G++ compiler bug on SCO OSR5

[Robert Lipe]

> ./../am2msdev-0.4.0/am/AmFileEditor.h:72: Internal compiler error in `dwarfout_
> finish', at dwarfout.c:6102

Yes, the same bug is triggered on the snapshots.  I can even duplicate it
on Linux if I invoke:

$ ./cc1plus  -O2 -fpic -gdwarf -g1 /tmp/*.ii
[ munch ] 
stream &operator<< (ostream &, const basic_string<charT, traits, Allocator> &) [with charT = char, traits = string_char_traits<char>, Allocator = __default_alloc_template<false, 0>]
../../am2msdev-0.4.0/am/AmFileEditor.h: At top level:
../../am2msdev-0.4.0/am/AmFileEditor.h:72: Internal compiler error in `dwarfout_finish', at dwarfout.c:6110
Please submit a full bug report.
See <URL: http://www.gnu.org/software/gcc/faq.html#bugreport > for instructions.


So it's probably a generic C++/Dwarf-1 thing and not an OpenServer thing.

> [[ Robert - should the Skunkware G++ compiler be emitting dwarf output?? ]]

The OpenServer GCC,  unless told otherwise will emit Dwarf-1 for ELF
and COFF thingies for COFF.

The UnixWare2 GCC always emits Dwarf-1.

The UnixWare7 GCC emits Dwarf-2 by default but can be "dumbed down" to Dwarf-2.


> gcc-lib/i386-pc-sco3.2v5.0.5/2.95.2/crtend.o /usr/ccs/lib/crtn.o
> /usr/local/lib/gcc-lib/i386-pc-sco3.2v5.0.5/2.95.2/libstdc++.a(cstrmain.o): fata
> l error: symbol `__pl__H3ZcZt18string_char_traits1ZcZt24__default_alloc_template
> 2b0i0_RCt12basic_string3ZX01ZX11ZX21PCX01_t12basic_string3ZX01ZX11ZX21` multiply
> -defined, also in file AmFileEditor.o
> collect2: ld returned 1 exit status

This looks like the fabled "system V weak" thing that was fixed on the
trunk but not on the 2.95 branch.  Ron respun binary kits of 2.95.2
with that patch applied and called them "pl1".  I don't know how he
distinguished them other than renaming the tarball to know which one you
have.

RJL