AIX Status

[David Edelsohn]

	The patch works for me.  Now you're almost up to the same point
that I am in debugging V3.

	One thing to watch out for is that the AIX assembler does not
produce correct ios.o and string-inst.o.  I believe that this is because a
few symbol names overrun some fixed buffer inside AIX as.  I manually have
been producing assembly files for those two files (in src subdirectory)
and assembling those with the GNU assembler.

	The GNU assembler complains about many debugger ".line" and
".bb"/".eb" directives which have a negative number as their argument.
The AIX assembler apparently ignores these silently.  For example:

        .bi     "/u/dje/src/GNU/gcc/libstdc++-v3/libsupc++/exception"
        .line   -79
        addi 1,1,72
        lwz 0,8(1)
        lwz 29,-12(1)
        mtlr 0
        blr
        .bb     -79
        .eb     -79
        .bb     -79
        .eb     -79
        .bb     -79
        .eb     -79
        .bb     -79
        .eb     -79
        .ef     709


	Once the library is built, the majority of the G++ testsuite
failures are missing vtables and missing typeinfo:

ld: 0711-317 ERROR: Undefined symbol: vtable for Y<double>

and

ld: 0711-317 ERROR: Undefined symbol: typeinfo for T4
ld: 0711-317 ERROR: Undefined symbol: typeinfo for T3
ld: 0711-317 ERROR: Undefined symbol: typeinfo for T2
ld: 0711-317 ERROR: Undefined symbol: typeinfo for T1
ld: 0711-317 ERROR: Undefined symbol: typeinfo for S2
ld: 0711-317 ERROR: Undefined symbol: typeinfo for S1

I don't know if these are fixable cases of G++ expecting weak support to
avoid duplication and never emitting any information in the fallback case.
Or more examples of the more difficult template problems.  Many of these
G++ testcases worked in the gcc-2.95 release.

Thanks, David