On Mon, May 23, 2016 at 7:49 PM, Matt Turner <[email protected]> wrote:
> On Mon, May 23, 2016 at 6:43 PM, Jason Ekstrand <[email protected]> > wrote: > > This prevents SPIR-V -> NIR from dying if we get a decoration we don't > know > > about. Compiling incorrect code is probably a better failure mode than > > dying horribly. > > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95573 > > We want to try to limp along after compiling a shader whose meaning we > can't determine? > > Maybe I don't understand, but shouldn't we just support whatever > decoration is missing? > We support all of the decorations that have meaning in the given context. However, sometimes glslang will throw on a pointless decoration we don't need. Sometimes hitting that means there's a bug (hence the warning) but sometimes it means we're getting a junk decoration. Ultimately, we should probably have every possible decoration in the switch with most of them going nowhere; I just haven't bothered to do a full audit of everything.
_______________________________________________ mesa-dev mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/mesa-dev
