Re: [libunwind] building recent bk libunwind on x86-64

[David Mosberger]

>>>>> On Fri, 4 Mar 2005 15:09:25 -0600, Nathan Froyd <address@hidden> said:

  Nathan> I compiled with a recent GCC snapshot so that
  Nathan> (theoretically) the necessary unwind info would be
  Nathan> generated.

I haven't checked very recent GCCs but one potential problem is that
so far GCC on x86/x86-64 (probably all DWARF-based targets) hasn't
emitted any unwind info for the epilogue, meaning that you couldn't
unwind while inside the epilogue.

  Nathan> I found that I had to apply to the attached patch to make
  Nathan> everything compile.

Thanks, there were indeed a couple of typos that had snuck in.  I
fixed those now in the bk repository.

  Nathan> However, even with the latest and greatest GCC, several
  Nathan> tests which the README indicates should not fail do, in
  Nathan> fact, fail:

  Nathan> {L,G}test-exc {L,G}test-resume-sig run-ptrace-mapper

run-ptrace-mapper also fails for me.
test-exc and test-resume-sig work for me with GCC v3.4.3.
test-exc is a bit sensitive to compilers so perhaps that's why you're
seeing it fail.  Not sure about test-resume-sig.

  Nathan> Is there a particular requirement for
  Nathan> libc/kernel/gcc/etc. for the information in the README to be
  Nathan> accurate?

The compiler that's installed on the system I have access to says:

  gcc version 3.4.2 20041017 (Red Hat 3.4.2-6.fc3)

FYI: I currently have two other higher-priority items I need to take
     care of before I can think about working on finishing x86{,-64}
     support.  Of course, patches are very welcome in the meantime.

Thanks,

        --david