Re: Patch: gcj access$0 error

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Alex" == Alexandre Petit-Bianco <apbianco@cygnus.com> writes:

Tom> I'm wondering if there is some strange case we need to support, for
Tom> instance a chain like this:
Tom> 
Tom> superclass -> derived inner -> top-level subclass -> inner-derived -> here

Alex> I'm taking that `->' is a search path that build_access_to_thisn walks
Alex> backward from `here.' I don't see how we would have to get to `derived
Alex> inner' from `top-level subclass' (I must say I don't see `->' working
Alex> any other way. It would clarify if you had an example.)

That's what it meant.  I couldn't construct an example of what I
(think I) was thinking of.

Alex> Meanwhile, your patch seems correct to me and improves things a
Alex> lot. I ran some supplemental tests, it shows no regressions and
Alex> improves the situation a lot with old snapshots of freenet and
Alex> tomcat. Please check it in both the branch and the trunk for the
Alex> time being (unless your clarification unravels more changes to
Alex> be made to it.)

I'm checking it in.  I think it is an improvement.  If there are
further cases that fail, well, they probably also fail without the
patch.

Tom