Re: PATCH: Reverse SUPPORTS_INIT_PRIORITY Change to iris6.h

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Jason" == Jason Merrill <jason@redhat.com> writes:

    Jason> If we wanted it to reflect just current usage, yes. 

    However, most of these features are rather version dependent, and
    the recent trend has been towards testing for these features at
    configure time,

Yes, I think that's the right solution.

In the short term, though, I'm not sure what you want to do.

We could make progress either by eliminating the flag in favor of the
configure-time stuff (good, but takes work, certainly Jeffrey's not
going to do that right now), or by making the
name/documentation/initialization reflect current reality (not as
good, but still improves the code, and is much less work). 

I think we're spending too much effort on this.  This all started as a
simple IRIX bugfix, which you then suggested could be done more
elegantly by using flag_gnu_linker, rather than a constant, so as to
automatically get it right everywhere, but then that didn't work
because flag_gnu_linker isn't actually true iff we are using the GNU
linker, so now we're wondering whether to make the name reflect the
current usage, or make the initialization reflect what we think it
should, but that all seems complicated and hard.  Schwoo. :-)

I guess my instinct is just to bail out, and try to use our energies
to fry bigger fish. :-)

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