[Info-vax] GCC on VMS, was: Re: What would you miss if DECnet got the chop? Was: "bad select 38" (OpenSSL on VMS)

[Simon Clubley]

On 2016-09-21, Chris <xxx.syseng.yyy at gfsys.co.uk> wrote:
> On 09/20/16 21:41, Simon Clubley wrote:
>>
>> That's exactly how I tried to bootstrap the FSF gcc/binutils
>> code base for VMS; I used Linux as the cross compiler host
>> and tried a cross compiler build for VMS Alpha.
>>
>> I tried four of the gcc frontends when building the cross compiler:
>> Ada, C, C++ and Fortran. C was the only one I was able to build
>> usable VMS binaries for with the cross compiler. Both the Ada and
>> C++ frontends failed to compile while building the cross compiler
>> itself and a simple Fortran test program built but fell over during
>> image activation (or very shortly afterwards; I can't remember).
>>
>
> Don't suppose you tried turning all optimisation off, or had a look
> at the asm output ?.
>

If you mean Fortran, I gave up pretty quickly as it was only a
secondary goal. My main goal was Ada and it became pretty obvious
rather quickly that the Ada problem wasn't going to be a 5 minute
fix as the build triggered an internal compiler error every single
time no matter what I tried...

> Only needing C, haven't tried to build C++ or any of the others yet,
> but the gcc version difference can have quite an effect...
>

I tried various versions but the earlier versions fell over even
earlier during the build process due to even more missing pieces
than the latest (at the time) gcc and binutils versions.

Simon.

-- 
Simon Clubley, clubley at remove_me.eisner.decus.org-Earth.UFP
Microsoft: Bringing you 1980s technology to a 21st century world