[Bug target/50678] [4.7/4.8 Regression] FAIL: c52104y on x86_64-apple-darwin10

[simon at pushface dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50678

simon at pushface dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #29360|0                           |1
        is obsolete|                            |

--- Comment #73 from simon at pushface dot org 2013-02-05 22:37:36 UTC ---
Created attachment 29361
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=29361
Patch to suppress register swap on Darwin >= 12, v2

(In reply to comment #72)

> We should probably check the return value of sysctl and immediately return 0 if
> there is a problem (it would be up to the caller to decide what to do with 0).
> Then I don't think that we need to check the length.

Done.

I do wonder when Apple introduced this problem. It was first reported with
Darwin 10, but GCC 4.6 and GNAT GPL 2011, 2012 (both based on GCC 4.5) don't
show the problem, so would we have known? Should the test be 10 <= version <12

> Minor points:
> 
>  1. This should be
> 
> static int
> __darwin_major_version (void)
> 
>  2. I think that you can pass NULL to strtol instead of dot_p.
> 
>  3. Double space before star-slash at the end of comments.
> 
>  4. No () for functions in comments.

Done.

> > As an aside, what's the recommended technique for rebuilding and reinstalling
> > just the RTS? ('cd gcc; make gnatlib' leaves the build in a state where 'make
> > install' doesn't work).
> 
> rm $(target)/libada/stamp-libada gcc/stamp-gnatlib*
> make all-target-libada

and 'make install-target-libada'.