On Sat, Jun 23, 2007 at 04:13:02PM +1000, Donald Bruce Stewart wrote:
I did extensive profiling: after reaching a given level, memory usage seems to stop increasing. But I need some feed back to understand if
That's interesting. Try setting a low bound on the heap size.
I'm making some experiment with that, with varied result: under 1 Mega the program exited because of the limit, now I've been running with +RTS -M1M -RTS and it keeps on running with increasing memory usage (now 8124 (vir) 47808 (res) 2244 (shr)). I've also tried a very long profiling (I run it many hours till it ended up with 9 Mega or resident memory, according to top), and these are the reports: http://gorgias.mine.nu/xmobar/xmo.hp http://gorgias.mine.nu/xmobar/xmo.prof http://gorgias.mine.nu/xmobar/xmo.ps This last one makes me wonder: it seems to me that the report is wrong. According to heap profiling there are no memory leaks, am I right? Am I reading the report correctly? I remember I've read something about heap profiling not showing memory leaks related with FFI, there were a discussion in the haskell-cafe mailing list. I'll keep investigating. If you guys have some clue that would be nice. BTW, gorgias is up and running almost (I was able to get a backup machine, an old, very old one). You cen get the darcs repository here: http://gorgias.mine.nu/repos/xmobar/ The server may be down during the next few days (it must be changed) so in case of problems please try coming back later. Ciao Andrea