xf86UDelay was used - removal breaks drivers

[Luc Verhaegen]

On Fri, Nov 06, 2009 at 05:50:35PM -0800, Alan Coopersmith wrote:
> Luc Verhaegen wrote:
> > On Fri, Nov 06, 2009 at 03:31:27PM -0800, Alan Coopersmith wrote:
> >> This commit:
> >>
> >>    commit 9522ab9a98e98bc394703038e34b1c12aa39a179
> >>    Author: Matt Turner <mattst88 at gmail.com>
> >>    Date:   Mon Aug 31 16:12:29 2009 -0400
> >>
> >>     Remove unused Delay.c
> >>
> >> has broken a number of drivers that used the previously exported
> >> xf86UDelay function, including:
> >> 	xf86-video-ast
> >> 	xf86-video-chips
> >> 	xf86-video-glint
> >> 	xf86-video-neomagic
> >> 	xf86-video-rendition
> >>
> >> Are updates going to be issued to all these drivers to use a different
> >> function, or should the removal be reverted?
> > 
> > Apart from the fact that i find changes like this horribly shortsighted 
> > and stupid, i wonder why tinderbox did not catch this.
> 
> We caught it while working on the updates to our packages, since we force
> resolution of all symbols when building drivers, but I don't think the
> X.Org upstream builds do that.   (For it to work for us on Solaris, we
> have to generate a linker mapfile listing the exported symbols from the
> Xorg binary that's used to satisfy the linker when checking.)
> 
> It does appear some distros have had bugs reported on it though:
> 	http://bugs.gentoo.org/291825
> 	https://bugzilla.redhat.com/show_bug.cgi?id=523800

Sounds like something useful that tinderbox could do too then :)

Not that that in any way should allow for changes like this to happen 
gratuitously.

Luc Verhaegen.