Re: gs_group_1 crashing on 13beta2/s390x

Hi,

On 2020-09-28 14:22:01 +0200, Christoph Berg wrote:
> Re: Andres Freund
> > > > Ok, but given that Debian is currently targeting 22 architectures, I doubt the PostgreSQL buildfarm covers all of them with the extra JIT option, so I should probably make sure to do that here when running tests.
> > > 
> > > +1.  I rather doubt our farm is running this type of test on anything
> > > but x86_64.
> > 
> > There's quite a few arm animals and at least one mips animal that do
> > some minimal coverage of JITing (i.e. the queries that are actually
> > somewhat expensive). I pinged two owners asking whether one of the arm
> > animals could be changed to force JITing.
> 
> I pushed a change that should enable LLVM-10-JIT-testing everywhere [*]
> and (admittedly to my surprise) all other architectures passed just
> fine:
> 
> https://buildd.debian.org/status/logs.php?pkg=postgresql-13&ver=13.0-2

Thanks!


> For the record, the architectures with llvm disabled are these:
> 
> clang-10 [!alpha !hppa !hurd-i386 !ia64 !kfreebsd-amd64 !kfreebsd-i386 !m68k !powerpc !riscv64 !s390x !sh4 !sparc64 !x32],

!powerpc doesn't exclude ppc64, I assume?


> After the tests I realized that LLVM 11 is also already packaged, but
> s390x still segfaults with that version.
> 
> Christoph
> 
> [*] apparently pgbench --temp-config=/no/such/file is not an error,
> which makes verifying this change a bit harder

pgbench? I assume you mean pg_regress?

FWIW, an easy way to enable JIT for just about all tests, including tap
tests, is to set
PGOPTIONS='-c jit=1 -c jit_above_cost=0 ...'
in the environment before starting the tests.


Can a non-debian dev get access to a s390x? It'd be nice to isolate this
enough to report a bug to LLVM - and that's probably a lot easier for me
than you... My guess would be that some relocation processing or such is
off.

Greetings,

Andres Freund