Re: [MacPorts] #9053: BUG gmp: crash in config.guess due to mfpvr instruction
#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 mcalhoun@…): Replying to [comment:15 vinc17@…]:
FYI, I tried trapping SIGILL in the past, but this didn't fix the problem (I suppose that the kernel doesn't have a way to know that SIGILL is trapped...).
Now, I'm wondering... What MacPorts is supposed to do concerning binaries? Produce binaries that can only run on the local machine? But in such a case, this is a bit contrary to universal support. Or produce binaries that can run on any declared architecture? For instance, any PowerPC-based machine if PowerPC is declared. In the latter case, GMP's config.guess should be patched to avoid particular CPU detection (as a side effect, this would avoid the crash in config.guess). One could test to see whether binaries would be significantly slower.
I think it is fair to say that the primary focus of MacPorts is the building of binaries to run on the machine doing the building.[[BR]] Intel/PowerPC universal support depends almost entirely on the individual ports.[[BR]] In the case of gmp, full universal support has proven to be difficult, which is why the muniversal PortGroup is now used. configure scripts, as a general strategy, attempt to do things (compile code or run programs).[[BR]] Often the scripts expect the attempts to fail and behave accordingly.[[BR]] Correct me if I'm wrong, but the objection in this case is that the failure is registered in the OS as a crash instead of just a failure.[[BR]] I can see how this is an annoyance, but fixing it seems like allot of effort for little reward. I would still vote that this ticket be closed as "wontfix." -- Ticket URL: <http://trac.macports.org/ticket/9053#comment:17> MacPorts <http://www.macports.org/> Ports system for Mac OS
participants (1)
-
MacPorts