Re: GCC support for QNX?

[jtc at acorntoolworks dot com (J.T. Conklin)]

Jack Lloyd <lloyd@acm.jhu.edu> writes:
> Does anyone know what the status is of GCC's support for QNX (x86)? [I'm
> specfically thinking about using the free-beer QNX 6.1.0 release.] I don't
> see it mentioned at all on the builtstat pages for 3.0 or 3.1. I'll
> probably try building it tommorow night, but if it's not supported at all,
> or has been allowed to rot, I'd like to find out now rather than try to
> debug a bad build.

Because of extremely bad experiences with the C++ support in QNX 6.1
(after generally favorable experiences with C++ in QNX 6.0), my boss
asked me to look into rolling our own (at least for the x86 platform
which we use).  A day or so later, plus some small tweaks to some of
the QNX header files (there were some __GNUC__ == 2 preprocessor
conditionals that needed to be changed to to __GNUC__ >= 2, etc.) and
I had a functional cross compiler (which is how I recommend using it).

Unfortunately, I didn't get C++ running because the new Dinkumware C
library headers conflict with the GNU stdc++ library headers (I'm not
exactly sure who's at fault here.  I hope to have the bandwidth to
raise the issue on the libstdc++ mailing list shortly) and the new
calling convention means we couldn't link with the provided libraries.
At the time, I thought the swiftest course of action was to see what
was necessary to licence the source code for the QNX-modified Dinkum 
libraries and build them with new toolchain.  After many blind alleys,
in retrospect I should have focused on getting libstdc++ running.

Earlier in the week I heard from RMS that he had signed the assignment
for my new employer, so my port can be contributed.  It probably makes
more sense to see if I can resolve the libstdc++ issues and figuring
whatever fixinclude magic is necessary so manual editing of headers
isn't necessary.  But I can make it available as it's refined until
then.

        --jtc

-- 
J.T. Conklin