[Gc] Crashs and Deadlocks in the GC on windows-xp
hans.boehm at hp.com
Tue Mar 14 13:21:25 PST 2006
Any chance we could include your stress test in the GC distribution?
GC7 tries to clean up a lot of this, and more thorough tests would be
Can you get stack traces corresponding to the deadlocks? Those don't
sound familiar at all.
The Dll magic has been a source of problems in the past. Gc7 uses the
other approach by default.
> -----Original Message-----
> From: gc-bounces at napali.hpl.hp.com
> [mailto:gc-bounces at napali.hpl.hp.com] On Behalf Of Lothar Scholz
> Sent: Monday, March 13, 2006 11:53 PM
> To: gc at napali.hpl.hp.com
> Subject: [Gc] Crashs and Deadlocks in the GC on windows-xp
> I just spend this morning to write a multithreading, garbage
> collector stress test program and found two serious problems.
> First, after a few seconds of thread creation and destruction
> i always run into an error that GC_stop_world calls
> SuspendThread on an invalid handle. The windows
> GetLastError() code is 6. This happens with 6.6 and 6.7 and
> is very repeatable for me. First i tried to catch this
> situation and simply close the handle in the thread table
> when such a situation came up but this did not help very much.
> Also the comment above in the function told me that i'm not
> the first one who had this problem.
> So i solved it by using GC_CreateThread and linking to a
> static compiled library instead of using the DLL magic. I
> never had invalid handles anymore.
> But there is still one problem left: It looks like i run into
> a deadlocks after some time. But it maybe something different
> as my dual processor is having a 50% CPU load. Is there any
> spinlock blocking that might cause this behaviour ? Is
> anybody out there who also saw this in the past ?
> At the moment, i'm 75% convinced that it is a GC bug.
> Best regards,
> Lothar Scholz mailto:scholz at scriptolutions.com
> Gc mailing list
> Gc at linux.hpl.hp.com
More information about the Gc