Xaos display problem with compiz

I'm using Gutsy updated.

My graphic card is an ATI IGP340M working with open source driver "ati"

I can confirm this with the intel driver. It works fine with the (binary-only) nvidia one.

It seems to be not releated to XSHM. I suspect its something with the color managment that leads it to generate a alpha channel for some reason.

I suspect a color managment problem as well.
I tried to launch xaos from Hardy alpha 3 (with some updates in the middle) in VirtualBox (so no compiz in it)

No colors at all.
See screenshot.

Closing compiz bug since it has problems without compiz as well.

@Travis

In Hardy it has problem without compiz.

In Gutsy it has problem only with compiz.

2008/2/19, Travis Watkins <email address hidden>:
> Closing compiz bug since it has problems without compiz as well.
>
> ** Changed in: compiz (Ubuntu)
> Status: New => Invalid
>
> --
> Xaos display problem with compiz
> https://bugs.launchpad.net/bugs/192882
> You received this bug notification because you are a direct subscriber
> of the bug.
>

@Travis

Just to clarify:
If you see the picture (hardy in virtualbox) and the video (Gutsy on my laptop) you see that xaos have different color behaviour.
So they could be two distinct bug due to different situations.
I think that before putting compiz issue as invalid it would be better to investigate more on the problem and, in case of two bugs, I'll open another one for Hardy.

I'm using Hardy Alpha 6 live cd and it's the bug is still present (see screenshot)
It's the same behaviour as Gutsy.
I think that Virtualbox problem (that seems to be different) is related to VirtualBox itself.

Xaos is doing something stupid here. It is asking for transparency and it is getting it. They probably never realized because you couldn't do transparency before.

This is a somewhat hacky way to fix this problem. Since xaos doesn't actually want alpha just disable it.

For these kinds of workarounds, I'd suggest a shell wrapper instead of changing the code.

The "visuals" code in xaos dates from a time when transparency was not thought of yet. It would be good if someone knowledgeable could go through it. See xalloc_display() in src/ui/ui-drv/x11/xlib.c. Is there an easy and good way to avoid transparency?

XLIB_SKIP_ARGB_VISUALS basically disables 32 bit depth visuals, so the equivalent workaround in xaos would be to not even try to match on those:

--- ../xlib.c.orig 2008-03-07 21:10:11.000000000 +0100
+++ src/ui/ui-drv/x11/xlib.c 2008-03-07 21:12:57.000000000 +0100
@@ -337,7 +337,7 @@ xalloc_display (CONST char *s, int x, in
   new->params = params;

   found = 0;
- for (i = 32; i > 13 && !found; i--)
+ for (i = 31; i > 13 && !found; i--)
     if (XMatchVisualInfo (new->display, new->screen, i, TrueColor, &vis))
       {
        found = 1;

Thanks for the patch!

I sponsored Travis debdiff for now. A better upstream fix would be better of course.

@Tormod: what are the chances that this might go upstream? is there a upstream active still?

Thanks,
 Michael

Michael, can you please look at the pending merge in bug #197261? I am afraid your upload screwed that debdiff a little again, but only the changelog.

Upstream is not dead, but they're mostly busy with a new-generation fork that eventually will be merged in. So I don't see this get in upstream very soon. We could try to get it into Debian, but I guess they would prefer a wrapper or a proper fix(TM).

I opened another bug for display problem with VirtualBox, due to probably different cause.
Bug #200650

My patch above went in upstream, so I included it in the merge in bug #197261.

This bug was fixed in the package xaos - 3.2-7ubuntu1

---------------
xaos (3.2-7ubuntu1) hardy; urgency=low

  * Merge from debian unstable (LP: #197261), remaining changes:
    - Drop build dependency on libsvga1-dev.
    - Add build dependency on sharutils.
    - Add .desktop file and uuencoded icon.
  * src/ui/ui-drv/x11/ui_x11.c: added window-id option (LP: #189475)
  * src/ui/ui-drv/x11/ui_x11.c: turn off root/fullscreen if windowid
  * src/ui/ui-drv/x11/xlib.c: fix -root and -fullscreen crashers
  * src/ui/ui-drv/x11/xlib.c: fix compiz display problems (LP: #192882)

 -- Tormod Volden <email address hidden> Sat, 01 Mar 2008 13:18:58 +0100

@Tormod

Bad news.

Using XaoS 3.2-9ubuntu1~tormod3 from your PPA in Gutsy I get black screen in XaoS either with or without compiz. (see screenshot, btw I had to take a desktop screenshot because trying window one gnome-screenshot crashes.)

I don't know with 3.2-7ubuntu1, because it's not been backported yet.

Could you provide me this version so I can test it?
Thanks.

I am a co-maintainer of the upstream project. I normally work on the Mac OS X port, but I did apply some patches that Tormod submitted against the newly released 3.3.

One of the patches we recently rolled into 3.3 was Tormod's patch for alpha transparency with Beryl and Compiz, and now we have a bug report that HSV TrueColor modes show all black. Is there any chance that you were using one of those modes when you took the screenshot? I am wondering if these could be related.

See https://sourceforge.net/tracker/index.php?func=detail&aid=1917866&group_id=5771&atid=105771 for more details.

I think it is the same bug.
I get black screen just after starting XaoS which has "0" and "True color
outcoloring mode->hsv2" as defaults for incoloring and outcoloring modes
respectively.

If I select other modes (except hsv or hsv2) for incoloring and outcoloring
I can see results on screen.

xdpyinfo attached if you need.

2008/3/19, J.B. Langston III <email address hidden>:
>
> I am a co-maintainer of the upstream project. I normally work on the
> Mac OS X port, but I did apply some patches that Tormod submitted
> against the newly released 3.3.
>
> One of the patches we recently rolled into 3.3 was Tormod's patch for
> alpha transparency with Beryl and Compiz, and now we have a bug report
> that HSV TrueColor modes show all black. Is there any chance that you
> were using one of those modes when you took the screenshot? I am
> wondering if these could be related.
>
> See
>
> https://sourceforge.net/tracker/index.php?func=detail&aid=1917866&group_id=5771&atid=105771
> for more details.
>
>
> --
> Xaos display problem with compiz
> https://bugs.launchpad.net/bugs/192882
> You received this bug notification because you are a direct subscriber
> of the bug.
>

I think this is a separate bug unrelated to the Beryl/Compiz transparency patch.

The user who reported the HSV issue to us tested the following on Debian sid, and could not get HSV to work on any of them:

1) Unmodified 3.3 sources
2) 3.3 sources with the Beryl/Compiz transparency patch rolled back
3) 3.2 with the Beryl/Compiz transparency patch applied
4) Unmodified 3.2 sources

This last one is the most surprising, but it surely confirms that nothing we changed in 3.3 is causing the problem.

I tried HSV coloring with several versions that I had readily available to me. Here are the results:

1) Mac OS X 3.3 - hsv does NOT work
2) Mac OS X 3.2.2 - hsv does NOT work

3) Windows 3.2 - hsv DOES work
4) Windows 3.3 - hsv DOES work

Most surprising here is the fact that HSV works on Windows 3.3. So again that suggests that nothing we changed in 3.3 inherently caused the problem. Also the fact that Mac 3.2.2 version does not work supports this conclusion.

More likely there is some incompatibility between the X11 and Mac-specific code and the HSV coloring modes.

3.2.2 is the earliest version that is compatible with Mac OS X so I cannot test any further back than that. However, I would be interested to hear what you see when you try earlier versions of XaoS on Linux.

However, we should probably move this discussion to a separate bug (or to the HSV bug on the sourceforge site that I linked earlier), since it does not appear to be related to Beryl/Compiz in any way.

Yes, as I wrote in this comment
https://bugs.launchpad.net/ubuntu/+source/xaos/+bug/192882/comments/19

it appears even without compiz.

So, I put this as fixed.

Could you open a new bug please, so you can explain thing in a more technical and clear way?

Thanks.

You can tell me to test what you want, just ask. :-)

Can we move this HSV bug discussion to the XaoS Sourceforge tracker, please?

https://sourceforge.net/tracker/index.php?func=detail&aid=1917866&group_id=5771&atid=105771

I think that is appropriate since this affects more than just Ubuntu (or Linux, for that matter). If you want to open a new lanuchpad bug for HSV coloring and link to the XaoS tracker, that would be fine by me.

"Fixed"?

I still have this bug, and it doesn't occur under metacity, yet does under compiz.

Should I file a new one, or just attach my logs (and which ones?)

ffm, this was fixed in Hardy. If you still have problems in up-to-date Hardy, please report a new bug and attach /var/log/Xorg.0.log and a screenshot if you can.