Re: C++ PATCH: Fix -O3

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Daniel" == Daniel Berlin <dan@www.cgsoftware.com> writes:

    Daniel> If it's not, it's quite possible we are confusing SGI's
    Daniel> dwarf2 optimizations in the assembler and linker (I would
    Daniel> imagine it's more linker than anything else, but you never
    Daniel> kow).

We're using SGI as/ld.  If you look at the SGI compiler output, there
are these funny notations in the .s file that get turned into DWARF2
debugging information by the assembler.  But, we pass -g0 to the
assembler which keeps it from doing this kind of thing.  Certainly,
the GCC 2.95.2 DWARF2 works OK, so *something* about what's different
is confusing things.  I noticed that the order of some of the DWARF2
fields is different, but I'm not sure if/how that manners.

If you're still stuck tomorrow, let me know, and I can get you the
actual .o/a.out files to look at if that would be helpful.

Thanks,

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com