Possible hash synchronization bug exposed by AWT?

Mon Jul 26 19:01:00 GMT 2004

I have come up with a test case that reproduces the bug consistently on 
multi-threading and multiprocessor systems (both x86 and PPC) - I've 
posted it under the bug report at http://gcc.gnu.org/PR16662
It does seem that the bug only occurs on MP/MT systems.
Jacob Gladish wrote:
>Oh... I forgot. This code is running on both dual p3 and dual p4
>machines. The p4's have hyperthreading disabled.
>>>-----Original Message-----
>From: java-owner@gcc.gnu.org [mailto:java-owner@gcc.gnu.org] On Behalf
>Of Jacob Gladish
>Sent: Monday, July 26, 2004 9:58 AM
>To: 'Bryce McKinlay'
>Cc: hans.boehm@hp.com; java@gcc.gnu.org
>Subject: RE: Possible hash synchronization bug exposed by AWT?
>>>I have been unable to repro the IllegalMonitorStateException. The test
>case I sent out that that revealed the deadlock in the hash sync was
>what I used to do the timing experiment. I was slightly modified to
>track timing. Now that I think about it, the runtime was built with a
>hash sync table size of 1. This may have been the cause of the slowness.
>>Yes, quite possibly.
Regards
Bryce