Fix PR15293: ARM codegen ice - expected larger existing stack allocation

[Renato Golin]

On 28 March 2013 12:06, Stepan Dyatkovskiy <stpworld at narod.ru> wrote:

> I'm not sure about time, since I don't know some pitfalls, may be. But I
> still want try to fix llvm and clang :-)
>

Great! That's the spirit! ;)


You say it was fixed once and rejected then. Did somebody guys remember
> some thread names or revisions, may be?
>

It wasn't fixed, it had partial support and some people (including me)
tried to use only to find broken stuff down the pipe. IIRC it was disabled
(possibly removed from source) to avoid people doing it again. You may find
some asserts along the way, too.

I think the best way to proceed is to make Clang lower struct byval for ARM
as it does on x86 and go fixing the broken tests one by one.

I'm not sure how Clang folks would prefer, a huge patch at the end with it
enabled by default, or a flag --use-broken-byval-on-arm (disabled by
default) for now until all tests pass with it, when you can remove the flag
altogether. You better ask them directly on cfe-dev at .

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130328/245c0e07/attachment.html>