Joshua Gatcomb <[EMAIL PROTECTED]> wrote:
> parrot trans_9.pasm
> ok 1
> ok 2
> parrot -j trans_9.pasm
> ok 1
> not ok 2
[ ... ]
> (gdb) add-symbol-file arithmetics_26.o 0
> add symbol table from file "arithmetics_26.o" at
> .text_addr = 0x0
> (y or n) y
> Reading symbols from arithmetics_26.o...done.
> (gdb) s
> jit_func () at arithmetics_26.pasm:1
> 1 set I0, 130
Good. So you can step through the JIT code.
Now comes the easy thing: find out, why the results differ:
- run the test through the plain core and trace it:
$ parrot -t t/op/trans_9.pasm
...
50 cosh N2, I1 - N2=1.543081, I1=1
- run it through JIT debugging:
...
# after the 2nd cosh
#
(gdb) p N2
$1 = 1.5430806348152437
(gdb) p I1
$2 = 1
You can turn on the register window of ddd too (or type "info float" and
"info registers" in gdb). Something must happen here that gives
differing results and segfaults.
leo
