Re: semi-random crashes on some macppc hardware
"John D. Baker" <[email protected]> Mon, 16 May 2016 12:38:33 -0500 (CDT)
| Newsgroups | gmane.os.netbsd.ports.powerpc |
|---|---|
| Message-ID | <[email protected]> |
I have for quite some time experienced similar crashes but it was always in C/C++ compiler when building packages, particularly those with a long list of dependencies to work through. The system in question is a G4 PowerMac (AGP Graphics), commonly referred to, IIUC, as "Sawtooth". It boots and runs from local disk. Using GCC 4.8.5 and prior, one would be working through dependencies until GCC would fail with "internal compiler error: Segmentation fault". If one simply restarted the build, it would likely complete the file it was working on previously, but then segfault while processing the next source file or perhaps one following. The only workaround I ever found was to reboot the machine and pick up where I left off. Then it would either complete the list of dependencies and possibly the target package, or at least run for a reasonably long time before having to repeat the process. With the switch to GCC 5.3.0, the problem occurs more frequently, but the reboot workaround seems not especially helpful or necessary. For some reason, I conciously chose not to reboot the machine after the compiler segfaulted, but did do "make clean clean-depends" before restarting the package build. It would certainly get much farther along between incidents (not just the next source file as before), but still more frequent than a freshly-rebooted machine running GCC 4.x would. I applied the patch posted previously and while it does not eliminate the problem, it seems to have reduced the frequency to at least the prior level. I have not rebooted the machine following the segfaults, but continue to run "make clean clean-depends" before restarting a package build. I've been hesitant to consider hardware problems as back when it last had MacOS X (10.4.11) installed, it worked just fine. Perhaps I should stick another disk in it and verify that it still does. The other viable system is a Titanium PowerBook G4 (667MHz) which runs NetBSD strictly via NFS (MacOS X 10.4.11 on internal disk). It has never exhibited the compiler segfault problem except when trying to perform the final link step of "firefox". I have not yet tried it with a GCC 5.3.0-built system or the native GCC 5.3.0. -- |/"\ John D. Baker, KN5UKS NetBSD Darwin/MacOS X |\ / jdbaker[snail]mylinuxisp[flyspeck]com OpenBSD FreeBSD | X No HTML/proprietary data in email. BSD just sits there and works! |/ \ GPGkeyID: D703 4A7E 479F 63F8 D3F4 BD99 9572 8F23 E4AD 1645