Re: [Dri-devel] Stange multi-context problem in Radeon driver (and probably others)

[Dieter <Die...@ha...>]

Am Donnerstag, 3. Juli 2003 03:18 schrieb Ian Romanick:
> Dieter N=FCtzel wrote:
> > Am Mittwoch, 2. Juli 2003 22:26 schrieb Ian Romanick:
> >>Dieter N=FCtzel wrote:
> >>>Some xdemos apps show it, too.
> >>>"wincopy" isn't working.
> >>>
> >>>Mesa/xdemos> ./wincopy
> >>>glXMakeContextCurrent failed in Redraw()
> >>>glXMakeContextCurrent failed in Redraw()
> >>>glXMakeContextCurrent failed in Redraw()
> >>>[-]
> >>
> >>wincopy doesn't work because it's trying to use an unsupported GLX 1.3
> >>call.  glXMakeContextCurrent & glXMakeCurrentReadSGI are the functions
> >>that I'm trying to add support for.  On my system with my patches,
> >>wincopy works fairly well. :)
> >
> > I want Mesa 5.1 (5.2), now ;-)
>
> You won't have to wait that long.  SGI_make_current_read will probably
> be supported for indirect contexts tomorrow.  I think I'm going to go
> ahead and commit all of the device independent parts tomorrow, and
> commit the device dependent parts as they are fixed.
>
> >>>"manywin" update only the lasted created window in hardware mode
> >>> (n<=3D28). All additional windows (indirect mode) would be updated
> >>> simultaneously. Sometimes it locks X after the second try.
>
> Could you try the attached patch?  It should fix the problem for both
> r100 and r200 based cards.  I had to hand-edit the patch to remove some
> other changes from glxcmds.c, so let me know if you have any problems
> getting it to apply cleanly.

That worked.

> Hopefully this will be the *last time*
> that I have to modify GetDRIDrawable! :)

But one error:

[-]
_MESA -DUSE_X86_ASM    glxcmds.c
glxcmds.c: In function `glXSwapBuffers':
glxcmds.c:725: structure has no member named `currentReadable'
make[5]: *** [glxcmds.o] Error 1

[-]
=2D    if ((dpy =3D=3D gc->currentDpy) && (drawable =3D=3D gc->currentDrawa=
ble)) {
+    gc =3D __glXGetCurrentContext();
+    if ((gc !=3D NULL) && (dpy =3D=3D gc->currentDpy) &&
+       ((drawable =3D=3D gc->currentDrawable) || (drawable =3D=3D=20
gc->currentReadable)) ) {
[-]

> >>*This* might very well be related.  I notice that if manywin is run with
> >>'-s' it works fine.  Is there a bug filed, either in the DRI database or
> >>the XFree86 database, for this?
> >
> > No? --- Not from me. It worked with tdfx...;-)
>
> Okay.  I created one in the XFree86 database.  It's bug #453.

Thanks!
	Dieter