On Thu, Oct 19, 2017 at 4:51 PM, Kenneth Graunke <kenn...@whitecape.org> wrote:
> On Wednesday, October 18, 2017 10:31:53 PM PDT Jordan Justen wrote: > > Signed-off-by: Jordan Justen <jordan.l.jus...@intel.com> > > --- > > src/compiler/glsl/builtin_variables.cpp | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/src/compiler/glsl/builtin_variables.cpp > b/src/compiler/glsl/builtin_variables.cpp > > index ea2d897cc8..d3cf12475b 100644 > > --- a/src/compiler/glsl/builtin_variables.cpp > > +++ b/src/compiler/glsl/builtin_variables.cpp > > @@ -318,6 +318,7 @@ per_vertex_accumulator::per_vertex_accumulator() > > : fields(), > > num_fields(0) > > { > > + memset(fields, 0, sizeof(fields)); > > } > > This shouldn't fix anything, we're calling the fields() constructor, > which should value-initialize every element of the fields array, > calling the constructor for glsl_struct_type on each element. > > That constructor is supposed to zero initialize items. Maybe it is > missing some of them? > No, the problem is that, when we do the user data check in the blob code, valgrind complains about the padding bytes in-between field elements. C++ constructors don't initialize that data. That said, I agree that this is a tad sketchy. I just wish there were some way to shut up valgrind.
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev