> The problem is that the 4.01 hsc seg-faults from time to time, and the
> exact circumstances are hard to reproduce. Good candidates
> are scanners
> generated by lx (but not all of them). This is independent of the
> Linux system used (SuSe 5.3 = libc5, SuSE 6.0 = libc6, with the
> correspondingly compiled 4.01s), so I suspect some GC problems. Hmm,
> once I had an error like "GC: strange closure type", but the programs
> were quite large and hard to boil down. This was one of the
> reasons I'm
> trying to compile the latest-and-greatest ghc...
I fixed at least one problem of this kind in 4.01's garbage collector (it
was the one reported by Sergey a while back), so updating to the latest GHC
will likely help here.
Cheers,
Simon
--
Simon Marlow
Microsoft Research Ltd., Cambridge, U.K.
