Re: Possible patch for miscompiled defer-pops

[Per Bothner <per at bothner dot com>]

Richard Henderson <rth@redhat.com> writes:

> If so, it would be good to have commentary to that effect.

I checked in my patch as is, but with this comment:

/* Should perhaps use expand_label, but this is simpler and safer. */

The label_chain chain stuff in expand_expr seems to be only used for
error checking - at least searching for "label_chain" didn't find
anything else.  This is arguably useless, as such checking should be
done in the front-end, at least for languages that represent entire
functons as trees.  One might argue it would be cleaner to use
expand_label, but I thought it safer to stick to the minimal change,
especially for gcc 3 - and it was easier to use the same patch for both.
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/per/