[cairo] cairo-xlib-surface.c:402:_swap_ximage_to_native:Assertion `NOT_REACHED' failed.

[Daniel Amelang]

On 11/20/06, Erik Ohrnberger <Erik at echohome.org> wrote:
> > -----Original Message-----
> > From: cairo-bounces at cairographics.org
> > [mailto:cairo-bounces at cairographics.org] On Behalf Of Daniel Amelang
> > Sent: Monday, November 20, 2006 10:59 PM
> > To: Erik AT echohome.org
> > Cc: cairo at cairographics.org
> > Subject: Re: [cairo]
> > cairo-xlib-surface.c:402:_swap_ximage_to_native:Assertion
> > `NOT_REACHED' failed.
> >
> > On 11/19/06, Erik Ohrnberger <Erik AT echohome.org> wrote:
> > > > -----Original Message-----
> > > > From: cairo-bounces at cairographics.org
> > > > [mailto:cairo-bounces at cairographics.org] On Behalf Of
> > Daniel Amelang
> > > > Sent: Sunday, November 19, 2006 11:10 PM
> > > > To: Vladimir Vukicevic
> > > > Cc: cairo at cairographics.org; Erik AT echohome.org
> > > > Subject: Re: [cairo] cairo-xlib-surface.c:402:
> > > > _swap_ximage_to_native:Assertion `NOT_REACHED' failed.
> > > >
> > > > On 11/19/06, Vladimir Vukicevic <vladimir at pobox.com> wrote:
> > > > > ----- Daniel Amelang <daniel.amelang at gmail.com> wrote:
> > > > > > On 11/18/06, Erik Ohrnberger <Erik AT echohome.org> wrote:
> > > > > > > Generally, I'm running an X Window server emulator on
> > > > my Windows
> > > > > > > PC and pipe the display up there.  The PC is set to 32
> > > > bit color
> > > > > > > depth, and when the DISLPAY env var is set to the IP of the
> > > > > > > Windows PC, that is when the applications fail with
> > this error.
> > > > > > >
> > > > > > > Is there a setting someplace on the Linux side where I
> > > > can force a
> > > > > > > similar color depth as to the Option "Pixmap" "32" you
> > > > mention below?
> > > > > > >
> > > > > > > Thanks for the help, I really appreciate it.
> > > > > >
> > > > > > Hmmm...your question is outside the scope of this
> > list, as your
> > > > > > problem is not cairo-related. You'll find better answers here
> > > > > > (especially the cygwin-xfree part):
> > > > >
> > > > > Well, it is sort of cairo related -- these X severs /are/
> > > > out there (including Xvnc and various others), and having cairo
> > > > assert out on them is not a good solution, IMO.
> > > >
> > > > Yea, totally agree. I was reponding to his question about
> > cygwin/X.
> > > > Sorry for the confusion, I didn't mean to imply that no
> > one is going
> > > > to address the root of the problem. As I said in my first
> > response
> > > > to Erik, I think this should problem should be in
> > bugzilla and get
> > > > the proper attention.
> > > >
> > > > Dan
> > > >
> > >
> > > I have to admit that it's not cygwin/X, but an old program called
> > > VisionWare, and has served me faithfully for a long number of years.
> >
> > Ah, sorry for jumping to conclusions. Unfortunately, I have
> > no idea how to get VisionWare's XVision into 32bpp mode. In
> > fact, the following posting seems to say that XVision doesn't
> > support 32bpp Pixmaps at all :(
> >
> > http://www.cygwin.com/ml/cygwin-xfree/2002-04/msg00366.html
> >
> > So I'm all out of ideas for workarounds. Sorry.
> >
> > > Specific version information that could proove helpful:
> > >
> > > gtkmm-2.8.1, gtk+-2.6.10, cairo-0.1.6           works
> > >
> > > gtkmm-1.2.9-r2, gtk+-2.8.19, cairo-1.0.4
> > does not work
> >
> > Cairo has changed so much from 0.1.6 to 1.0.4, especially in
> > the place in xlib-surface where the problem is. I can see the
> > before and after code paths, but it doesn't help solve the
> > problem. Thing is, libpixman in now used in places that
> > previously only contained X calls. And it's pixman's lack of
> > support for 24bpp that is the root of the problem. At least
> > that's my understanding.
> >
> > > Should I go and put this information in bugzilla?  I'd be
> > willing to
> > > do so, but I'd have to find out where the bugzilla site is.
> >
> > Here:
> >
> > https://bugs.freedesktop.org/enter_bug.cgi?product=cairo
> >
> > Thanks for reporting this.
> >
> > Carl/whoever is listening: could a quick stopgap measure be
> > to do a conversion from 24bpp to 32bpp before the image is
> > handed over to pixman (or write a conversion function in
> > pixman that cairo-xlib can call)? Ideally, one would want
> > pixman to support 24bpp directly, but since that's a lot of
> > work, and this isn't a common case, it might be worth
> > considering a slow and messy approach. People don't call
> > _acquire_(source|dest)_image in their display loop anyways, right?
> >
> > Dan
>
> Well, I put the bug in the database as best I could.  I'm sure that the
> receiver of the bug will make some adjustments to fix it up right.
>
> I understand your position, as to pixman not supporting 24 bit, and the fact
> that it would take quite of a bit of code to handle this peculiar situation,
> but then, that doesn't really make me want to have it working any less . . .
> :{)       But then, cairo doesn't really revolve around my wants either.
> :{(
>
> I'd have to agree with the sentiments that 24 bits is a valid situation.
> Just think of some of the systems out there with graphic cards that can't be
> configured to 32 bits, but could support 24 bits.  They'd be in pretty much
> the same boat, wouldn't they?

Yes, AFAIK, cairo right now is broken on all X servers that use 24bpp
for Pixmaps. And I would imagine that many of those servers are that
way because the graphics card doesn't support 32bpp.

On the bright side, the stopgap measure that I proposed in my previous
email should actually be pretty straightforward to implement. In case
it wasn't clear, I was proposing that we can avoid all that work in
pixman if we adopt this other quick (to implement), slow (to run) and
dirty (to look at) approach.

Dan