Integer.decode() patch for negative numbers

[Tom Tromey]

>>>>> "Mark" == Mark Wielaard <mark@klomp.org> writes:

Mark> I have gone through the changes to Integer in Classpath but it
Mark> is actually a large rewrite (that will bring some other
Mark> important improvements).

Mark> I have tested this with mauve and it fixes the decoding of
Mark> negative haxadecimal values but it seems to introduce some new
Mark> failures with to[Hex|Octal|Binary]String(). This might be my
Mark> fault for not applying to correct patches though.

It would be nice if we could approach a 100% Classpath merge.  Then we
could just copy in changes.  I think this will mean becoming a little
more strict about testing on the Classpath side, but that will be good
all around.

Mark> Would it be OK for me to checkin my original patch only to the
Mark> 3.1 branch and do the real merging of Integer (and other wrapper
Mark> classes) after 3.1 is released?

Yes.  Thanks, and thanks for looking into the Mauve regressions.

Mark> Note that the String change is actually part of the patch that
Mark> Eric posted earlier this month and which seems to not have been
Mark> approved yet.
Mark> <http://gcc.gnu.org/ml/java-patches/2002-q1/msg00699.html>

I'm not even looking at that until 3.1 ships.  I don't have time.
Plus, String is a tricky part of libgcj.  While it might be possible
to merge it (for Eric, apparently, anything is possible :-), I
definitely don't want to do it just before a release.

Tom