is-prime test failure in rakudo-star-2013.02 on OS X 10.8.2 #3111
Hi. on the perl6 IRC channel, it was suggested i open a bug on this test failure, in is-prime.t :
On Tue Apr 23 13:48:44 2013, email@example.com wrote:
I have experienced the same bug on OS X 10.8.3.
Here is more info on the symptom. When I run the code below, it always
(Lines wrapped for clarity)
Note that at 172, *even* numbers start creeping in. Definitely not prime!
The fix is in commit 16fa719.
The problem was in src/vm/parrot/ops/nqp_bigint.ops .
My take/guess on what was happening:
On my system, C's (int)s are 32-bit, but Parrot's I-registers are
If the high half of the register happens to be 0, then everything works
If anyone thinks my analysis is in error, even after this ticket is
On Tue Apr 30 15:44:27 2013, util wrote:
Yes, fixes the problem for me on FreeBSD and OS X.
I think your analysis is spot on. We had quite a few problems like
Any cast to a pointer of a different size (or potentially a different
It's also strange that the bug had all the symptoms of a GC bug