On 10 September 2005 13:31, Gaal Yahas wrote:
> In compiling pugs, usually at Pugs.AST.Internals or Pugs.Parser (large
> files), I get GHC crashes. The error is typically "strange closure
> type 37108" but other numbers also appear (0 sometimes).
>
> I tried this with:
>
> An oldish GHC 6.5 snapshot
> GHC 6.4 release
> Latest GHC 6.4.1
>
> My system is an x86 with plenty of RAM running Windows XP. I'm setting
> -M450m for the compilation: up till recently, as 350m was enough so
> this ought to be sufficient.
>
> Is there anything I can do to give you more information? Or to work
> around this problem? My own pugs development is stalled because of
> this :-(
I just built Pugs 6.2.9 on Windows with -M450m, using a recent 6.4.1
snapshot, and it went through without errors.
Did this start to happen recently for you? Does it happen more often
with certain heap size settings? With certain GHC versions? If you
have plenty of RAM, you should be able to give a really big -M setting
(say -M1.5G) - does that make it more stable?
I'm trying to establish whether this is a bug in the compacting garbage
collector. Compacting collection gets turned on when the residency
approaches the maximum heap size (the -M setting), so the compacting
collector gets a lot less testing than the normal copying collector.
Also, it is known that these random GC errors are sometimes a symptom of
hardware failure. Your system might otherwise seem stable, but GHC
gives your memory a good work-out and can coax hardware problems out of
the woodwork. If you can reproduce the errors on another machine that
would help to eliminate this possibility.
Cheers,
Simon
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
