Re: Cores almost on demand in patcompile()

[Peter Stephenson <p.stephenson@xxxxxxxxxxx>]

On Wed, 19 Oct 2016 13:14:47 +0200
Sebastian Gniazdowski <sgniazdowski@xxxxxxxxx> wrote:
> I also did step by step walk through patcompile(). Unluckily it wasn't
> a core-triggering dump (next ended with 0x0). I attach screenshots
> with some crucial places.

Almost ceratinly you need to hit misbehaviour (or be about to hit
misbehaviour :-() before we're going to get anywhere.  It's not
fundamentally broken in normal cases.

If you can provoke it frequently enough, valgrind is the most likely
thing to help.  In your case, you might think a few runs under valgrind
should produce some interesting out.

I tried turning on the additional valgrind integration for heaps, but
that was doing weird things, in fact rather dishearteningly two sits of
different weird things on two different machines (one rather out of
date).

pws