<offlist> Re: C compiler cannot create executables

[Ralf Corsepius]

On Mon, 2007-04-02 at 12:44 -0500, Joel Sherrill wrote:
> Ralf... do you still feel like 4.7.1 is justified based on this thread?
Now that you sent this to the list, let me reply to the list.

Yes, I still think so, because

* the RTEMS bug wrt. CFLAGS handling in tools/ will hit all users of
pc386-based BSPs (which I assume to be one of the most popular BSPs to
getting started) on non-i386 hosts (e.g. on x86_64-hosts).

* the other issues the OP reports, partially seem to originate from
Ubuntu issues. However, due to lack of Ubuntu installations, I can't
verify. So far, I am having doubts on some of the issues he reports, but
also doubt, it's RTEMS who is to blame.

Releasing a 4.7.1 with the "fix" applied should work around both of
these issues and will avoid having to answer these questions the next
time.

One issue not having been fixed is building the posix BSP on x86_64.

AFAICT, so far, the POSIX-BSP doesn't support x86_64 cpus nor does
RTEMS's configuration gracefully handle (e.g. reject) such
build-environments.

I.e. at the moment, building the posix BSP on x86_64/amd64 systems,
probably would require a multilib'ed gcc (applicable e.g. on RH/Fedora
based distros) or i386 gcc installed in parallel to an i86_64/amd64-gcc
(I could be wrong, but AFAIK, this is how many Debian-based distros
package GCC).

> Just checking.

> >> Thus I did away with my 64bit Ubuntu (no negative impact to me)
> >> and went 'backwards' with 32bit Ubuntu. The problem seems to be
> >> more of a gcc issue.
> >>     
> > Well, there are at least two issues I am aware about.
> >
> > 1. there is a bug in rtems-4.7.1

Ugly typo, this should have been: rtems-4.7.0

The rtems-4.7.0.tar-ball contains the bug. It is supposed to be fixed on
rtems-4-7-branch and rtems-cvs-HEAD/rtems-4.8.

Ralf