Actually, I just resurrected the project again. It was never "desurrected", but I haven't put much time on it lately.
But now I've continued with OpenJDK, and I have a non-hacky setup that can be built! A big thing actually! Basically I've copied UNIX/Windows implantation to my newly added AmigaOS parts of the build process, and will hopefully soon begin to implement them.
I haven't worked much on the AWT stuff, but I think I'll be able to reuse what I've done for OpenJDK.
In your blog I found a hint that you do this work on an NG Amiga:
"...since I actually now use JAmiga to compile Java code when I'm developing JAmiga on my X1000. We kind of almost have a Java implementation so we can build Java ourselves on the Amiga!"
Now I'm wondering if there is any chance that classic Amiga systems (OS 3.x / WarpOS) will also benefit from your work (68k port), or if it will be released for NG OS4.x-Amigas exclusively?
Would JAmiga run sufficiently fast on e.g. a CyberstormPPC with 68060 @ 50 mHz / PPC604e @ 200 mHz and 128 mB Ram? Or would it only run at decent speed with e.g. AmigaForever on a fast PC?
tl;dr: 68k won't happen from me, and I don't put much effort in trying to keep compatible to pre OS 4. But if someone manages to build stuff on my work that support 68k, I'm not reluctant to incorporate patches.
Now, long rant: In order to have Java, you need, 1) the virtual machine to execute bytecode, and 2) the JDK that actually builds up the Java standard.
A 68k port won't happen from me. Mostly because I don't have the time. Currently I'm using the Java virtual machine "jamvm" which support various platforms, but not 68k. The old JAmiga virtual machine can run on 68k, as can the really old GNU Kaffe VM (available on aminet). These do however only support Java 1.4, using GNU classpath. When I tried updating the old JAmiga VM it lacked a few functions to support newer Java versions, which is why I ported jamvm.
As for my OpenJDK porting efforts, these use the Amiga API, so in theory it'd be possible to use that as a base for an OS 3.x implementation. I do however use new stuff in the AOS 4.1 API, and I really don't bother to check whether it is compatible to OS 3.
Regarding jamvm, my port supports Amiga libraries, i.e. no .so-stuff, and I aim to keep it that way. So, one possibility is to add 68k support to jamvm. Can't remember if jamvm has a clean C implementation. I know that OpenJDK's VM Hotspot has a "zero" implementation in C++ that could possibly be compiled for 68k. But I really doubt it's an easy port, and I really doubt even a 060 would be able to run at decent speed.
Might be possible to take the 68k code from JAmiga, and add to jamvm though. (Like the comments in the JAmiga code suggests, this has already been done the other way, taking PPC code from jamvm to JAmiga.)
(Edit: changed url to my GitHub, with the code I'm working with now, instead of the official jamvm repo.)
When you say "68k" do you really mean OS3? i.e. It uses functions that OS4 has but OS3 does not?
Perhaps both. If someone has patches that add 68k support for jamvm (like a CallNative.S linked in the post above), I don't mind adding it to my repo. Also, if someone has patches that makes my OpenJDK code compilable for OS 3 (or MorphOS/Aros, for that matter), I don't mind adding that aswell.
there's still some assembler code needed when doing JNI-calls
Ah, that makes sense. Vaguely similar: I seem to recall I had to add some 68k machine code(!) to PortablE, to handle hook calls in AmigaOS3. (No, I didn't write the machine code myself...)