Patrick R. Michaud via RT wrote:
On Wed, Aug 08, 2007 at 09:12:20PM -0600, Lloyd Miller wrote:
Patrick R. Michaud via RT wrote:
On Tue Aug 07 20:00:03 2007, millerlf <!-- x --> at telus.net wrote:
when I ran "make test" I get 1 failure. Looks like this ...
not ok 16 - examples/shootout/regexdna.pir
# Failed test (t/examples/shootout.t at line 103)
[...]
# Segmentation fault (core dumped)
[...]
I am running kubutu fiesty x86_64 on a pentium-D 925 & 1gb ram
The general consensus is that this recurring and sporadic problem is
related to an ongoing GC bug. I've merged #44495 into RT#43481, which
has some other discussion of the same issue.
...
What happens if you run the test directly from parrot, both with and
without the -G flag?
$ parrot -G examples/shootout/regexdna.pir
With the -G switch it seemed to run normally without the seg fault and
produced the proper output
Yes. The -G switch disables the garbage collector and causes
the program to run properly (which is why we think the segfault
is garbage collection related).
Also, very important -- what svn revision of Parrot is showing the
segfault for you?
just the download tar file version 0.4.14
Okay. We _think_ this particular bug may have been fixed
(or otherwise circumvented) in the current svn HEAD for Parrot.
It's hard to know for sure when we've fixed the bug, because
garbage collection errors are notoriously difficult to track --
often the only way you know it's fixed is that it stops appearing.
At any rate, the bug existed for me as well in 0.4.14, but
appears to be working as of r20576 in the HEAD branch.
(Like you, I'm also running kubuntu feisty on an x86_64 system,
gcc 4.1.2.)
OK, I tried svn r20585 and this fail is gone from make test.
--
VE6LFM Terminal insomniac
