[Mesa-dev] Mesa (master): tnl: Initialize gl_program_machine memory in run_vp.

[Ian Romanick]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 12/13/2010 07:27 AM, Keith Whitwell wrote:
> On Mon, 2010-12-13 at 07:09 -0800, Brian Paul wrote:
>> On 12/10/2010 03:27 PM, Vinson Lee wrote:
>>> Module: Mesa
>>> Branch: master
>>> Commit: ef3f7e61b314236cbb7ed2cf24d34c6f90d9cfca
>>> URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=ef3f7e61b314236cbb7ed2cf24d34c6f90d9cfca
>>>
>>> Author: Vinson Lee<vlee at vmware.com>
>>> Date:   Fri Dec 10 14:24:05 2010 -0800
>>>
>>> tnl: Initialize gl_program_machine memory in run_vp.
>>>
>>> Fixes piglit valgrind glsl-array-bounds-04 failure (FDO bug 29946).
>>>
>>> NOTE:
>>> This is a candidate for the 7.10 branch.
>>> This is a candidate for the 7.9 branch.
>>>
>>> ---
>>>
>>>   src/mesa/tnl/t_vb_program.c |    2 +-
>>>   1 files changed, 1 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c
>>> index 76f8fde..7e7c59a 100644
>>> --- a/src/mesa/tnl/t_vb_program.c
>>> +++ b/src/mesa/tnl/t_vb_program.c
>>> @@ -311,7 +311,7 @@ run_vp( struct gl_context *ctx, struct tnl_pipeline_stage *stage )
>>>      struct vp_stage_data *store = VP_STAGE_DATA(stage);
>>>      struct vertex_buffer *VB =&tnl->vb;
>>>      struct gl_vertex_program *program = ctx->VertexProgram._Current;
>>> -   struct gl_program_machine machine;
>>> +   struct gl_program_machine machine = { 0 };
>>>      GLuint outputs[VERT_RESULT_MAX], numOutputs;
>>>      GLuint i, j;
>>
>> I think there's a better fix.  The above will initialize the whole 
>> object to zeros for every function call (and be a performance hit).  I 
>> think we really only need to do it once to avoid the valgrind warning.
>>
>> I've got a new patch that I'll commit.
> 
> Also, this idiom of partially initializing structures with {0} seems to
> cause gcc to squawk about all the fields which weren't included in the
> initializer.

What's the specific warning?  That seems odd as the C spec explicitly
says what this is supposed to do for both arrays and structures.  It's
particularly convenient when used with C99 designated initializers.

I'm mostly just curious here.  I agree with Brian's assessment of the patch.

> MSVC apparently doesn't complain, and it is a convenient idiom providing
> the compiler likes it.
> 
> On balance, I like getting those warnings from gcc as they can spot real
> bugs elsewhere, so I'd prefer not to have code which causes us to want
> to turn the warning off...

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk0GZNMACgkQX1gOwKyEAw87iQCgkePZN61YPk4kb96MDI709Rbf
9x8AnRVk+2O0l9AT8ECfJTKSkL8YT/To
=C08x
-----END PGP SIGNATURE-----