Build Failures on Master (mostly C++ issues)

Joel Sherrill joel at rtems.org
Sat Apr 9 15:54:39 UTC 2016


On Sat, Apr 9, 2016 at 10:26 AM, Daniel Gutson <
daniel.gutson at tallertechnologies.com> wrote:

>
> El 9/4/2016 12:24, "Joel Sherrill" <joel at rtems.org> escribió:
> >
> >
> >
> > On Sat, Apr 9, 2016 at 10:17 AM, Daniel Gutson <
> daniel.gutson at tallertechnologies.com> wrote:
> >>
> >>
> >> El 9/4/2016 12:15, "Joel Sherrill" <joel at rtems.org> escribió:
> >> >
> >> > Hi
> >> >
> >> > These are the BSPs which do not build all tests with C++ and the
> in-tree network stack enabled:
> >> >
> >> > + epiphany_sim - GCC ICE building mongoose.c. No investigation.
> >>
> >> Gcc 6? Could you please attach more data, such as the gcc output and
> the test?
> >
> > Most targets are like this:
> >
> > arm-rtems4.12-gcc (GCC) 6.0.0 20160327 (RTEMS 4.12, RSB
> 648e2d5d6c308227c456434238da5f3f881bddc7, Newlib 2.4.0)
> >
> > But epiphany hasn't been merged yet and has its own issues:
> >
> > epiphany-rtems4.12-gcc (GCC) 4.9.0 20140411 (RTEMS 4.12, RSB
> 648e2d5d6c308227c456434238da5f3f881bddc7, Newlib
> 8b1ede3ce11d53292036aadfcfb6043df0235f9c)
> >
> > I can post the ICE but not even sure what good it would do.
>
> We can fix it :)
>
> https://devel.rtems.org/ticket/2686#ticket

But I will probably need to bump the newlib version soon on it. All targets
should be using newlib 2.4.0. There are some pending POSIX additions which
require a newer newlib.

--joel


> >
> >>
> >> >
> >> > + m32c - does not even build a C++ compiler. This is a small CPU and
> there isn't any general support for C++. This was why it was on my list of
> automatically disabled.
> >>
> > This is gcc 6. And I think we just disable C++ for m32c.
> >
> >>
> >> > + moxiesim - cxx_iostream.exe fails to link bcause it can't find
> __dso_handle. I don't know if C++ works for moxie-elf or if this is RTEMS
> specific.
> >
> > This is gcc 6. It has had this issue for months. I don't think C++ has
> ever worked.
> > I have investigated but don't have a clue which tiny bit is wrong.
> >
> >>
> >> > + or1k - needs newlib bumped to match tree.
> >
> > or1k was also using older gcc 4.9.3. It hasn't been merged yet. I am
> updating it
> > now to the latest newlib.
> >
> >>
> >> > + sh/gensh4 - undefined reference to __gnu_cxx::__atomic_add(int
> volatile*, int) in cxx_iostream test
> >>
> > This is gcc 6. It has had this issue for months. It broke when the
> atomic changes
> > and our C++ configuration changed. I have investigated but don't have a
> clue
> > which tiny bit is wrong.
> >
> >
> >>
> >> > --joel
> >> >
> >> > _______________________________________________
> >> > devel mailing list
> >> > devel at rtems.org
> >> > http://lists.rtems.org/mailman/listinfo/devel
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20160409/8e6451e3/attachment-0002.html>


More information about the devel mailing list