Re: GCC 3.1 Ada status for various platforms

[Joel Sherrill <joel dot sherrill at OARcorp dot com>]

Rainer Orth wrote:
> 
> I've recently tried to bootstrap various GCC 3.1 configurations with ada
> support, namely sparc-sun-solaris2.5.1, sparc-sun-solaris2.8,
> mips-sgi-irix6.2, mips-sgi-irix6.5, alpha-dec-osf4.0f, alpha-dec-osf5.1,
> and i386-pc-solaris2.8.
> 
> I had to start from widely different binary distributions to get an
> acceptable bootstrap compiler (i.e. GNAT 3.13p or newer), available from
> 
>         ftp://ftp.cs.nyu.edu/pub/gnat/private/old/
> 
> * GNAT 3.12p (alpha-dec-osf4.0d)
> 
> * GNAT 3.12p (i386-pc-solaris2.6)
> 
> * GNAT 3.11p (mips-sgi-irix5.3)
> 
> * GNAT 3.13p (sparc-sun-solaris2.5.1)
> 
> Having to bootstrap up to at least 3.13p over one or two steps is
> certainly a mess and nothing that can be reasonably expected from GCC
> users.  Are there any plans to provide more binary distributions of GNAT
> bootstrap compilers for the GCC 3.1 release, either 3.13p, 3.14p or 3.15p
> (i.e. GCC 3.1 itself)?
> 
> Anyway, those bootstraps finally succeeded (getting from alpha-dec-osf4.0d
> to osf4.0f was messiest since I had to replace the fixincluded files
> with matching ones from V4.0F to allow the compiler to be usable at all),
> and the resulting GCC 2.8.1-based gcc/GNAT was usable even on osf5.1 with
> -Wa,-oldas added.
> 
> With those bootstrap GNAT installations, I was able to successfully build
> GCC 3.1 20020502 (C and Ada only) on all the platforms mentioned above,
> albeit only serially at that time.  On IRIX 6.[25], I had to manually
> disable optimization when compiling a-except.adb to avoid an ICE:
> 
> | 3.13p  (20000509) (mips-sgi-irix6.2) Gigi abort, Code=999                |
> | Error detected at /vol/gnu/src/gcc/gcc-3.1-branch-dist/gcc/ada/a-except.adb:1392:7|
> 
> Even make gnatlib_and_tools succeeded on all platforms but
> mips-sgi-irix6.[25], where I get many ICEs like the following:
> 
> | 3.1 (20020501) (mips-sgi-irix6.2) GCC error:                             |
> | Internal compiler error in handle_epilogue_set, at function.c:7521       |
> | Error detected at a-chahan.adb:428:4                                     |
> 
> This error persists even without any optimization.  I've filed PR ada/6552
> for this issue.
> 
> Overall, Ada status on the branch looks reasonable over the platforms
> investigated, though I cannot be sure due to the lack of a testsuite.

Great work!!  There is a script harness for the ACATS that Laurent
Guerby put together.  I have used it for native GNU/Linux x86 and
GNAT/RTEMS testing.  It is available at:

ftp://ftp.oarcorp.com/pub/rtems/snapshots/contrib/acats/

It really shouldn't be under rtems but that's another problem. :)

It is pretty easy to run and will automatically download
the source for the acats if it needs to.


> As has been pointed out, it's critically important that the build procedure
> be improved in time for the GCC 3.2 release: the current manual procedure,
> lack of readily available bootstrap compilers, and missing support for
> parallel bootstraps will certainly deter all but the most determined users.
> 
>         Rainer

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