[nightly] 15-Apr-2004 build of HEAD on Mingw32 (water)

Tue Apr 20 17:56:29 EDT 2004

Sigh.  No we haven't made up our minds, we're trying to discuss what the
right thing to do would be.

Good point about building GHCi libraries.  I wonder how that works now;
perhaps the cygwin ld produces the same result as the mingw ld.

Cheers,
	Simon

On 20 April 2004 16:48, Sigbjorn Finne wrote:

> You've already made your mind up over there it seems,
> so do whatever you like..
> 
> How do you create those GHCi object files?
> 
> --sigbjorn
> 
> ----- Original Message -----
> From: "Simon Marlow" <simonmar at microsoft.com>
> To: "Sigbjorn Finne" <sof at galois.com>; "Simon Peyton-Jones"
> <simonpj at microsoft.com>
> Cc: "Mike Thomas" <mike.thomas at brisbane.paradigmgeo.com>;
> <cvs-ghc at haskell.org>
> Sent: Tuesday, April 20, 2004 08:43
> Subject: RE: [nightly] 15-Apr-2004 build of HEAD on Mingw32 (water)
> 
> 
>> I'd prefer not to be required to change my PATH to build GHC.  That
>> would be a pretty annoying restriction, because it's something you
>> have to do every time you want to do something in the build tree,
>> rather than just once when you configure the build.
>> 
>> The --with-gcc option is *supposed* to work properly.  Indeed, I use
>> it for testing different versions of gcc.  We should fix any cases
>> where it doesn't work properly.  Why is GMP still working?
>> 
>> As for binary distributions: they use a different method for finding
>> gcc, which is hardcoded into SysTools.hs.
>> 
>> --with-ld isn't necessary, because linking uses GCC.  As far as I can
>> see, --with-dllwrap is the only missing option.  Perhaps it could be
>> derived automatically from the --with-gcc option, pretty much like we
>> derive ghc-pkg from --with-ghc.
>> 
>> Cheers,
>> Simon
>> 
>> On 20 April 2004 16:32, Sigbjorn Finne wrote:
>> 
>>> Quite the opposite, I'd say. We _want_ to always be using
>>> mingw-based tools, so insisting that they're available ahead
>>> of anything else sounds more robust. How do you know that
>>> the --with-gcc setting is honoured throughout all the Makefiles
>>> inside of fptools/ ? (which it isn't btw; cf. ghc/rts/gmp/ ) Using
>>> --with-gcc=... is also incompatible with creating portable binary
>>> distributions (as the path WhatGccIsCalled is hardcoded
>>> into main/Config.hs)
>>> 
>>> Presumably you want to add --with-ld=... too?
>>> 
>>> --sigbjorn
>>> 
>>> ----- Original Message -----
>>> From: "Simon Peyton-Jones" <simonpj at microsoft.com>
>>> To: "Sigbjorn Finne" <sof at galois.com>; "Mike Thomas"
>>> <mike.thomas at brisbane.paradigmgeo.com>
>>> Cc: <cvs-ghc at haskell.org>
>>> Sent: Tuesday, April 20, 2004 02:25
>>> Subject: RE: [nightly] 15-Apr-2004 build of HEAD on Mingw32 (water)
>>> 
>>> 
>>>> Well, if you are a Cygwin user, you might well expect to get the
>>>> Cygwin gcc when you type 'gcc' on the command line.  In general I
>>>> think it's more robust if mingw/bin doesn't appear in your PATH *at
>>>> all*.  That's why I always configure with
>>>> --with-gcc=c:/mingw/bin/gcc 
>>>> 
>>>> Sounds as if I should add
>>>> --with-dllwrap=c:/mingw/bin/dllwrap
>>>> 
>>>> Is that right?  Would configure understand that?  If not, could one
>>>> of you make it do so?  Let me know and I'll update the
>>>> instructions. 
>>>> 
>>>> 
>>>> (Of course, if you are using MSYS, then msys/bin will be in your
>>>> path, and that gcc probably is the same as mingw/bin/gcc; ditto
>>>> dllwrap.  But doing the --with stuff ensures you are independent
>>>> of the PATH and build platform.) 
>>>> 
>>>> Simon
>>>> 
>>>>> -----Original Message-----
>>>>> From: cvs-all-bounces at haskell.org
>>>>> [mailto:cvs-all-bounces at haskell.org] On Behalf Of Sigbjorn Finne
>>>>> Sent: 16 April 2004 01:16 To: Mike Thomas
>>>>> Cc: cvs-ghc at haskell.org
>>>>> Subject: Re: [nightly] 15-Apr-2004 build of HEAD on Mingw32
>>>>> (water) 
>>>>> 
>>>>> Hi Mike,
>>>>> 
>>>>> good to hear that your change did the trick.
>>>>> 
>>>>> There may have been a good reason for arranging your PATH to
>>>>> not include mingw/bin ahead of cygwin path entries sometime
>>>>> in the distant past, but there's no reason not to do this today.
>>>>> I've been building fptools/ bits using that setup for years now.
>>>>> 
>>>>> --sigbjorn
>>>>> 
>>>>> ----- Original Message -----
>>>>> From: "Mike Thomas" <mike.thomas at brisbane.paradigmgeo.com>
>>>>> To: "Sigbjorn Finne" <sof at galois.com>;
>>>> <miketh at brisbane.paradigmgeo.com>
>>>>> Cc: <cvs-ghc at haskell.org>
>>>>> Sent: Thursday, April 15, 2004 17:01
>>>>> Subject: RE: [nightly] 15-Apr-2004 build of HEAD on Mingw32
>>>>> (water) 
>>>>> 
>>>>> 
>>>>>> Hi Sigbjorn.
>>>>>> 
>>>>>> Thanks for the excellent observation.
>>>>>> 
>>>>>>> I've seen this build error flit past for quite a while.
>>>>>> 
>>>>>> Yes; I'm so deeply enmired in GNU Common Lisp and Maxim Windows
>>>>>> ports  at the moment that I just run an occasional GHC build to
>>>>>> make myself feel better, but with little regard for the output.
>>>>>> 
>>>>>> 
>>>>>>> As far as I'm able
>>>>>>> to diagnose, it appears as if it might be due to a local mingw
>>>>>>> setup issue; I'm able to build HDirect just fine using multiple
>>>>>>> mingw versions/installations. 
>>>>>>> 
>>>>>>> 'dllwrap' perhaps being slurped in from a location other than
>>>>>>> the mingw installation tree? 
>>>>>>> 
>>>>>> 
>>>>>> Well spotted.  I just ran a test with the dllwrap entry in
>>>>>> "hdirect/mk/paths.mk" hardwired to the MinGW32 dllwrap and it
>>>>>> all went perfectly. 
>>>>>> 
>>>>>> This is going wrong because I am building according to the
>>>>>> instructions on the GHC web site which explicitly forbid putting
>>>>>> MinGW32 modules ahead of Cygwin (fow the past few months anyway,
>>>>>> I once used to ignore this aspect of the build instructions
>>>>>> which I regard(ed) as bad): 
>>>>>> 
>>>>>> "Do not add any of the mingw binaries to your path. They are only
>>>>>> going to get used by explicit access (via the --with-gcc flag you
>>>>>> give to configure later). If you do add them to your path you are
>>>>>> likely to get into a mess because their names overlap with
>>>>>> Cygwin binaries." 
>>>>>> 
>>>>>> I'll aim to do a build tonight with MinGW32 back in it's rightful
>>>>>> place in the path to make sure the whole system works correctly.
>>>>>> 
>>>>>> 
>>>>>> Cheers
>>>>>> 
>>>>>> Mike Thomas.
>>>>>> 
>>>>> 
>>>>> _______________________________________________
>>>>> Cvs-ghc mailing list
>>>>> Cvs-ghc at haskell.org
>>>>> http://www.haskell.org/mailman/listinfo/cvs-ghc
>>> 
>>> _______________________________________________
>>> Cvs-ghc mailing list
>>> Cvs-ghc at haskell.org
>>> http://www.haskell.org/mailman/listinfo/cvs-ghc