Daichi Kawahata wrote:
> Christian Biere <[EMAIL PROTECTED]> wrote:
> > but this means that we must not use format strings with ngettext().
 
> Although in my knowledge, it means human him/herself must confirm that
> whether its format is proper instead of compiler's auto checking,
> doesn't it? ...um, I don't know actually.

You can, of course, successfully compile gtk-gnutella no matter how
much warnings the compiler barfs at you. The warnings cause no fatal
errors. However, I don't want people to get used to warnings because
some of those may actual indicate (severe) bugs. Also if there are
many warnings, it's easy miss the important ones in the noise. And no,
I don't want to use specific compiler flags to suppress specific
warnings or filter them with grep from the compile log.
 
> > I assume this is no different from gettext() but GCC "trusts" the
> > latter to not mess with format strings.
 
> I don't always trust GCC, so I can understand what you want to say in
> different aspect though...well...I wish I had have MIPSpro compiler so
> that I can acquire certain view point.

Well, if GCC trusts gettext(), it doesn't make a lot of sense to not
trust ngettext() equally. Thus, I've added NG_() (as short-hand for
ngettext() using the GCC attribute format_arg for the prototype as
suggested by R�diger Kuhlmann. I don't know whether gettext checks
the strings only during generation of the message catalogs or also at
load-time.

-- 
Christian

Abo(r)t, ret(r)y, (r)efail?

Attachment: pgpSLpxXVOm2H.pgp
Description: PGP signature

Reply via email to