On Wed, 19 Oct 2016 07:54:05 -0700, [email protected] wrote:
> > [jdv@wieldy ~]$ perl6 test_grammar_n_threads.p6
> > *** Error in `/home/jdv/rakudo/install/bin/moar': double free or
> > corruption (fasttop): 0x000000000533c8a0 ***
> > ======= Backtrace: =========
> > /lib64/libc.so.6(+0x77d9e)[0x7f6585d5fd9e]
> > /lib64/libc.so.6(cfree+0x5b5)[0x7f6585d6b9f5]
> > //home/jdv/rakudo/install/lib/libmoar.so(MVM_interp_run+0x13b83)[0x7f658620d813]
> > //home/jdv/rakudo/install/lib/libmoar.so(+0x1774be)[0x7f658621c4be]
> > //home/jdv/rakudo/install/lib/libmoar.so(+0x23cf17)[0x7f65862e1f17]
> > /lib64/libpthread.so.0(+0x752a)[0x7f65857cb52a]
> > /lib64/libc.so.6(clone+0x6d)[0x7f6585de822d]
>
> This one I've only managed to tickle once out of a hundred runs, and
> not with exactly that backtrace either. Will have to look into it a
> bit further.
>
dogbert17++ managed to get some useful backtraces out of it, one of which
pointed directly to the problem so I could fix it. Now the SEGV doesn't seem to
happen even in hundreds of runs (must have reached into the thousands of runs).
Added a test of a golf of the issue to S17-promise/start.t.
