Bug 96950

Do you still see the issue after 64d35f81 "vbo: fix attr reset"?

(In reply to Rob Clark from comment #1)
> Do you still see the issue after 64d35f81 "vbo: fix attr reset"?

Yes.

Created attachment 125104 [details] [review]
Patch for debugging

The remaining thing that I could see to get into this situation would be that  newSize is equal to zero on entry to in vbo_exec_wrap_upgrade_vertex.
The only case where I could not rule out newSize being zero just now is with evaluators.
If this is our problem, the attached patch may bring us to a backtrace that matters.
Can you help me verify that the above thought heads into the right direction?

Mathias

Or Can you provide an apitrace?
I have no such system to reproduce at hands.

Created attachment 125255 [details] [review]
Fix for an other assert in immediate mode rendering with the linux turbine

Brian,

I tried to reproduce the described problem with the linux demo and stepped ontpo an other assert in vbo_exec_vtx_wrap(). This assert may be related to the one you mentioned. With this patch applied the linux turbine demo runs fine on i965.
Can you check if the windows demo gets fixed by this patch too?

Rob,

The patch fixes the problem you observed with 0ad in a different way. Can you check if this approach also fixes your problem?

Thanks

Mathias

(In reply to Mathias Fröhlich from comment #6)
> Created attachment 125255 [details] [review] [review]
> Fix for an other assert in immediate mode rendering with the linux turbine
> 
> Rob,
> 
> The patch fixes the problem you observed with 0ad in a different way. Can
> you check if this approach also fixes your problem?

it does not seem to regress 0ad..

fwiw, if you are curious, this should reproduce my original problem (if you revert my earlier fix and don't apply your patch)

https://people.freedesktop.org/~robclark/0ad-cycladic-archipelago.trace.xz

(In reply to Mathias Fröhlich from comment #6)
> Created attachment 125255 [details] [review] [review]
> Fix for an other assert in immediate mode rendering with the linux turbine
> 
> Brian,
> 
> I tried to reproduce the described problem with the linux demo and stepped
> ontpo an other assert in vbo_exec_vtx_wrap(). This assert may be related to
> the one you mentioned. With this patch applied the linux turbine demo runs
> fine on i965.
> Can you check if the windows demo gets fixed by this patch too?

Yeah, your patch seems to fix things.  Thanks!

Tested-by: Brian Paul <brianp@vmware.com>

As for the patch itself, I think we want to s/boolean/bool/ and maybe #include <stdbool.h> just to be safe.

Otherwise, Reviewed-by: Brian Paul <brianp@vmware.com>

I have verified 0ad with the provided trace here, updated as requested and pushed.
Thanks for testing and review on your side.

I assume the originator verifies and closes the bug?

Mathias

Verified.  Thanks!