cvs commit: fptools/libraries Makefile.common fptools/libraries/HUnit Makefile.nhc98 fptools/libraries/HaXml/src Makefile.nhc98 fptools/libraries/OpenGL Makefile.nhc98 fptools/libraries/base Makefile.nhc98 fptools/libraries/haskell-src Makefile.nhc98

[Sven Panne]

Malcolm Wallace wrote:
> Unfortunately, this change causes all my library builds to fail :-(
> 
>     nhc98 built with
> 	hbc		still seg-faults on Language.Haskell.Parser,
> 			  even with 170Mb
> 	ghc-5		exhausts 1Mb stack on Language.Haskell.Parser
> 			  and Text/XML/HaXml/OneOfN,
> 			also on the profiling builds of Text/Html,
> 			  Language/Haskell/Syntax, and Text/XML/HaXml/Parse.
> 	ghc-6		same as for ghc-5
> 	nhc98		exhausts 16Mb heap on profiling build of
> 			  Language/Haskell/Parser
> [...]

Strange, I tested this with hbc (my "new" version) and with GHC from CVS (using -O),
and both worked. Currently programs generated by nhc98 don't work on my box, e.g.:

    Trying to get tag from unevaluated node in TABLESWITCH at 08057b18!
    Node is:
     0805abb3 at 401790b4

But it's a mystery to me how PARANOID gets defined...

> Besides, I think it is better policy to set the heap/stack sizes
> individually for each library package [...]

It's OK with me, I just wanted to keep the version madness centralized.

Cheers,
    S.