Re: libg++-2.8.1-980207 and egcs-2.91.08 980214

[Jason Merrill <jason at cygnus dot com>]

>>>>> Alexandre Oliva <oliva@dcc.unicamp.br> writes:

> In fact, the problem is the way HUGE_VAL is defined: it depends on
> named initializers, which eg++ no longer supports (the original
> support was broken).

I've reenabled them in the parser, with a sorry if they would actually make
a difference.  In HUGE_VAL, it doesn't.

> Perhaps this is a job for fixincludes...  Unless we get named
> initializers support fixed and back.

A volunteer to port the C frontend's initializer parsing to the C++
frontend would be welcome.  rms rewrote the C code back in 1993, but the
changes never made it to C++:

Thu Jul 22 14:59:38 1993  Richard Stallman  (rms@mole.gnu.ai.mit.edu)

        * c-typeck.c: Output most constructors an element at a time.
        (digest_init): Deleted code for raw constructors.
        (process_init_constructor): Function deleted.
        (start_init, finish_init): New functions.
        (really_start_incremental_init)
        (push_init_level, check_init_type_bitfields, pop_init_level)
        (set_init_index, set_init_label, process_init_element)
        (output_init_element, output_pending_init_elements): New functions.
        * c-tree.h: Declare most of those functions.

        * c-parse.in (cast_expr): Use new initializer parsing functions.
        (initdcl, notype_initdcl): Likewise.
        (init): Likewise.
        (initlist_maybe_comma, initlist1): These replace initlist.
        (initelt): New nonterminal.
        Change specified index syntax to `[INDEX]='.
        Change specified field syntax to `.NAME='.

Jason