Hi, Markus.
Thanks for the patch. It's not obvious to me why the original code
would leak, so thanks twice. I've applied it as revision 2205.
I've counted ok() calls, and t/10_units/04_drivers/01_sysrw.t test 13
seems to be "buffer exhausted (got N messages wanted 0)" at line
132. I don't think I can debug that one by remote control.
--
Rocco Caputo - [EMAIL PROTECTED]
On Jul 23, 2007, at 11:06, Markus Jansen wrote:
Hi Rocco and others,
thanks for all the effort you put in POE.
On POE-0.9999 (released today), I found one defect causing lots of
tests
to fail on my Perl-5.8.8 Solaris 2.6 box, with the Perl process trying
to allocate as much memory as possible.
The problem (root cause found at POE::Wheel::SocketFactory Socket6
code)
is a painful one; please find the context diff of my correction
(tested) below.
In addition to the Socket6 non-existence problem described above,
test t/10_units/04_drivers/01_sysrw.t (test 13) seems to have a
race condition
(passes ca. 50-60 percent of times it gets invoked).
Unfortunately, I don't have time to invest the root cause.
Finest regards,
Markus Jansen
Test descripton / context diff
==============================
Since Socket6 is missing on Solaris 2.6, the 'require Socket6' fails
(obviously only the first time!). As a consequence, the Perl process
grows unless the out of memory bound is reached.
N.B. the patched code is tested but not too nice ...
[patch and signature pruned]
