to write out good and bad into terms file even though these are
> not customized, and then always read from terms file to learn what
> words are used for good and bad.

Yes, indeed.

> But I do not think it is a good idea to penalize the normal case by
> writing the terms file and reading them back from it when the user
> is bisecting with good/bad in the first place, so....

No strong opinion on that, but creating one file doesn't cost much, and
one advantage of writing it unconditionally is that it unifies bad/good
and old/new more in the code. Just the creation of BISECT_TERMS becomes
a special-case.

-- 
Matthieu Moy
http://www-verimag.imag.fr/~moy/