cvs commit: fptools/ghc/rts Makefile

[Marcin 'Qrczak' Kowalczyk]

Mon, 29 Jan 2001 04:00:20 -0800, Simon Marlow <simonmar@microsoft.com> pisze:

> 	- stage 1 compiler: set GhcRtsHcOpts += -optc-DGHCI.
> 	- stage 2 compiler: set GhcWithInterpreter = YES, set
> 	  GhcHcOpts += -ldl.

Now I understand the reason of my confusion. It's not a single fixed
point! The resulting compiler is not supposed to be able to compile
itself with all whistles turned on. Different options are needed not
just for bootstrapping during the transition from pre-GHCi-enabled
compilers to GHCi-enabled ones, but each time a GHCi-enabled compiler
is to be compiled.

I hope this is a temporary solution and in future there will be a
single good configuration able to compile itself.


PS. The compiler loops eating memory trying to compile PrelCError
and I don't know why. Reducing the list of
    | errno == eWHATEVER = (WhateverError, "whatever description")
below a few elements lets it pass, removing -O too, so perhaps some
code generation bug for the implementation of large case constructs.

-- 
 __("<  Marcin Kowalczyk * qrczak@knm.org.pl http://qrczak.ids.net.pl/
 \__/
  ^^                      SYGNATURA ZASTĘPCZA
QRCZAK