Had some flooding (leaking pipes, rather values) issues to fix. Will 
look at this asap.

On 04-Aug-16 10:58, Martin Panter wrote:
> Martin Panter added the comment:
>
> Okay, so to be clear, I am assuming XLC supports all of the following fields, and uses unsigned bit fields by default:
>
> struct UNSIGNED_BITS {
>      unsigned int AU: 1;
>      int A: 1; /* Equivalent to unsigned int */
>      signed int AS: 1;
>      unsigned short MU: 1;
>      short M: 1; /* Equivalent to unsigned short; triggers warning */
> };
>
> and that it cannot compile the following:
>
> struct SIGNED_BITS {
>      signed short MS: 1; /* Not supported */
> };
>
> Attached is what I think a patch to resolve this would look like. However it needs a line in Modules/_ctypes/_ctypes_test.c completed to detect the compiler:
>
> #ifndef /* Something to identify XLC */
>
> Can you figure out a way to test for XLC (but not GCC, which the AIX buildbot uses), and then try my patch out? Hopefully you see no more compiler warnings, test_ints() should now pass, and test_shorts() should be skipped.
>
> ----------
> keywords: +patch
> Added file: http://bugs.python.org/file44005/disable-signed-short.patch
>
> _______________________________________
> Python tracker <report@bugs.python.org>
> <http://bugs.python.org/issue27643>
> _______________________________________