[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-devel
Subject:    Re: [PATCH] second try at fixing up socklen issues on Solaris
From:       Carsten Griwodz <griff () ifi ! uio ! no>
Date:       2001-04-05 11:25:41
[Download RAW message or body]

Hi,

I can't imagine why any OS should use more than sizeof(int) bytes for
socklen_t, but who knows? My current problem is that AIX uses unsigned
long instead of unsigned int in 32 bit mode, probably for backward
compatibility. But is it guaranteed by any standard that socklen_t is a
basic data type everywhere?

The idea of passing a pointer to int (or something else) seems reasonable
to me.

Regards,
   Carsten


On Thu, 5 Apr 2001, Oswald Buddenhagen wrote:

> > > Actually, this sounds perfect. I hadn't thought of forcing everyone
> > > over to int externally instead of socklen_t. I won't get a chance to
> > > look at this until at least tomorrow, so if somone else builds a
> > > patch, that's fine. It's not complicated, just insidious....
> > 
> > Which is exactly the reason I created the ksocklen_t type in the first 
> > place. Actually, I didn't want to include sys/socket.h in the code just 
> > to have socklen_t defined, so I created a type that would be exactly that.
> > 
> why didn't you just use int? ;)
> never mind.
> 
> > Maybe the problem was making it a typedef instead of #define.
> > 
> i don't think so. this should not matter in the source (except some
> possible warnings about type mismatches).
> 
> > So, my recommendation is: - use a general type for all KDE exported
> > functions, from KExtendedSocket, KSocket, K*SocketAddress and
> > KSocks. Should be "int", "unsigned" or a #define'd ksocklen_t
> > 
> i'm voting for unsigned. it's compatible with sizeof() (no gcc warning
> ...) and it's a primitive type.
> 
> > - make sure all our calls to the system functions cast that to 
> > "ksize_t", which is #define'd to be the third parameter of getsockname()
> > 
> yes, but note, that casting is possible only for values. don't cast 
> pointers - they need the temporary step i presented in the previous mail.
> 
> > And "ksize_t" is a misnomer. It's the one that should be "ksocklen_t", 
> >
> yeah ... :(
> 
> > but we can't change that anymore.
> >
> why? the goal is not to export it. and running a perl -pi on the whole
> kde sources is no problem. am i missing something?
> 
> > Or it should be "size_t", not the third parameter of getsockname().
> > 
> ???
> 
> > I don't think typedef is a good alternative anymore because some 
> > compilers use the new type's name in the mangling of the C++ function, 
> > while others (like gcc) will use the primitive type.
> > 
> sounds sensible.
> 
> best regards
> 
> -- 
> Hi! I'm a .signature virus! Copy me into your ~/.signature, please!
> --
> Nothing is fool-proof to a sufficiently talented fool.
>  
> >> Visit http://master.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<
> 

 
>> Visit http://master.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<

[prev in list] [next in list] [prev in thread] [next in thread]