GNAT 3.13p patches for GCC 2.95.3?

Joel Sherrill joel.sherrill at OARcorp.com
Tue Apr 30 18:35:39 UTC 2002



"Rude, Randy [Cont]" wrote:
> 
> > GNAT never officially supported 2.95.x or 3.0.  GNAT has been merged
> > into
> > gcc and will be part of the 3.1 release.  Thanks to Laurent Guerby, I
> > have gotten far enough to have build reports for the rtems targets.
> 
> Do you have any idea how close GCC 3.1 is?  ;-) but see below.

Release preview snapshots have been available for about a week now.
Everyone is working to test and close high priority PRs.  I don't recall
the exact date that is now being targetted but it shuld be in the next
2-3 weeks.

> > If you want a particular CPU that
> > is not powerpc or sparc, then you should watch for my report on
> > the gcc mailing list and help fix. :)  It really looks like a
> > handful of configurery problems are causing most of the
> > failures (not finding include files that are there causes about
> > 4 targets to fail).
> 
> We're interested in the Coldfire 5307.  I'm talking with Bill Butler
> about the version he has running (presumably against a recent snapshot).

A Coldfire person can speak up but I think 2.8.1 does not understand
Coldfire.
 
This was the failure for gcc 3.1 gnat source for the m68k:

m68k-rtems -
        raise.c: In function `__gnat_eh_personality':
        raise.c:601: built-in function `__builtin_eh_return_data_regno'
not
currently supported

I assume this is something that needs to be addressed in the m68k
port but someone more knowledgeable about gcc should comment.  

> It appears then that we have two possible approaches: 1) help with the
> GCC 3.x work on RTEMS; or 2) roll back to a snapshot that builds with GCC
> 2.8.1 so we can get GNAT working.  Since we are still in the evaluation
> stage on RTEMS we'll have to go with 2.8.1 for now.  Does the latest
> snapshot build with GCC 2.8.1?  If not, how far back do we need to go to
> find a snapshot that does?

It should but no port except the sparc and powerpc have been tried 
recently.  The big issue will be that m68k support in gcc 2.8.1 is
COFF and m68k-rtems is now ELF.  This usually results in some
parts of the linker script being unacceptable.

The big issue is Coldfire support in gcc 2.8.1.  The effort to
get 3.1 useable would be better spent.


>         Randy

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
Support Available                (256) 722-9985



More information about the users mailing list