dl_iterate_phdr() deadlock patch

[Jeff Sturm]

On 1 Mar 2003, Tom Tromey wrote:
> Is this test still valid when we reduce the count?  I'd like it to at
> least not time out.  Really I'd like it to take very little time at
> all (unreasonable, I know)

Not unreasonable at all.  If you use jikes as your source code compiler,
this test takes just a couple seconds.

So this particular failure really does indicate a deficiency in gcj.

Jeff