Re: Octave's m4/acx_pthread.m4 & OSX

[Ben Abbott]

On Sep 7, 2010, at 7:21 AM, Jarno Rajahalme wrote:

> On Sep 7, 2010, at 3:30 , ext Ben Abbott wrote:
> 
>> I understand the patch is needed for MacOS 10.6, but what of the early 
>> versions?
>> 
>> Thomas / anyone else, can you comment. Is the patch needed for ealier 
>> versions of MacOS?
>> 
>> Also does anyone understand how the ACX_PTHREAD macro can be modified to 
>> detect the need for the -pthread option?
>> 
>> For the details for why the patch is needed on MacOS 10.6, see the link 
>> below.
>> 
>>      
>> https://www-old.cae.wisc.edu/pipermail/bug-octave/2010-January/010241.html
>> 
>> Ben
> 
> The change is 10.6 is in /usr/include/architecture/i386/math.h:
> 
> /*    lgamma and lgammaf are not thread-safe.  The thread-safe variants
> *     lgamma_r and lgammaf_r are available on OS X 10.6 and later.
> *
> *     To use the thread-safe variants, you must define the _REENTRANT symbol.
> */
> 
> The -pthread compiler option causes gcc to define the required _REENTRANT 
> symbol. I have not tested whether lgamma_r and lgammaf_r depend on any 
> specific threading libraries, but if they do the -pthread option should cause 
> them to be linked in.
> 
> OSX before 10.6 does not have the _r variants, but it seems to me that having 
> the -pthread option set should not break anything. Haven't tested it, though.
> 
>  Jarno

Ok! If the decision is not to detect (by a functional test) that -pthread is 
needed, the case statement should be modified to reflect 10.6 only.

The Darwin version numbers for the various MacOS versions are listed at the 
link below.

        http://en.wikipedia.org/wiki/Darwin_(operating_system)

Would the case statement function properly if it were written as ...

case "${host_cpu}-${host_os}" in
        *solaris*)

        # On Solaris (at least, for some versions), libc contains stubbed
        # (non-functional) versions of the pthreads routines, so link-based
        # tests will erroneously succeed.  (We need to link with -pthreads/-mt/
        # -lpthread.)  (The stubs are missing pthread_cleanup_push, or rather
        # a function called by this macro, so we could check for that, but
        # who knows whether they'll stub that too in a future libc.)  So,
        # we'll just look for -pthreads and -lpthread first:

        acx_pthread_flags="-pthreads pthread -mt -pthread $acx_pthread_flags"
        ;;

        *-darwin10*)

        # On Darwin 10.x (MacOS 10.x), lgamma and lgammaf are not thread-safe. 
        # The thread-safe variants lgamma_r and lgammaf_r are available on
        # OS X 10.6 and later. To use the thread-safe variants, you must define 
the
        # _REENTRANT symbol. The -pthread compiler option causes gcc to define
        # the required _REENTRANT symbol.

        acx_pthread_flags="-pthread"
        ;;
esac

I'm not familiar with the tests done in these macros, but looking at 
lo-specfun.cc:rc_lgamma, It doesn't look to me like creating a test for this is 
straight forward (if I understand correctly a test would depend upon 
HAVE_LGAMMA_R being defined true).

Thoughts anyone? Is there an approach that can detect this problem, or should 
the case statement be modified as done above?

Ben