darcs patches and questions
simonmarhaskell at gmail.com
Mon Aug 21 08:07:05 EDT 2006
Esa Ilari Vuokko wrote:
> Esa Ilari Vuokko wrote:
>>Fri Aug 11 21:07:21 FLE Standard Time 2006 Esa Ilari Vuokko <ei at vuokko.info>
>> * Include ghcplatform.h for primops' benefit in Windows
>>This patch is needed to generate all the primops for async-stuff under
>>mingw. One of the recent patches removed the include that had hostos
>>defines, this adds the minimal (I think) include needed.
> Sorry, that's just not true. The patch isn't useful at all.
> I was using make -j2, and it appears that there isn't correct dependency
> between primops.txt.pp and ghc_boot_platform.h. This does cause an error
> while preprocessing primops.txt.pp, but that isn't consider fatal for some
> reason. Without ghc_boot_platform.h's mingw32-defines, resulting
> primops.txt doesn't have async-stuff.
> Adding simple dependency into makefile causes havoc, as the
> ghc_boot_platform.h got created during boot and during stage1 (and
> probably stage2, but I haven't got that far), and it just so happens
> to interfere with util/FastString, header file being partial during
All the compiler sources depend on ghc_boot_platform.h, but this dependency
isn't expressed explicitly. We could add that dependency, but it'll result in
too much recompilation: everything would be recompiled when the Makefile
changes, which is not usually what you want.
So currently we make sure ghc_boot_platform.h is built early enough by including
it in the boot target. A better solution might be to put the Makefile code that
builds ghc_boot_platform.h into a separate Makefile so we can add a dependency
ugh, this is horrible, I can't find a good way to fix it. I've committed a hack
that seems to work for me, let me know if it helps you.
Would you like to resubmit your other patches, I'm not sure which ones are still
More information about the Cvs-ghc