I just sent out a few updated patches (latest can be seen in
https://github.com/imirkin/mesa/commits/fp64-5) -- with all that in
place the mysterious failure in one of the initializer tests is gone,
and the constbuf locations being used correspond to the ones that we
expect instead of having *tons*
This is v3 of the patchset which includes fixes based on Ian's
feedback with the exception of his comment on patch 23 (glsl: validate
output types for shader stages), i.e. whether it needs to exist at all
and isn't that being checked elsewhere. I am unfortunately not
familiar enough with the overal
