Kiyo Inaba wrote:
> Still it does not work, but I finally found where the bug is...
>
> According to the assembler listing of the method 'loadClass' (which
> is attached at the very end of this mail), the argument one (which is
> passed by register r1) is once stored in register r6, and has a copy
> in [r11, #-72]. But at adress 0x68d910, the saved area are contaminated
> by register r7, rather than r6. It is also true for the 'hidden'
> argument ('this' variable) passed by register r0, and stored in r5.
>
> Maybe some register spill portions should have bug, and I need extra
> time to fix it. The register allocation algorithm in kaffe is tough
> to debug ;-<
Thank you very, very much for all the ARM JIT fixes, Kiyo. I think the
global register workaround is OK for the release ;)
cheers,
dalibor topic
_______________________________________________
kaffe mailing list
kaffe@kaffe.org
http://kaffe.org/cgi-bin/mailman/listinfo/kaffe
