[Bug 54832] New: Regression: 185ed2105829d6f5eb19edb9abbf0d7977e157c3 breaks swtcl on r300g

[Bug 54832] New: Regression: 185ed2105829d6f5eb19edb9abbf0d7977e157c3 breaks swtcl on r300g

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

             Bug #: 54832
           Summary: Regression: 185ed2105829d6f5eb19edb9abbf0d7977e157c3
                    breaks swtcl on r300g
    Classification: Unclassified
           Product: Mesa
           Version: git
          Platform: Other
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: medium
         Component: Drivers/Gallium/r300
        AssignedTo: dri-devel@lists.freedesktop.org
        ReportedBy: tstellar@gmail.com
                CC: brianp@vmware.com


Lightsmark stopped working on my r300 laptop with swtcl following this commit:

commit 185ed2105829d6f5eb19edb9abbf0d7977e157c3
Author: Brian Paul <brianp@vmware.com>
Date:   Fri May 25 09:44:53 2012 -0600

    draw: simplify index buffer specification

    Replace draw_set_index_buffer() and draw_set_mapped_index_buffer() with
    draw_set_indexes() which simply takes a pointer and an index size.


When I start Lightsmark, it crashes during the "loading..." screen after a few
seconds, before the demo starts rendering the scene.  Prior to the crash, the
debug statement ("warning index out of range") @ draw_pt_vsplit_tmp.h:59 is
printed several times, followed by the message:

draw_pipeline_run: max_index (7589) outside vertex buffer (300)

and then a segfault.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug 54832] Regression: 185ed2105829d6f5eb19edb9abbf0d7977e157c3 breaks swtcl on r300g

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

--- Comment #1 from Tomasz P. <son_of_the_osiris@interia.pl> 2012-09-13 13:05:08 UTC ---
Someone on the mailing list suspects that in r300 compiler is somekind of bug.
He wrote:

In playing with Coccinelle, I discovered a signed/unsigned bug in
radeon_rename_regs.c:rc_rename_regs.

unsigned new_index;
unsigned writemask;
struct rc_variable * var = var_ptr->Item;

if (var->Inst->U.I.DstReg.File != RC_FILE_TEMPORARY) {
        continue;
}

new_index = rc_find_free_temporary_list(c, used, used_length,
                                                RC_MASK_XYZW);
if (new_index < 0) {
        rc_error(c, "Ran out of temporary registers\n");
        return;
}

unsigned new_index is compared with < 0.

I don't know the code, but I can't imagine that you'd need an unsigned
to represent a register index value.

Matt Turner

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug 54832] Regression: 185ed2105829d6f5eb19edb9abbf0d7977e157c3 breaks swtcl on r300g

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

--- Comment #2 from Tom Stellard <tstellar@gmail.com> 2012-09-13 13:26:16 UTC ---
(In reply to comment #1)
> Someone on the mailing list suspects that in r300 compiler is somekind of bug.
> He wrote:
> 
> In playing with Coccinelle, I discovered a signed/unsigned bug in
> radeon_rename_regs.c:rc_rename_regs.
> 
> unsigned new_index;
> unsigned writemask;
> struct rc_variable * var = var_ptr->Item;
> 
> if (var->Inst->U.I.DstReg.File != RC_FILE_TEMPORARY) {
>         continue;
> }
> 
> new_index = rc_find_free_temporary_list(c, used, used_length,
>                                                 RC_MASK_XYZW);
> if (new_index < 0) {
>         rc_error(c, "Ran out of temporary registers\n");
>         return;
> }
> 
> unsigned new_index is compared with < 0.
> 
> I don't know the code, but I can't imagine that you'd need an unsigned
> to represent a register index value.
> 
> Matt Turner

Please file a separate bug for this.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug 54832] Regression: 185ed2105829d6f5eb19edb9abbf0d7977e157c3 breaks swtcl on r300g

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

--- Comment #3 from Tomasz P. <son_of_the_osiris@interia.pl> 2012-09-13 14:21:30 UTC ---
done, bug: 54867

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug 54832] Piglit Regression: r300g swtcl 185ed21 draw: simplify index buffer specification

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

Tom Stellard <tstellar@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Regression:                 |Piglit Regression: r300g
                   |185ed2105829d6f5eb19edb9abb |swtcl 185ed21 draw:
                   |f0d7977e157c3 breaks swtcl  |simplify index buffer
                   |on r300g                    |specification

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug 54832] Piglit Regression: r300g swtcl 185ed21 draw: simplify index buffer specification

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

--- Comment #4 from Tom Stellard <tstellar@gmail.com> 2012-09-17 00:43:08 UTC ---
This commit also regresses the following piglit tests:

Segfault:

primitive-restart-DISABLE_VBO
primitive-restart-VBO_COMBINED_VERTEX_AND_INDEX
primitive-restart-VBO_INDEX_ONLY
primitive-restart-VBO_SEPARATE_VERTEX_AND_INDEX
primitive-restart-VBO_VERTEX_ONLY

Fail:

draw-batch

Note: This can be reproduced on any r300g supported hardware by setting the
environment variable: RADEON_NO_TCL=1

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug 54832] Piglit Regression: r300g swtcl 185ed21 draw: simplify index buffer specification

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

--- Comment #5 from Tom Stellard <tstellar@gmail.com> 2012-09-17 00:46:53 UTC ---
Created attachment 67261
  --> https://bugs.freedesktop.org/attachment.cgi?id=67261
Backtrace from primitive-restart-DISABLE_VBO piglit test

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug 54832] Piglit Regression: r300g swtcl 185ed21 draw: simplify index buffer specification

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

--- Comment #6 from Marek Olšák <maraeo@gmail.com> 2012-09-17 01:37:25 UTC ---
This should fix this:

http://lists.freedesktop.org/archives/mesa-dev/2012-September/027435.html

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 54832] Piglit Regression: r300g swtcl 185ed21 draw: simplify index buffer specification

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

--- Comment #7 from Tom Stellard <tstellar@gmail.com> 2012-09-17 02:13:28 UTC ---
(In reply to comment #6)
> This should fix this:
> 
> http://lists.freedesktop.org/archives/mesa-dev/2012-September/027435.html

It does fix it, thanks!

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug 54832] Piglit Regression: r300g swtcl 185ed21 draw: simplify index buffer specification

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=54832

Marek Olšák <maraeo@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #8 from Marek Olšák <maraeo@gmail.com> 2012-09-22 14:34:07 UTC ---
I pushed the fix into master and 9.0. Closing.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel