[Bug 20448] New: Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

Summary: Touhou 7 and 10 segfault at startup with Intel drivers
                    >= 2.7, 64-bit
           Product: Wine
           Version: 1.1.32
          Platform: PC
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: directx-d3d
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Created an attachment (id=24300)
 --> (http://bugs.winehq.org/attachment.cgi?id=24300)
Adds a NULL pointer check before dumping FBO attachments

Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit. The
reason for this is context_check_fbo_status() being passed a NULL pointer in
context->current_fbo by content_apply_fbo_state().

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #1 from Chow Loong Jin   2009-10-23 17:25:22 ---
Created an attachment (id=24301)
 --> (http://bugs.winehq.org/attachment.cgi?id=24301)
Terminal output

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #2 from Chow Loong Jin   2009-10-23 17:25:43 ---
Created an attachment (id=24302)
 --> (http://bugs.winehq.org/attachment.cgi?id=24302)
Backtrace of segfault.

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

Chow Loong Jin  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #24302|application/octet-stream    |text/plain
          mime type|                            |

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #3 from Henri Verbeet   2009-10-23 19:33:28 ---
That situation can only happen when your graphics driver setup is broken. The
window system provided framebuffer can never be invalid, and 0 is not a valid
return code for glCheckFramebufferStatus().

I'd be ok with printing an ERR instead of crashing, but I doubt it'll make the
application work correctly. I think there's already a bug filed against Mesa
for this.

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

Vitaliy Margolen  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
          Component|directx-d3d                 |-unknown
         Resolution|                            |INVALID

--- Comment #4 from Vitaliy Margolen   2009-10-23 20:12:41 ---
Invalid then per comment #3.

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

Vitaliy Margolen  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #5 from Vitaliy Margolen   2009-10-23 20:12:50 ---
Closing - not Wine bug.

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #6 from Chow Loong Jin   2009-10-24 00:45:56 ---
On the contrary, the attached patch makes the said applications work very
nicely indeed. Would you be fine with accepting a patch that prints an ERR
instead then?

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #7 from Chow Loong Jin   2009-10-24 01:10:11 ---
Actually digging around context.c has led me to see one case where
context->current_fbo can remain NULL.
> From context_apply_fbo_state(),

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #8 from Chow Loong Jin   2009-10-24 01:21:49 ---
Sorry, please ignore my previous comment. It was submitted prematurely.

Actually digging around context.c has led me to see one case where
context->current_fbo can remain NULL.
> From context_apply_fbo_state(), if !context->render_offscreen,

context->current_fbo = NULL;
context_bind_fbo (context, GL_FRAMEBUFFER, NULL);

In context_bind_fbo(),
f = 0;                                             
if (context->fbo_read_binding == f && context->fbo_draw_binding == f) return;   
context->fbo_read_binding = f;                                      
context->fbo_draw_binding = f;
gl_info->fbo_ops.glBindFramebuffer(target, f);

The above code doesn't touch context->current_fbo, does it? Unless I'm mistaken
and somewhere in gl_info->fbo_ops.glBindFramebuffer(target, f) sets it somehow?

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #9 from Henri Verbeet   2009-10-24 05:54:01 ---
Created an attachment (id=24309)
 --> (http://bugs.winehq.org/attachment.cgi?id=24309)
patch

Yes, it's NULL if the application is rendering onscreen. But in that case
"status" in context_check_fbo_status() should *always* be
GL_FRAMEBUFFER_COMPLETE. I'll submit something like the attached patch.

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #10 from Chow Loong Jin   2009-10-24 07:43:35 ---
That'd be cool. Something I noticed after applying the patch is that
performance really sucks unless I disable all the err and fixme messages via
`export WINEDEBUG=fixme-all,err-all` prior to starting the game, but that's
another issue.

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #11 from Henri Verbeet   2009-10-26 04:02:06 ---
Yes, debug output can slow things down quite a bit. If you care about
performance you usually want to run with WINEDEBUG="-all".

For what it's worth, the relevant Mesa bug report is
http://bugs.freedesktop.org/show_bug.cgi?id=23335. It seems there's at least a
configuration issue on your side as well, since you generally want to avoid
indirect rendering. Is your 32-bit libGL setup properly?

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #12 from Chow Loong Jin   2009-10-26 04:10:48 ---
I'm pretty sure my configuration is quite fine. In fact, I managed to play
Counter Strike 1.6 without the patch, using the OpenGL rendering method, as
opposed to using D3D.

Also, I think I'm having direct rendering on -- `glxinfo | grep Direct` says
yes.

Is there a way I could check if my 32-bit libGL stuff are fine?

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #13 from Henri Verbeet   2009-10-26 04:14:40 ---
(In reply to comment #12)
> Also, I think I'm having direct rendering on -- `glxinfo | grep Direct` says
> yes.

Yeah, but that uses your 64-bit GL setup.
> Is there a way I could check if my 32-bit libGL stuff are fine?

Runnning with WINEDEBUG="+wgl" should show if you're using direct rendering or
not.

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #14 from Chow Loong Jin   2009-10-26 04:52:22 ---
trace:wgl:X11DRV_WineGL_InitOpenglInfo Direct rendering enabled: False

Hmm, it does look like direct rendering is disabled after all. I guess I'm
missing some lib32 library or other. What should I be looking for?

[Bug 20448] Touhou 7 and 10 segfault at startup with Intel drivers >= 2.7, 64-bit

http://bugs.winehq.org/show_bug.cgi?id=20448

--- Comment #15 from Henri Verbeet   2009-10-26 04:58:40 ---
(In reply to comment #14)
> trace:wgl:X11DRV_WineGL_InitOpenglInfo Direct rendering enabled: False
>
> Hmm, it does look like direct rendering is disabled after all. I guess I'm
> missing some lib32 library or other. What should I be looking for?

That's really a question you should be asking somewhere specific for your
distro instead of Wine bugzilla, but you probably need a 32-bit Mesa or video
driver package.