From @lizmat

$ 6 'my $x; multi sub foo($n where True) { temp $x; }; foo(True) for ^22’
(no pb)

$ 6 'my $x; multi sub foo($n where True) { temp $x; }; foo(True) for ^23'
Segmentation fault​: 11

$ 6 MVM_SPESH_DISABLE=1 6 'my $x; multi sub foo($n where True) { temp $x; }; foo(True) for ^23’
(no pb)

This is basically what test 5 in t/spec/integration/weird-errors.t is about.

From @jnthn

On Thu Apr 16 12​:27​:25 2015, elizabeth wrote​:

$ 6 'my $x; multi sub foo($n where True) { temp $x; }; foo(True) for
^22’
(no pb)

$ 6 'my $x; multi sub foo($n where True) { temp $x; }; foo(True) for
^23'
Segmentation fault​: 11

$ 6 MVM_SPESH_DISABLE=1 6 'my $x; multi sub foo($n where True) { temp
$x; }; foo(True) for ^23’
(no pb)

This is basically what test 5 in t/spec/integration/weird-errors.t is
about.

Some further analysis​: If you run it for longer without spesh it explodes anyway​:

jnthn@​(none)​:~/dev/rakudo$ MVM_SPESH_DISABLE=1 ./perl6-m -e 'my $x; multi sub foo($n where True) { temp $x; }; foo(True) for ^150;'
jnthn@​(none)​:~/dev/rakudo$ MVM_SPESH_DISABLE=1 ./perl6-m -e 'my $x; multi sub foo($n where True) { temp $x; }; foo(True) for ^5550;'
Segmentation fault

So not a dynamic optimizer bug after all.

The RT System itself - Status changed from 'new' to 'open'

From @jnthn

On Wed Jun 10 02​:55​:11 2015, jnthn@​jnthn.net wrote​:

Some further analysis​: If you run it for longer without spesh it
explodes anyway​:

jnthn@​(none)​:~/dev/rakudo$ MVM_SPESH_DISABLE=1 ./perl6-m -e 'my $x;
multi sub foo($n where True) { temp $x; }; foo(True) for ^150;'
jnthn@​(none)​:~/dev/rakudo$ MVM_SPESH_DISABLE=1 ./perl6-m -e 'my $x;
multi sub foo($n where True) { temp $x; }; foo(True) for ^5550;'
Segmentation fault

So not a dynamic optimizer bug after all.

Furthermore, you could actually get Valgrind to whine on a single iteration. Turned out to be a bad interaction between bind checks for the where and exit handlers. Fixed, and test in integration/weird-errors.t unfudged.

@jnthn - Status changed from 'open' to 'resolved'