#9053: BUG gmp: crash in config.guess due to mfpvr instruction ----------------------------------------+----------------------------------- Reporter: vincent-opdarw@… | Owner: gwright@… Type: defect | Status: assigned Priority: Low | Milestone: Port Bugs Component: ports | Version: 1.0 Resolution: | Keywords: Port: gmp | ----------------------------------------+----------------------------------- Comment(by vinc17@…): Replying to [comment:20 mcalhoun@…]:

-force_cpusubtype_ALL is required on PPC architectures or else gmp won't work properly.

What is the cause? (I think it may be related to broken CPU subtype detection.)

ABI=32 is required or else gmp will try to build a 64-bit library.

Yes, this is what I'm saying: MacPorts fixes a bad choice from GMP. And I think it should do something similar concerning CPU subtype detection.

I do not see why -mpcu would necessarily be a problem or why disabling the CPU subtype detection would be a good thing.

Why? The gcc man page says: {{{ The subtype of the file created (like ppc7400 or ppc970 or i686) is determined by the flags that specify the ISA that GCC is targetting, like -mcpu or -march. The -force_cpusubtype_ALL option can be used to override this. }}} So, avoiding CPU subtype detection and -mcpu in the first place would avoid the use of -force_cpusubtype_ALL (which may be buggy here since it does not control the correctness of assembly code, which GMP uses).

Just to help me understand, are we still talking about the crash in config.guess?

Not directly, but a fix to the above problem can avoid the crash as a side effect. -- Ticket URL: <http://trac.macports.org/ticket/9053#comment:21> MacPorts <http://www.macports.org/> Ports system for Mac OS