On 1/8/10 1:53 PM, Vincent Rivière wrote:
Mark often complains about the malloc() bug, what is it about ?Often when building a large software package I'll get an error like this: Alan Hourihane p�e v St 11. 03. 2009 v 21:47 +0000: > > make: ../mintlib/malloc.c:187: __free: Assertion `s == r' failed. > > > > I just had something similar happen to me on aranym.... > > sed: ../mintlib/malloc.c:212: __free: Assertion `s == (struct mem_chunk *) _heap_base' failed. things like this happen from time to time since ages (> 10 years). > Funny thing is, this has never happened to me on my Falcon. Even if it's random I think it sort of depends on the amount of RAM and I guess you got more memory in your ARAnyM setup than in your Falcon. PetrIt's always on line 187 of malloc.c for me. Frank Naumann said the problem is fixed, it must be fixed, you're simply using something that is linked to the old mintlib. So I rebuilt all packages, that make depends on and make still crashed with this error. I haven't reproduced this for a little while but I'm 100% sure there's still something somewhere causing this. Keith Scroggins might be able to comment if he's been building any large packages natively recently. It may be hard to find but I do believe something is there and it rears it's ugly head on long native compiles - the kind needed to rebuild sparemint for instance ;) Thanks, Mark |