Raphael Manfredi wrote:
> Quoting Christian Biere <[EMAIL PROTECTED]> from ml.softs.gtk-gnutella.devel:
> :IMHO, VAL_EAGAIN is unnecessary.
> 
> It's a standard metaconfig symbol.
> 
> :#if defined(EWOULDBLOCK) && EWOULDBLOCK != EAGAIN
> :...
> :#endif
> :
> :should take care of that, in other places a macro ERROR_RETRY() might be
> :more handy.
> 
> Perhaps, but as I said, Configure already computes VAL_EAGAIN for us,
> and it simplifies things a bit.

Well this requires that all functions return either EAGAIN or
EWOULDBLOCK. If those are actually different and some (library)
function use the other variant, you lose. I don't really know
why you would need VAL_EAGAIN anyways. If you #undef EAGAIN
and then #define EAGAIN appropriately from override.h, it has
exactly the same effect and you wouldn't have this odd VAL_EAGAIN
thing in the sources.

Personally I despise compile checks which needlessly require
running code making cross-compiling impossible and are far
from being "correct".

-- 
Christian

Attachment: pgpO2QCQnoGVo.pgp
Description: PGP signature

Reply via email to