[Pkg-hhvm-team] Bug#824573: Bug#824573: hhvm: Enable hhvm to be built on ppc64el

[Faidon Liambotis]

severity 824573 wishlist
thanks

Hi Breno,

On Tue, May 17, 2016 at 01:12:12PM -0400, Breno Leitao wrote:
> HHVM is being ported to ppc64el arch[1], and at the moment, 100% of the 1041
> test runs fine on the platform.
> 
> HHVM, as current version 3.12, is able to run in interpreted mode (JIT
> disabled)[2], so, I think we should go ahead and enable it to be built on ppc64el.
> I will working to have it enabled completely in the ppc64el platform.
> 
> The following patch just enable the built on ppc64el.

Thanks for your bug report and patch, appreciated -- and very sorry for
the late response.

It's good to know that HHVM works on ppc64el these days -- I've seen
work happening on the upstream git as well, so that's definitely great
to hear! :) It looks like upstream is working on arm64 as well, so that
may be an option for us as well.

Unfortunately, the package will currently FTBFS if I apply your patch:
we build-depend on ocaml-native-compilers and that's only available for
ppc64el in experimental at the moment (ppc64el was added in 4.03.0-2).

Regardless, I tried building the newly-uploaded (but generally old as an
upstream release) HHVM 3.12.11 package as-is in an experimental chroot
on the ppc64el porterbox (plummer.debian.org) and while it built fine,
the resulting binary immediately segfaults, even with no arguments
and/or with -m interp.

The backtrace shows:

Program received signal SIGSEGV, Segmentation fault.
(gdb) bt full
#0  HPHP::jit::(anonymous namespace)::memory_effects_impl (inst=...) at
./hphp/runtime/vm/jit/memory-effects.cpp:1066
No locals.
Backtrace stopped: Cannot access memory at address 0x400000000a90

Do you know if this is a known issue, fixed in newer upstream releases?
Any clues on what this may be?

Thanks,
Faidon