[kaffe] More mipsel jit3

[Timothy Stack]

> So far, I can get a pretty major application (Jetty) to nearly finish
> loading itself, but still end with a segfault. It runs an awful lot of
> code before it fails.

Do all of the test cases pass?

> The problem now looks like an invalid `load_offset_int' (which becomes
> a `lw a0,20[a0]'). I *think* it is invalid because a0 is loaded with
> memory that contains an instruction, not data, so $a0+20 isn't a valid
> address, obviously.

Need more context here...

> So, maybe, areas of data are being overwritten
> with jitted instructions. Or is this a garbage collection issue?

I just fixed the "gcstate" macro in the gdbinit file which prints out the 
GC related stuff for any address.  That might help out a bit...

> Also, I'm trying to cross-compile the xdebugging assembly, so I can
> at least have some way to set up an xdebug, but I get this error:
> 
> - -> {standard input}:54830: Error: Rest of line ignored. First ignored character is `0'.

What command is being executed here?  Is it choking on the C, the 
generated asm?

> My gcc is some craptastic customized version of gcc 3.0. Is this just
> out of date?

I've successfully built the xdebugging stuff on gcc 2.95 and 3.x, but 
other than that I don't know what to tell ya.

> Casey Marshall || rsdio at metastatic.org

tim