[MLton-user] Finalizable bug

[Matthew Fluet]

On Sun, 18 Nov 2007, Florian Weimer wrote:
> * Sean McLaughlin:
>>   I found a bug using the Finalizer module.  The tar of the files is
>> attached.
>> On multiple machines it gave segfaults.
>
> The references to MLton.Pointer are not necessary, so it's possible to
> minimize the test case even further.  The crash occurs in the
> mark-compact collector during the first mark-compact collection of this
> program.

Thanks for the diagnosis.  If the bug is only in the mark-compact 
collector (as opposed to in the mutator, and it only manifests itself in 
the m-c collector), then you may be able to work around it with various 
runtime options to avoid the m-c collector.  It's been on my todo to 
augment the GC with some options to enable/disable/force the 
copying/mark-comact/generational collectors.  This would make it easier to 
avoid bugs in particular collectors and also make it easier to stress test 
collectors.