Peter Thiemann complained about this bug:
> ghc-4.00 -c -cpp -syslib misc -fglasgow-exts -H27M
> -DAFMPATH=\\\"/usr/local/tex/Adobe\\\" -DRGBPATH=\\\"/usr/lib/X11\\\"
> CommandLine.hs -o CommandLine.o
> /usr/libexec/ld.so: warning: /usr/lib/libc.so.3.0: minor
> version 0 older
> than expected 1, using it anyway
> ghc-4.00: module version changed to 1; reason: no old .hi file
> /tmp/ghc13164.hc:4106: `__ap_31_upd_info' undeclared (first
> use in this
> function)
> /tmp/ghc13164.hc:4106: (Each undeclared identifier is
> reported only once
>
> /tmp/ghc13164.hc:4106: for each function it appears in.)
> gmake: *** [CommandLine.o] Error 1
I'm afraid that's my fault. I left a safety net commented out because I was
interested in finding out whether it happened very often or not. I guess my
question has now been answered... patch follows.
RCS file: /local/fp/src/cvsroot/fptools/ghc/compiler/codeGen/CgExpr.lhs,v
retrieving revision 1.13.2.17
diff -c -r1.13.2.17 CgExpr.lhs
*** CgExpr.lhs 1998/09/29 15:19:32 1.13.2.17
--- CgExpr.lhs 1998/11/17 09:14:06
***************
*** 341,347 ****
| length args + 1 == arity
&& all isFollowableRep (map idPrimRep fvs)
&& isUpdatable upd_flag
! --&& arity <= mAX_SPEC_AP_SIZE
-- Ha! an Ap thunk
= cgStdRhsClosure bndr cc bi srt fvs [] body lf_info payload
--- 341,347 ----
| length args + 1 == arity
&& all isFollowableRep (map idPrimRep fvs)
&& isUpdatable upd_flag
! && arity <= mAX_SPEC_AP_SIZE
-- Ha! an Ap thunk
= cgStdRhsClosure bndr cc bi srt fvs [] body lf_info payload
--
Simon Marlow
Microsoft Research Ltd., Cambridge, U.K.
