[Openmcl-devel] find reason for crashes and prevent mcl from going into kernel debugger?
stoye at stoye.com
Thu Jan 19 20:58:02 PST 2006
Am 13.01.2006 um 00:24 schrieb Gary Byers:
> a) there's a fencepost in the code that calculates the required
> size of the bitmap
> b) there's an object whose size is incorrect (possibly as a
> result of some sort of memory corruption near the end of
> the heap, and this causes the GC to try to set bits that
> aren't allocated and shouldn't be.
b) is the winner:
the coder "optimized" by declaring an array with fill-pointer to be
of type simple-array :-(
("there really is no reason not to use an simple-array, it just has
to be an simple-array....")
testing with $gc-integrity-check-bit set told me about corrupt objects,
the way to find the bug was straightforward - assuming that mcl's
memory handling isn't buggy - messing around with pointers or wrong
declarations could be the only reasons.
sorry to pollute the list with this dull unnecessary traffic
just writing this so nobody worries about errors in the core of openmcl
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Openmcl-devel