Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!usc!snorkelwacker!bloom-beacon!LARRY.MCRCIM.MCGILL.EDU!mouse
From: mouse@LARRY.MCRCIM.MCGILL.EDU
Newsgroups: comp.windows.x
Subject: Re:  Compiling X11R4 with gcc on Sun-3 (4.0.3)
Message-ID: <9008110909.AA14843@Larry.McRCIM.McGill.EDU>
Date: 11 Aug 90 09:09:28 GMT
Sender: daemon@athena.mit.edu (Mr Background)
Organization: The Internet
Lines: 34

> [... X11 with GNU gcc ...]  I forced shared libraries off and was
> able to build the server & client entirely with gcc.  When I run the
> new server it is unable to locate any screens on my 3/60.  (operation
> not supported on socket, ioctl inappropriate for device)

> I'm sure I'm not the first person to try this :-) Is there any magic
> I have missed?

(If this isn't a FAQ, it should be!  This is about the third case of
this problem in a row I've seen here.)

Almost certainly, gcc's fixincludes script wasn't run, was run wrong,
is broken, or had its generated include files damaged afterwards.  In
short, your gcc include files are broken, and nothing using ioctls will
work right.  To test this theory, create a little test file (foo.c,
say) with these two lines in it

#include <sys/ioctl.h>
SIOCGIFCONF

and run it through cc -E and gcc -E.  Compare the last line of output
between the two; on a working system they will be identical (in a
functional sense - irrelevant differences such as whitespace can be
ignored).  Most likely you will find that the gcc output has an 'x'
where the cc output has an 'i' instead.

Until you can get your include files into a state where the above test
works, you haven't much hope of making gcc-built binaries of anything
using ioctls (such as the X server) work.

					der Mouse

			old: mcgill-vision!mouse
			new: mouse@larry.mcrcim.mcgill.edu