Re: GNAT on MinGW

[Fionn Mac Cumhaill <invisible@hiding.from.spam>]

On Thu, 1 Mar 2007 08:56:36 -0800, "Vo, Anh \(US SSA\)"
<Anh.Vo@baesystems.com> wrote:

>-----Original Message-----
>From: comp.lang.ada-bounces+anh.vo=baesystems.com@ada-france.org
>[mailto:comp.lang.ada-bounces+anh.vo=baesystems.com@ada-france.org] On
>Behalf Of Martin Krischik
>Sent: Wednesday, February 28, 2007 11:05 PM
>To: comp.lang.ada@ada-france.org
>Subject: Re: GNAT on MinGW
>
><< Michael Bode schrieb:
>> Martin Krischik <krischik@users.sourceforge.net> writes:
>> 
>>> I was able to create a gcc-4.1.x on a MinGW/Cygwin combo - However it
>>> was GNAT/Pro 6.0.1 being created using GNAT/Pro 6.0.0w. - Not much
>>> help for the Open Source Community.
>> 
>> I thought you were successfull cross compiling a mingw gcc from linux?
>
>No I created a cross compiler for linux. Next step would have been to 
>use that cross compiler to create a native MinGW compiler. But that 
>proved difficult.
>
>I have now tried all sorts of tricks and nothing worked out - there was 
>always some show stopper down the line. >>
>
>Danny Smith at mingw.org has been building GNAT for Windows. The latest
>version is 3.4.5. The MinGW/GNAT has less tools when compared to
>GNAT-GPL.
>
>Way back I successfully built GNAT using cygwin. The problem was that
>the GNAT built did not support tasking run-time because GNAT under
>cygwin, used to build the newer GNAT, does not have tasking run-time
>support. Therefore, any Ada codes involving task will not run.
>
>Currently, I port MinGW/GNAT to cygwin and build gcc/gnat-4.1.2 using
>cygwin. The build hit a snap after going one fourth of the way. I
>scratch my hair with error message as shown below. I would be thankful
>if any one can tell me what headers it looks for.
>
>[...]
>mv tmp-specs specs
>if true ; then \
>          cat ../../gcc-4.1.2/gcc/limitx.h ../../gcc-4.1.2/gcc/glimits.h
>../../g
>cc-4.1.2/gcc/limity.h > tmp-xlimits.h; \
>        else \
>          cat ../../gcc-4.1.2/gcc/glimits.h > tmp-xlimits.h; \
>        fi
>mv tmp-xlimits.h xlimits.h
>echo |  ./xgcc -B./ -B/usr/gnat/mingw32/bin/ -isystem
>/usr/gnat/mingw32/include
>-isystem /usr/gnat/mingw32/sys-include
>-L/cygdrive/e/build-4.1.2/gcc/../ld -E -d
>M - | \
>          sed -n -e 's/^#define \([^_][a-zA-Z0-9_]*\).*/\1/p' \
>                -e 's/^#define \(_[^_A-Z][a-zA-Z0-9_]*\).*/\1/p' | \
>          sort -u > tmp-macro_list
>/bin/sh ../../gcc-4.1.2/gcc/../move-if-change tmp-macro_list macro_list
>echo timestamp > s-macro_list
>The directory that should contain system headers does not exist:
>  /mingw/include
>make[2]: *** [stmp-fixinc] Error 1
>make[2]: Leaving directory `/cygdrive/e/build-4.1.2/gcc'
>make[1]: *** [stage1_build] Error 2
>make[1]: Leaving directory `/cygdrive/e/build-4.1.2/gcc'
>make: *** [bootstrap] Error 2
> 

Hmm. I have been trying to build gcc 4.2 snapshots with MinGW and
MSYS. With MSYS, the /mingw/include location for headers tends to
baffle configure scripts which expect /usr/include. I used the
configure option  --oldincludedir=/mingw/include 

My MinGW/MSYS build goes quite a long way before it crashes. It
completes all three stages of building the compiler (configured to
build c, c++, and Ada) and then fails building the Ada rtl. I posted a
request for help on the gcc-help mailing list, but got no answers;
apparently the Ada wizards don't pay much attention to gcc-help.
(See <http://gcc.gnu.org/ml/gcc-help/2007-02/msg00378.html>

The Ada error message that I got was
"a-calend.adb:44:17: "Ada.Calendar" is not a predefined library unit"

Predefined library units seem to be some kind of private club, and the
stage three 4.2 snapshot compiler seems to be confused as to what the
membership requirements are.