From: Jonathan Nieder <jrnieder@gmail.com>
Date: Tue, 22 May 2018 18:38:07 GMT
Subject: Re: Why do we have both x*() and *_or_die() for "do or die"?
Message-ID: <20180522183807.GK10623@aiede.svl.corp.google.com>
In-Reply-To: <CACsJy8BnsbOvz+iSz=9ug99Cs5Dm17Fgk65K2nXusBg0dML4xg@mail.gmail.com>

Duy Nguyen wrote:
> On Tue, May 22, 2018 at 7:49 PM, Ævar Arnfjörð Bjarmason
> <avarab@gmail.com> wrote:

>> Just a side-question unrelated to this patch per-se, why do we have both
>> x*() and *_or_die() functions in the codebase?
>
> I wondered about that myself shortly after suggesting
> repo_read_index_or_die(). My only guess is xfoo is better version of
> foo, which sometimes involves dying inside but that's not the only
> possible improvement. Later I guess people go with _or_die() more
> because it describes what the function does much better.

In particular, there are functions like xwrite that don't die on error
and write_or_die that do.

I'd probably be in favor of a series using cocinelle to rename the
functions that do die on error to _or_die.  The main case where I
pause for a moment is xmalloc, since I'm worried about the verboseness
of malloc_or_die, but I suspect I would get used to it.

Thanks,
Jonathan