[ILUG] gcc and printf style functions

[kevin lyda]

On Fri, Mar 04, 2005 at 07:27:59PM +0100, Brian Foster wrote:
>  tip:  unless you are doing something sneaky elsewhere,
>       it is a good idea to use __format__ and __printf__,
>       i.e.:  __attribute__((__format__(__printf__, 2, 6)))
>       because those names are less likely to be #define'd
>       by some idio^H^H^H^H^Hone else.

yes, good point.  and gcc supports that.

>   |[ ... ]
>   | i've read through the gcc docs and as far as i can see, i'd need to
>   | change gcc in order to support new % format codes.  am i missing
>   | something?
>  I've always assumed --- but without research! --- that
>  the % codes are hardwired into gcc.  in addition to
>  the answer to Kevin's question, I would be interested
>  in knowing:

they're defined in c-format.c in the gcc sources it appears.

>    - is it possible, without changing gcc, to remove
>      or redefine some of the % codes it does know?

i think it should be.  and i think the argument for it is that the
format escapes are actually handled by the target libc.  gcc seems to
know glibc and various versions of ansi/iso c.  but gcc targets platforms
with other libc's - the mingw gcc uses the ms libc.

>  it's never been too important, but if someone just
>  happens to know ....  thanks.

yes, but it's a great way to catch some silly errors.

kevin

-- 
kevin lyda ~ dems for torture: salazar(co/10) landrieu(la/08) pryor(ar/08)
kevin at ie.suberic.net ~ nelson(fl/06) nelson(ne/06) lieberman(ct/06) 2/2/04
  Those who refuse to raise their voices against something as clearly evil
  as torture are enablers, if not collaborators.  --Bob Herbert, 2/11/04