site_archiver@lists.apple.com Delivered-To: darwin-dev@lists.apple.com
I think I have solved the problem. I thought the use of only -fno-common in gmp to compile shared libs looked suspect. Sure enough if I compile gmp 4.2 with...
./configure --enable-cxx CC='gcc -fPIC' CXX='g++ -fPIC'
Jack, That's ... interesting. PIC-ness affects the correctness of the code for the failing test? (Were you able to reproduce the t-hamdist fail?) Something seems amiss. Is your 'gcc' FSF's 4.1.0, or one of Apple's? (Incidentally, you can pass CFLAGS=-fPIC CXXFLAGS=-fPIC as configure params, that guarantees that your flags come last, and not overridden by anything else the maintainer defined.)
...the resulting build fully passes make check on the G5. I don't believe that -fno-common provides PIC code in the same fashion as -fPIC. Googling those flags show a lot of code using both -fPIC and -fno-common in the compiles.
Anyhow, thanks for the update, good to hear that someone's got it working. Now if we can trust the coverage of 'make check', then we're good to go. :) David _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev@lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.appl... This email sent to site_archiver@lists.apple.com