Re: [RFC] Suggested replacement for specs and switch handling

[Neil Booth <neil at daikokuya dot demon dot co dot uk>]

Alexandre Oliva wrote:-

> Except that this time we're not splitting up, but collecting.  I.e.,

It's text manipulation just the same, and therefore straightforward.

> Well, it's certainly system-dependent.

Right, so it goes in the target hooks, just like the target-dependent
CPP, assembler and linker handling will.

> That's precisely the problem of this suggestion.  I agree with
> changing the format of specs to something more readable (think of
> moving from sendmail to postfix :-), but I think it's very important
> to retain the ability to change this stuff at run-time.

I don't agree.  There's way too much hair and fluff at the moment, and
like all hair balls, it just collects more hair :-)

>From the other parts of this thread, it seems to be believed that we
can achieve in other, cleaner ways, the same things we are using specs
for at present.  Specs give extreme run-time configurability that I'm
not offering, but no-one seems to be using that.  If GCC had handled
options the way I suggest from the beginning, do you think the
solution to the issues that people have brought up would be to scrap
it in favour of the complications we have now?  I don't - we'd have
just found a better way to achieve it.  So we should do that now.

I think my proposal, with some modification, is a healthy step
forward, and that we should grasp the opportunity to do so.

Neil.