[cfe-dev] sizeof(long) in OpenCL C

[David Tweed]

Hi,

Apologies for not pointing Erik at the patch directly (annoyances of on
being in the UK time zone and all that...)

> Could this be reconsidered?

| It looks more like the idea fizzled than was rejected. I also think
| it's a good plan for anything that's specified by the language.

My reading of things was that there was a general feeling that this wasn't
something the other OpenCL implementers felt belonged in open source clang,
so I didn't feel it was appropriate to push to change hearts and minds; but
maybe that has changed. Having said that, one thing that I think has
happened is that quite how OpenCL (and associated things like SPIR) split
between the open source clang/LLVM and implementers choices stuff is
something that might be something worthy of an actual discussion. (This is
particularly relevant since the clang frontend does a comparatively large
amount of code processing in order to detect and issue good diagnostics.) It
might be something that interested parties might want to get together to
discuss at the upcoming North American LLVM developer meeting: certainly if
there's an interest one of the ARM attendees would join the discussion and
try to help make the situation better.

Cheers,
Dave