h8300-rtems configure problem in libstdc++-v3
Joel Sherrill
joel.sherrill@OARcorp.com
Tue Feb 13 11:17:00 GMT 2001
Carlo Wood wrote:
>
> On Tue, Feb 13, 2001 at 11:05:08AM -0600, Joel Sherrill wrote:
> > if (eval "$ac_cpp conftest.$ac_ext") 2>&5 |
> > egrep "ok" >/dev/null 2>&1; then
>
> In a case like this you can best edit `configure' directly
> and add a few lines to know exactly what is going on.
>
> A few suggestions, add just prior to the lines above:
>
> echo "$ac_cpp conftest.$ac_ext" > /tmp/junk
> eval echo "$ac_cpp conftest.$ac_ext" >> /tmp/junk
> (eval "$ac_cpp conftest.$ac_ext" >> /tmp/junk 2>&1)
> (eval "$ac_cpp conftest.$ac_ext") 2>&5 | echo "we get here" >> /tmp/junk
>
> etc.
>
> Then examine /tmp/junk (after rerunning configure).
> Add new echo's and rerun configure until you know *exactly* what is happening.
Thanks. With this added, I am now seeing this output in /tmp/junk:
$CXXCPP $CPPFLAGS conftest.C
/lib/cpp conftest.C
++ /lib/cpp conftest.C
# 1 "conftest.C"
# 2574 "configure"
# 1 "confdefs.h" 1
# 2575 "configure" 2
we get here
which indicates that configure is using the native CPP not the
cross one just built. This is inconsistent with the other
cross configurations I have built using the same source
tree. In fact, the i960-rtems configuration actually
got an internal compiler error in libstdc++-v3 while answering
this email.
Jeff Law suggested that the h8 is not supported for C++ because
low end models have 16-bit address spaces. But I would think
the generic configurery should be able to correctly guess CPP
regardless.
Any thoughts or should I just disable C++ for h8300-rtems
and move on. :)
> --
> Carlo Wood <carlo@alinoe.com>
--
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
More information about the Gcc-bugs
mailing list