Re: Really obscure C++ mangling error with -fnew-abi

[Alex Samuel <samuel at codesourcery dot com>]

"Martin v. Loewis" <martin@loewis.home.cs.tu-berlin.de> writes:

  Alex> I think that comment is simply out-of-date.  It should be
  Alex> removed.

  Martin> It's not a comment; it's a normative statement; it says that
  Martin> the function name, although a name, is not added to the
  Martin> compression table.

Point taken.  I'll rephrase: I think it's an out-of-date statement
that should be removed.

  Martin> The "entities appearing in syntax tables" statement is a
  Martin> non-normative remark, though. The function certainly appears
  Martin> in the syntax table (at least in gcc, and in any other
  Martin> compiler).

The way the ABI committee mangling discussions have played out, as I
perceive them at least, is that the normative statements have been
adjusted to suit this philosophy.  As I understand it, the name of the
function without its full signature was deemed by the committee as
inappropriate for substitution.  Did you get a different read?

Issues of the specification aside, do you see anything technically
wrong with the mangling scheme as currently implemented in g++?

  Martin> <function name> is still part of <encoding>...

But <encoding> is not a <substitution>.

Regards
Alex Samuel
CodeSourcery LLC