[Gc] Re: win32 infinite loop in marking

Boehm, Hans hans.boehm at hp.com
Thu Feb 22 11:33:58 PST 2007


Thanks.  That's interesting.

Perhaps this should generate a warning if you catch more than N of these
per GC cycle.  I don't think we've ever legitimately seen more than 1,
so reporting a case in which you see 1000 seems benign.

I would also recommend at least going through the CVS README.changes
file to make sure that none of the more recent bug fixes affect you.
7.0alpha5 is moderately old at this stage, and I think there were some
fairly serious bugs.

Hans


> -----Original Message-----
> From: gc-bounces at napali.hpl.hp.com 
> [mailto:gc-bounces at napali.hpl.hp.com] On Behalf Of Peter Ross
> Sent: Thursday, February 22, 2007 7:59 AM
> To: gc at napali.hpl.hp.com
> Subject: [Gc] Re: win32 infinite loop in marking
> 
> On 2/21/07, Peter Ross <pro at missioncriticalit.com> wrote:
> > Eventually after a while the gc goes into an infinite loop 
> in the mark routines.
> >
> I found the problem.
> 
> GC_mark_some wraps GC_mark_some_inner with an exception 
> handler which catches segmentation faults, and then 
> effectively reruns marking.
> 
> My problem was the GC_malloc_heap_l structure had been 
> corrupted by some library code and hence a segmentation fault 
> was always being thrown when this structure was being 
> examined hence the infinite loop.
> 
> Lots of fun debugging that one.
> 
> Hope no-one wasted too much time thinking about it.
> 
> Regards,
> Peter
> --
> Peter Ross - Mission Critical Australia
> AU: +61 424 651 834
> BE: +32 485 482 559
> _______________________________________________
> Gc mailing list
> Gc at linux.hpl.hp.com
> http://www.hpl.hp.com/hosted/linux/mail-archives/gc/
> 



More information about the Gc mailing list