[cmake-developers] [PATCH] Fix broken PGI builds for liblzma.

[Chuck Atkins]

On Fri, Aug 1, 2014 at 2:43 PM, "C. Bergström" <cbergstrom at pathscale.com>
wrote:

> I'm not intimately following this issue, but I can get you access to
> pathcc/pathCC to see if we work on this as well.


I've got an older ekopath 4.0.9 and 4.0.10 install that you gave us a while
back when I was setting up some LAPACK testing.  The build was fine without
this patch so no issues there.  It would be nice to get an updated version
though going on a nightly dashboard.  Let's discuss further off-list.



> (C99 doesn't work "correctly" on some of the older linux) So while "c99"
> is claimed to be supported - it was in fact broken. SLES10 is one example I
> know 1st hand - there are others as well.


In this particular case it's a C99 language feature not library or include
issue so I wouldn't expect the OS version to matter but I suppose could.
We've seen similar issues with trying to use AVX instructions on RHEL5.
MAny compilers will detect the CPU as supporting it and try to generate the
code but the kernel doesn't support the instructions and will segfault so
we need to explicitly disable AVX optimizations on RHEL5.


feature/functional tests imho are always preferred vs #if..
>

If the "if C99" tests are actually reliable and I could trust that C99
would work then I'd rather since it really bogs down the configure time.
That being said though, in this case it seems to not be since
__STDC_VERSION__ is getting set to 199901L by the compiler but clearly not
all C99 features are working.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/cmake-developers/attachments/20140801/b882caa2/attachment-0002.html>