From: Alexander Potashev <aspotashev@gmail.com>
Date: Sun, 07 Jun 2009 11:12:39 GMT
Subject: Re: [PATCH v2 1/3] Introduce die_errno() that appends strerror(errno) to die()
Message-ID: <2325a7950906070412k3a440558h35db7511cc390e4b@mail.gmail.com>
In-Reply-To: <20090606221320.GB30064@coredump.intra.peff.net>

2009/6/7 Jeff King <peff@peff.net>:
> On Sat, Jun 06, 2009 at 04:44:51PM +0200, Thomas Rast wrote:
>
>> Code by Jeff King and Alexander Potashev, name by Johannes Sixt.
>> [...]
>> +void die_errno(const char *err, ...)
>> +{
>> +     va_list params;
>> +     char msg[1024];
>> +
>> +     va_start(params, err);
>> +
>> +     vsnprintf(msg, sizeof(msg), err, params);
>> +     die("%s: %s", msg, strerror(errno));
>> +
>> +     va_end(params);
>> +}
>> +

This breaks theoretical usage of custom format string in custom die
routines:
        1. A custom die routine might not support %s.
        2. If the die routine adds a custom format specifier, vsnprintf
                will fail.

But I'm not sure that passing a format string and a va_list to
die_routine/error_routine/warn_routine is really necessary. If there
is no much external code depending on them, I would convert them to the
style of usage_routine:
        void (*usage_routine)(const char *err)

CC'ing Petr Baudis <pasky@suse.cz>, see commit 39a3f5ea


>
> No, this approach is much more elegant than what I posted, so no need to
> credit me, at least. ;)

It was your idea of using 1024-character buffer here ;)


                                        Alexander