Re: [PATCH] ARM: fix macro expansion with a pld insn as a macro argument

[Roland McGrath <mcgrathr at google dot com>]

On Mon, Nov 12, 2012 at 11:51 PM, Jan Beulich <JBeulich@suse.com> wrote:
> That's pretty unfortunate. I'd really like to see problems like this
> to be addressed in their entirety, rather than just piecewise -
> that's, in my opinion, part of the reason why there are so many
> half baked things in binutils, which the unsuspecting way too
> easily stumbles across.

Cleanup is nice, but so is making forward progress.  I have a fix for the
one bug that is actually preventing me from doing work, and it is a very
simple change.  Handling the # case is much less straightforward and
probably cannot be done cleanly without rewriting a bunch of hairy stuff,
greatly increasing the likelihood of regressions, etc.

Holding my simple fix hostage to the prayed-for cleanup of the whole mess
is just not reasonable.

> Fact is that this pre-parsing has been causing problems every
> once in a while, but I don't see anyone stepping up and fixing
> this altogether (not the least because of how fragile this is, and
> hence how high the potential of regressions would be).

It seems likely to be that the entire APP (i.e. whitespace
canonicalization) pass should just go away.  Certainly it's extremely loony
that this canonicalization pass happens before macro expansion.

But that's neither here nor there.  Please don't stand in the way of a
simple, practical fix to a real problem affecting a real use while you
demand an enormous amount of work to clean things up for no known immediate
practical benefit.


Thanks,
Roland