Re: pg11.1 jit segv

Hi,

On 2018-11-15 17:03:35 -0600, Justin Pryzby wrote:
> On Thu, Nov 15, 2018 at 02:47:55PM -0800, Andres Freund wrote:
> > > (gdb) bt
> > > #0 0x00007f08127e814e in ?? ()
> > > #1 0x0000000000000000 in ?? ()
> >
> > Could you enable jit_debugging_support and reproduce? That should give
> > a more useful backtrace.

> Core was generated by `postgres: pryzbyj ts [local] EXPLAIN '.
> Program terminated with signal 11, Segmentation fault.
> #0 0x00007f819e227cb9 in __memcpy_ssse3_back () from /lib64/libc.so.6
> Missing separate debuginfos, use: debuginfo-install audit-libs-2.8.1-3.el7_5.1.x86_64 bzip2-libs-1.0.6-13.el7.x86_64 cyrus-sasl-lib-2.1.26-23.el7.x86_64 elfutils-libelf-0.170-4.el7.x86_64 elfutils-libs-0.170-4.el7.x86_64 glib4
> (gdb) bt
> #0 0x00007f819e227cb9 in __memcpy_ssse3_back () from /lib64/libc.so.6
> ...

Huh, that's the same crash? Because I don't see any evalexpr functions
in the stack, and without those the above bt should have worked...

> > It'd also be interesting if toggling jit_inline_above_cost=-1,
> > jit_optimize_above_cost=-1 and jit_tuple_deforming=false, each set
> > individually, make a difference.
>
> Crashes with jit_tuple_deforming=true (others off).

> Doesn't crash with other combinations:
> Options: Inlining false, Optimization true, Expressions true, Deforming false
> Options: Inlining true, Optimization false, Expressions true, Deforming false

Oh, interesting. That helps. Could you perhaps show an EXPLAIN VERBOSE
of the query plan?

> > Any way for me to reproduce this locally? I.e. can you narrow this down
> > to a reproducible test case?
>
> Will work on it.

Thanks!

Greetings,

Andres Freund