Caciocavallo bug on JDK > 8u40b06

[Severin Gehwolf]

On Thu, 2015-02-19 at 17:34 +0100, Roman Kennke wrote:
> Am Donnerstag, den 19.02.2015 um 12:39 +0100 schrieb Severin Gehwolf:
> > Hi,
> > 
> > There is a caciocavallo bug due to X11FontManager made final[1]. We hit
> > this problem in Thermostat with cacio-tta version 1.3 on JDK 9 (or a
> > more recent OpenJDK 8 build). Reproducer is attached.
> > 
> > Is there any chance for you to do a cacio release ([3] depends on the
> > change in [2]) and get patch in [2] pushed to JDK 9. Then backport [2]
> > to JDK 8.
> 
> I am working on getting the fix into JDK9. If it also affects JDK8, we
> should probably backport it.

Thanks! Yes they backported it to JDK8 and will be in 8u40 which will be
GA'ed in a month.

> > Perhaps cacio could be made compatible with more JDKs by checking the
> > version? I do see this problem surface soon for more people since JDK
> > 8u40 will be GA soon.
> 
> Yeah, this is a problem. However, there's a much more serious problem
> coming up, which is the module system:
> 
> http://mail.openjdk.java.net/pipermail/awt-dev/2015-February/008924.html
> 
> I'd love your comments on it. I hope it doesn't kill Cacio...

Reading that thread it seems the most viable solution is to set a
precedent for a module that uses java.desktop (and you becoming an
official OpenJDK module). Then, java.desktop only exports to your new
module. After all, the point of the module system is not to kill openjdk
projects... Why don't you try an propose this as a JEP?

I'm pretty sure there will be others that need migration paths like you
do.

Cheers,
Severin