[kaffe] JIT fails on NetBSD/sparc but not interpreter...

[Dalibor Topic]

Ciao Riccardo,

Riccardo wrote:
> Hello,
> 
> to be sure of my results, I did a double compile in two different 
> directories on NetBSD/sparc 1.6.2. One with interpreter and one with JIT. 
> BOth used jikes and ZIP (to test also that trouble didn't arise from 
> here).

Thanks a lot for testing!

> The intrp version passes internal self test and crunches over several 
> regression tests (I stopped after more than a dozen).

That means it passes some tests, right? That's something ;)

> Jit version still behaves as it did in my report some days ago: hangs in 
> one of the internal tests (boolean method)

I'm sorry, I'm rather unfamilar with both the jitter tests, and SPARC 
assembler. Would it be possible to poinpoint the change that broke the 
jit? That might help us find the real cause. Or if that doesn't work, to 
compare the build logs with some other sparc platform where the jit 
works for interesting differences in warnings?

That's all one can do without getting the hands dirty with the debugger, 
I fear, or going after the warnings one by one.

cheers,
dalibor topic