[Mesa-dev] [Review Request (master branch)] st/mesa: Fix topogun-1.06-orc-84k-resize.trace crash

[Neha Bhende]

> Would it be wise to just initialize everything explicitly? For example
> this is what _mesa_draw_arrays does:
>
>    memset(&prim, 0, sizeof(prim));
>    prim.begin = 1;
>    prim.end = 1;
>    prim.mode = mode;
>    prim.num_instances = numInstances;
>    prim.base_instance = baseInstance;
>    prim.draw_id = drawID;
>    prim.is_indirect = 0;
>    prim.start = start;
>    prim.count = count;
>
> and vbo_exec_Begin has:
>
>    exec->vtx.prim[i].mode = mode;
>    exec->vtx.prim[i].begin = 1;
>    exec->vtx.prim[i].end = 0;
>    exec->vtx.prim[i].indexed = 0;
>    exec->vtx.prim[i].pad = 0;
>    exec->vtx.prim[i].start = exec->vtx.vert_count;
>    exec->vtx.prim[i].count = 0;
>    exec->vtx.prim[i].num_instances = 1;
>    exec->vtx.prim[i].base_instance = 0;
>    exec->vtx.prim[i].is_indirect = 0;
>
> I do think that your change is all that's strictly needed though.
> Sorry about the breakage!


I think you are right, we should initialize all fields explicitly here. I will send out patch for this and

also modify commit message as per suggested by Brian.


Regards,

Neha

________________________________
From: Ilia Mirkin <imirkin at alum.mit.edu>
Sent: Tuesday, January 29, 2019 10:58:01 AM
To: Neha Bhende
Cc: Brian Paul; ML Mesa-dev
Subject: Re: [Mesa-dev] [Review Request (master branch)] st/mesa: Fix topogun-1.06-orc-84k-resize.trace crash

Would it be wise to just initialize everything explicitly? For example
this is what _mesa_draw_arrays does:

   memset(&prim, 0, sizeof(prim));
   prim.begin = 1;
   prim.end = 1;
   prim.mode = mode;
   prim.num_instances = numInstances;
   prim.base_instance = baseInstance;
   prim.draw_id = drawID;
   prim.is_indirect = 0;
   prim.start = start;
   prim.count = count;

and vbo_exec_Begin has:

   exec->vtx.prim[i].mode = mode;
   exec->vtx.prim[i].begin = 1;
   exec->vtx.prim[i].end = 0;
   exec->vtx.prim[i].indexed = 0;
   exec->vtx.prim[i].pad = 0;
   exec->vtx.prim[i].start = exec->vtx.vert_count;
   exec->vtx.prim[i].count = 0;
   exec->vtx.prim[i].num_instances = 1;
   exec->vtx.prim[i].base_instance = 0;
   exec->vtx.prim[i].is_indirect = 0;

I do think that your change is all that's strictly needed though.
Sorry about the breakage!

On Tue, Jan 29, 2019 at 1:31 PM Neha Bhende <bhenden at vmware.com> wrote:
>
> This patch fixes regression caused by
> bac8534267- st/mesa: allow glDrawElements to work with GL_SELECT feedback
> We also need to set instance count while creating new rastpos stage.
>
> Reviewed-by: Brian Paul <brianp at vmware.com>
> ---
>  src/mesa/state_tracker/st_cb_rasterpos.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c
> index fecaaf7..cca2fba 100644
> --- a/src/mesa/state_tracker/st_cb_rasterpos.c
> +++ b/src/mesa/state_tracker/st_cb_rasterpos.c
> @@ -206,6 +206,7 @@ new_draw_rastpos_stage(struct gl_context *ctx, struct draw_context *draw)
>     rs->prim.indexed = 0;
>     rs->prim.begin = 1;
>     rs->prim.end = 1;
> +   rs->prim.num_instances = 1;
>     rs->prim.start = 0;
>     rs->prim.count = 1;
>
> --
> 2.7.4
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Fmesa-dev&data=02%7C01%7Cbhenden%40vmware.com%7Ca611d0289a334ee9273708d6861bb82d%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C0%7C0%7C636843850945848627&sdata=03tQMDRMwfagcQPeALU9Ox2txEb0A8ajAxj2CUgtmIk%3D&reserved=0
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20190129/23015da5/attachment.html>