On 12 November 2012 22:55, Dennis Clarke wrote:
>
> I think at this point I am left wondering where the bug is. There must be a "bug" somewhere
> because this process of just bootstrapping should work. It doesn't. I just don't see that filing
> a bug report to Oracle would get me anywhere.
>
> Really, I'm trying to get a result here .. and not getting very far. Maybe a change in
> pretty-print.c is needed .. not sure.
>
> I'll stop what I am doing and try again with -D_XOPEN_SOURCE=600 and then if it blows up
> in the usual 71 secs .. that's it. Give up on Solaris 10.
I've just successfully bootstrapped x86_64-pc-solaris2.10 using GCC
4.4.3 as the bootstrap compiler, with in-tree gmp/mpfr/mpc, without
setting any CFLAGS at all, just
CC='gcc -m64' CXX='g++ -m64' ../gcc-4.7.2/configure
--build=x86_64-pc-solaris2.10
I had no problems with the iconv() definition in pretty-print.c, my
builddir/gcc/config.log shows it wasn't found:
| /* end confdefs.h. */
| #include <stdlib.h>
| #include <iconv.h>
| int
| main ()
| {
| iconv_t cd = iconv_open("","");
| iconv(cd,NULL,NULL,NULL,NULL);
| iconv_close(cd);
| ;
| return 0;
| }
configure:10309: gcc -m64 -o conftest -g -fkeep-inline-functions
-I/usr/local/include conftest.c /usr/local/lib/libiconv.so
-L/usr/local/lib -L/usr/openwin/lib -L/usr/local/ssl/lib
-L/usr/local/BerkeleyDB.4.2/lib -R/usr/local/lib >&5
ld: fatal: file /usr/local/lib/libiconv.so: wrong ELF class: ELFCLASS32
Which means the code failing for you in pretty-print.c isn't compiled,
because of:
#if defined ENABLE_NLS && defined HAVE_LANGINFO_CODESET && HAVE_ICONV
On your system presumably builddir/gcc/config.log shows iconv() is
found, and so configure defines HAVE_ICONV.
What does 'grep builddir/gcc/config.log' show?
What does 'env | grep PATH' show? Do you have something in your
LD_LIBRARY_PATH which causes the failing test above to pass?
I think the bug is that the configure test for iconv only tests:
iconv(cd,NULL,NULL,NULL,NULL);
which doesn't check for a POSIX-conforming signature.
