Hi,

thanks for the report - 

Thomas Nordin writes:
> Environment:
>   bash-2.01$ uname -a
>   CYGWIN32_NT TIGGER 4.0 19.3 i686 unknown
>   bash-2.01$ perl --version
>   This is perl, version 5.004_04 built for cygwin32
> 
> Building ghc-3.02 from hc files:
>   mk/config.mk.in:230 is "GhcLibWays" should be "GhcLibWays="
>   ghc/compiler/basicTypes/Unique.hc:2714 is "W_ crEE_;" should be "W_ crEG_;"

Eh? Looks like the tarball hit a cosmic ray while in-transit,
Unique.hc:2714 does say "W_ crEG_;" ! :-)

>   missing *.hi files for libraries

Yep, with the current booting-from-hc setup, you'll have to generate
these yourself after hsc has been successfully linked. The
booting-from-hc section in the installation doc have been updated
post-3.02 to include a paragraph about how to go about doing this.
(Not of much use to you, but I thought I'd mention it nontheless :-) 

>   ./configure:
>     documentation says "-enable-hc-booting" should be "--enable-hc-boot"
>     --help does not show --hc-boot flag
> 
> Doing 3.02 compilation with 3.02
>   when building the profiled runtime:
>     ghc/runtime/Itimer.c:81 TIMERR_NOCANDO undefined
>     ghc/runtime/Itimer.c:87 LPTIMECALLBACK undefined
>     ghc/runtime/Signal.c:409 TIMECALLBACK undefined
> 

The header files that come with cygwin32 don't have support for the
multimedia API, which we need to get at some decent timer support.
To fix, you'll need to patch the header files - let me know if you 
want the diffs and I'll mail them to you.

>     this does *not* work: ghc-3.02 -cpp -fglasgow-exts -Rghc-timing -I. -IcodeGen 
>-InativeGen -Iparser 
>-iutils:basicTypes:types:hsSyn:prelude:rename:typecheck:deSugar:coreSyn:specialise:simplCore:stranal:stgSyn:simplStg:codeGen:absCSyn:main:reader:profiling:parser:nativeGen
> -recomp   -Onot -H12m  -c specialise/Specialise.lhs -o specialise/Specialise.o -osuf 
>o
>     adding -v makes it work!(?)
> 

I see this intermittently too. Fingers point in the general direction
of how cygwin32 handles read-only FILEs - hopefully the ongoing
rewrite of IO will work around this problem..

The general workaround to this is to interleave the renamer's reading
of files with some output - -dshow-passes is a less noisy alternative
to -v.

--Sigbjorn

Reply via email to