Ben Burton wrote:
Doesn't a few days seem abnormally short notice for such a major NMU - not a bugfix but an entire new upstream release?
Kaffe has RC bugs which are open for 3 months (not counting your bug report about /usr/lib/jni) - all of them either include a patch or are easy to fix. Kaffe has been removed from testing because of this and now keeps other Java packages (like jikes, see #203054) from moving to testing.
The availability of the new upstream release was reported almost 2 months ago (with a list of bugs that are fixed by this upstream release) and I've sent Ean another mail about a month ago. How much longer should we wait?
The gij interpreter is quite advanced and for me works better than kaffe in almost all cases where I've done a comparison. There is no reason (in most cases) that an out-of-date kaffe should be a bottleneck for packages progressing into main.
People keep telling me that their favourite free JVM works quite well but whenever I try this particular JVM it fails running Ant. I have only been able to run Ant with Kaffe 1.1 - not with GIJ, not with SableVM. I would be very happy to not depend on Kaffe, so if you managed to run Ant with GIJ (or SableVM) please tell me how you did it.
Incidentally, several java packages could move from contrib into main if the maintainers could simply take the time to write their own Makefiles instead of relying on the default ant build system which is in contrib, e.g., #163168. It's a bit of work but it's certainly possible - see jython for an example.
I think changing the upstream build system is a bad idea, see e.g. #162426. You need to check each new upstream version for changes and sometimes don't notice missing files in JARs. I'd rather get Ant into main as I've written in my answer to #163168.
Stefan