Re: OpenOffice 1.9.X and gcj

[Andrew Haley <aph at redhat dot com>]

Caolan McNamara writes:
 > On Mon, 2004-11-22 at 16:41 -0700, Tom Tromey wrote:
 > > An access error like this is more likely to be PR 9369.
 > > This can be hacked around by making the constructor in question (if
 > > that is what it is) not `private'.
 > > 
 > > Caolan, if you can confirm that this is the problem, I'll make an OOo
 > > meta-bug and link this to it.
 > 
 > Going around turning things that seem to cause trouble into public has
 > allowed me to make much more progress, so that seems like the problem
 > there alright. 
 > 
 > I'm now using cvs head from a day or two ago rather than FC3's gcj to
 > get around the previous URL issue. Though with this version I've been
 > bitten a few times by seeming oddness with pre or post incremented
 > variables, 
 > 
 > int foo = 0;
 > 
 > ...
 > 
 > stuff[foo++] = new thingy
 > 
 > where stuff ends up as
 > 
 > stuff[0] = null;
 > stuff[1] = the thingy
 > 
 > whether this is restricted to where foo is a class member or not I'm not
 > sure, I just did a mass rejigging of lines that used the return of ++
 > operations to be seperate increment and assign and continued on my merry
 > way.

This is a really bad problem.  Can you please, as a matter of
importance, construct a test case that we can duplicate for this bug.
I will fix it quickly.

Thanks,
Andrew.