Re: PR3042

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "David" == David Edelsohn <dje@watson.ibm.com> writes:

    Gabriel> I'm confused as to why you're bringing "xlC name
    Gabriel> mangling" into this discussion.

    David> 	Because MARK(!) brought them into the discussion.
    David> This is really getting annoying.

I said there was no compatibility between xlC and G++.  

You had not yet explained at this point that you were looking at
things from the very high-level usage-model point of view.  Typically,
`compatibility' means inter-operation, so perhaps you will forgive my
misunderstanding your argument.

If xlC really behaves just like GCC, in terms of, for example, not
linking the example program I gave, that is very interesting.  I am
surprised that IBM, with the superior instantiation technology it has
in xlC, defaults to a mode in which conforming programs do not link.

If so, however, that does provide an argument for maintaining the
current behavior in that it makes it easier to compile xlC code with
G++.  It doesn't have any impact on compatibility with G++ on other
platforms, or with the compilation of free programs.

If the change I'm suggesting would up Benjamin's chances of making V3
work well on AIX, by simplifying the mental model he has to keep in
his brain, I would think that's important.  With the change, it's easy
to model the AIX semantics on GNU/Linux, which makes it easier to make
the library portable.

Anyhow, we're just going to have to agree to disagree.  Gaby,
Benjamin, and I are all long-time C++ programmers and we think the
current behavior is weird.  Jason is a long-time C++ programmer, and
he likes it.  Apparently, IBM likes it too.  At least some of us our
not going to get our way.

We probably have different favorite colors too.

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