Sorry, I forgot to mention. It's a PPC PowerBook G4 running Mac OS X 10.4.11. -bw On Jan 25, 2008, at 6:14 PM, Dale Johannesen wrote:> What target? > > On Jan 25, 2008, at 6:06 PM, Bill Wendling wrote: > >> Here are my test results for the 2.2 prerelease. >> >> Built objDir == srcDir, release, and with the pre-built LLVM-GCC >> binary. >> >> "make test" results: >> >> === Summary ==>> >> # of expected passes 2533 >> # of expected failures 6 >> >> >> -bw > > Looks good. > > <report.nightly.txt> >> _______________________________________________ > > Not so good, there are things failing that usually pass: lencod, > oggenc, a bunch of EH tests. > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
On Jan 25, 2008, at 9:35 PM, Bill Wendling wrote:> Sorry, I forgot to mention. It's a PPC PowerBook G4 running Mac OS X > 10.4.11. > > -bwOK, ppc32 EH used to work better than this (although I haven't tried it on 10.4). I'll have a look.> On Jan 25, 2008, at 6:14 PM, Dale Johannesen wrote: > >> What target? >> >> On Jan 25, 2008, at 6:06 PM, Bill Wendling wrote: >> >>> Here are my test results for the 2.2 prerelease. >>> >>> Built objDir == srcDir, release, and with the pre-built LLVM-GCC >>> binary. >>> >>> "make test" results: >>> >>> === Summary ==>>> >>> # of expected passes 2533 >>> # of expected failures 6 >>> >>> >>> -bw >> >> Looks good. >> >> <report.nightly.txt> >>> _______________________________________________ >> >> Not so good, there are things failing that usually pass: lencod, >> oggenc, a bunch of EH tests. >> >> _______________________________________________ >> LLVM Developers mailing list >> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu >> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
On Jan 25, 2008, at 9:43 PM, Dale Johannesen wrote:> > On Jan 25, 2008, at 9:35 PM, Bill Wendling wrote: > >> Sorry, I forgot to mention. It's a PPC PowerBook G4 running Mac OS X >> 10.4.11. >> >> -bw > > OK, ppc32 EH used to work better than this (although I haven't tried > it on 10.4). I'll have a look.I should clarify, what is the "pre-built LLVM-GCC binary"? Part of 2.2, and it's 4.2, right? I wouldn't expect 4.0 to work very well in this area.>> On Jan 25, 2008, at 6:14 PM, Dale Johannesen wrote: >> >>> What target? >>> >>> On Jan 25, 2008, at 6:06 PM, Bill Wendling wrote: >>> >>>> Here are my test results for the 2.2 prerelease. >>>> >>>> Built objDir == srcDir, release, and with the pre-built LLVM-GCC >>>> binary. >>>> >>>> "make test" results: >>>> >>>> === Summary ==>>>> >>>> # of expected passes 2533 >>>> # of expected failures 6 >>>> >>>> >>>> -bw >>> >>> Looks good. >>> >>> <report.nightly.txt> >>>> _______________________________________________ >>> >>> Not so good, there are things failing that usually pass: lencod, >>> oggenc, a bunch of EH tests. >>> >>> _______________________________________________ >>> LLVM Developers mailing list >>> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu >>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >> >> _______________________________________________ >> LLVM Developers mailing list >> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu >> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >
On Jan 25, 2008, at 9:43 PM, Dale Johannesen wrote:> > On Jan 25, 2008, at 9:35 PM, Bill Wendling wrote: > >> Sorry, I forgot to mention. It's a PPC PowerBook G4 running Mac OS X >> 10.4.11. >> >> -bw > > OK, ppc32 EH used to work better than this (although I haven't tried > it on 10.4). I'll have a look.I've confirmed that it still works on 10.5. I was less successful than you on 10.4: [dalejos-powerbook-g4-15:/Volumes/Data/tests] dalejo% /llvm-gcc4.2-2.2- ppc-darwin8.11.0/bin/llvm-g++ ctor1.C /usr/bin/ld: unknown flag: -macosx_version_min collect2: ld returned 1 exit status This is a ppc32 machine running 10.4.11, but does not have a recent tools release installed on it. What's the prerequisite here? I think the problem with EH is that you're getting the unwinding library that was built as part of the llvm-gcc build. That one will not work. We want to pick up that library from a gcc build (or it may work just to remove the one that's there, if we require some other tools be installed). It seems to use libgcc_s.10.[45].dylib on 10.5 and libgcc_s.dylib on 10.4. But it may be that 4.0 vs 4.2 is another variable, I'm not in a position to test all combinations. I've used llvm-gcc to compile as far as a .o, then used g++ to link, and that works. So I think this is the problem.