[Pkg-clamav-devel] Uploading 0.98.7+dfsg-4

[Andreas Cadhalpun]

Hi Sebastian,

On 27.11.2015 13:19, Sebastian Andrzej Siewior wrote:
> So let me add some of the informations I gathered:
> - llvm-3.5 will leave unstable before the release. I've been told today
>   the chance are exactly zero that it will remain.

I thought so, too.

> - I stumbled over https://llvm.org/bugs/show_bug.cgi?id=9605 saying and
>   I quote "THe MCJIT is working for PPC64 only.".
>   I'm going to ping him if something changed or will change but it
>   matches our (or better your) observation.

Looking at the crashing code [1] seems to confirm this:
void RuntimeDyldELF::resolveRelocation(const SectionEntry &Section,
                                       uint64_t Offset, uint64_t Value,
                                       uint32_t Type, int64_t Addend,
                                       uint64_t SymOffset) {
  switch (Arch) {
  case Triple::x86_64:
    resolveX86_64Relocation(Section, Offset, Value, Type, Addend, SymOffset);
    break;
  case Triple::x86:
    resolveX86Relocation(Section, Offset, (uint32_t)(Value & 0xffffffffL), Type,
                         (uint32_t)(Addend & 0xffffffffL));
    break;
  case Triple::aarch64:
  case Triple::aarch64_be:
    resolveAArch64Relocation(Section, Offset, Value, Type, Addend);
    break;
  case Triple::arm: // Fall through.
  case Triple::armeb:
  case Triple::thumb:
  case Triple::thumbeb:
    resolveARMRelocation(Section, Offset, (uint32_t)(Value & 0xffffffffL), Type,
                         (uint32_t)(Addend & 0xffffffffL));
    break;
  case Triple::mips: // Fall through.
  case Triple::mipsel:
    resolveMIPSRelocation(Section, Offset, (uint32_t)(Value & 0xffffffffL),
                          Type, (uint32_t)(Addend & 0xffffffffL));
    break;
  case Triple::ppc64: // Fall through.
  case Triple::ppc64le:
    resolvePPC64Relocation(Section, Offset, Value, Type, Addend);
    break;
  case Triple::systemz:
    resolveSystemZRelocation(Section, Offset, Value, Type, Addend);
    break;
  default:
    llvm_unreachable("Unsupported CPU type!");
  }
}

As you can see, you can't see 'Triple::ppc' there.
This is unchanged for llvm-toolchain-3.7, but llvm-toolchain-snapshot contains [2]:
  case Triple::ppc:
    resolvePPC32Relocation(Section, Offset, Value, Type, Addend);
    break;

So I guess that this should be fixed in 3.8, but...

> - I was going to check your testcase on llvm-3.7 and snapshots but llvm
>   failed to build on powerpc. So I can't.

... it fails to build on powerpc.

>>> What exactly do we lose if we disable llvm?
>>
>> I guess the JIT engine is supposed to be faster than the interpreter.
> 
> So no lack of functionality. Does not sound that bad compared to broken
> clambc in unstable like we have it right now.

Yes. We should just disable llvm on powerpc, at least until llvm 3.8.

Best regards,
Andreas

1: https://sources.debian.net/src/llvm-toolchain-3.6/1:3.6.2-1/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp/?hl=861#L861
2: https://sources.debian.net/src/llvm-toolchain-snapshot/1:3.8~svn250696-1/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp/?hl=1098#L1098