Compilation error with newer CVS on glibc -2.0

[Martin Cracauer]

Hi, guys,

since Jul 20 or so I get this compilation error on
glibc-2.1.3/linux-2.4.18:

gcc -c -I. -I. -I../../include -I../../include  -g -O2 -Wall -fno-strength-reduce -mpreferred-stack-boundary=2 -D__WINE__ -D_REENTRANT -I/usr/X11R6/include -o wrc.o wrc.c
In file included from wrc.h:25,
                 from wrc.c:61:
wrctypes.h:447: redefinition of `struct user'
make[2]: *** [wrc.o] Error 1
make[2]: Leaving directory `/disks/part9/home/cracauer/work/cvs-packages-work/stupids/wine/wine/tools/wrc'

Wine's definition of struct user is clashing with the one from
/usr/include/sys/user.h

Now, it appears that /usr/include/sys/user.h (a glibc-2.1.3 file) is
in error having this struct in first place, since newer kernels have
this struct in asm/user.h.

I fixed the problem in my /usr/include, but I guess it is better to
use a different name for the struct in Wine to be compatible with this
configuration (it's Debian-2.1 with an upgraded kernel).

Unless somebody disagrees with that analysis, I'll submit a patch to
rename the struct.  Suggestions about a suitable name welcome.

Martin

[Alexandre Julliard]

Martin Cracauer  writes:
> Unless somebody disagrees with that analysis, I'll submit a patch to
> rename the struct.  Suggestions about a suitable name welcome.

Before doing that it would be interesting to find out why sys/user.h
gets included. This is not supposed to happen.

[Martin Cracauer]

Alexandre Julliard wrote on Thu, Aug 01, 2002 at 11:45:27AM -0700:
> Martin Cracauer  writes:
> > Unless somebody disagrees with that analysis, I'll submit a patch to
> > rename the struct.  Suggestions about a suitable name welcome.
>
> Before doing that it would be interesting to find out why sys/user.h
> gets included. This is not supposed to happen.

This is the sequence:

make[2]: Entering directory `/disks/part9/home/cracauer/work/cvs-packages-work/stupids/wine/wine/tools/wrc'
gcc -c -I. -I. -I../../include -I../../include  -g -O2 -Wall -fno-strength-reduce -mpreferred-stack-boundary=2 -D__WINE__ -D_REENTRANT -I/usr/X11R6/include -o wrc.o wrc.c
In file included from /usr/include/sys/ucontext.h:31,
                 from /usr/include/ucontext.h:27,
                 from /usr/include/signal.h:317,
                 from wrc.c:59:
/usr/include/sys/user.h:26: #error traceme
make[2]: *** [wrc.o] Error 1
make[2]: Leaving directory `/disks/part9/home/cracauer/work/cvs-packages-work/stupids/wine/wine/tools/wrc'

Compilation error with newer CVS on glibc -2.1.3

[Martin Cracauer]

Martin Cracauer wrote on Thu, Aug 01, 2002 at 03:08:18PM -0400:
> Alexandre Julliard wrote on Thu, Aug 01, 2002 at 11:45:27AM -0700:
> > Martin Cracauer  writes:
> > > Unless somebody disagrees with that analysis, I'll submit a patch to
> > > rename the struct.  Suggestions about a suitable name welcome.
> >
> > Before doing that it would be interesting to find out why sys/user.h
> > gets included. This is not supposed to happen.
>
> This is the sequence:
>
> make[2]: Entering directory `/disks/part9/home/cracauer/work/cvs-packages-work/stupids/wine/wine/tools/wrc'
> gcc -c -I. -I. -I../../include -I../../include  -g -O2 -Wall -fno-strength-reduce -mpreferred-stack-boundary=2 -D__WINE__ -D_REENTRANT -I/usr/X11R6/include -o wrc.o wrc.c
> In file included from /usr/include/sys/ucontext.h:31,
>                  from /usr/include/ucontext.h:27,
>                  from /usr/include/signal.h:317,
>                  from wrc.c:59:
> /usr/include/sys/user.h:26: #error traceme
> make[2]: *** [wrc.o] Error 1
> make[2]: Leaving directory `/disks/part9/home/cracauer/work/cvs-packages-work/stupids/wine/wine/tools/wrc'

signal.h only includes ucontext.h when __USE_GNU is defined.

That is why a simple test program with its own 'struct user' and
#include <signal.h> does not fail.

Now, where does __USE_GNU come from? By definign GNU_SOURCE:
#define _GNU_SOURCE  /* for pread/pwrite */
./include/wine/port.h
#define _GNU_SOURCE /* we may need to override some GNU extensions */
./scheduler/pthread.c

The offending _GNU_SOURCE is in here for some time.
1.18         (julliard 07-Jan-02): #define _GNU_SOURCE  /* for   pread/pwrite */

I think the julliard 22-Jul-02 commits to this file triggered the
error.

Have to head for home now, will sniff more later today.

Martin

Compilation error with newer CVS on glibc -2.1.3

[Alexandre Julliard]

Martin Cracauer  writes:
> The offending _GNU_SOURCE is in here for some time.
> 1.18         (julliard 07-Jan-02): #define _GNU_SOURCE  /* for   pread/pwrite */
>
> I think the julliard 22-Jul-02 commits to this file triggered the
> error.

Yes, I see it now. We should probably avoid _GNU_SOURCE here, but this
would be a bit of work. And since we don't really need that struct
user at all I have simply removed it. Thanks for looking into this.