[pypy-issue] [issue978] metainterp bug

[Carl Friedrich Bolz]

Carl Friedrich Bolz <cfbolz at gmx.de> added the comment:

Ok, the strange value is of course stored to the fail_boxes_ptr in
grab_frame_values. I don't know the backend well enough to sanely continue, I
have a screen (14694.pts-1.tannit) running on tannit that shows the issue: just
start gdb like this:

~/converge/compiler $ gdb --args ../vm/converge-bug-lldebug-O0
/home/cfbolz/converge/compiler/convergec -I /home/cfbolz/converge/lib/Stdlib -o
Compiler/Code_Gen.cvb Compiler/Code_Gen.cv

The following break point makes gdb stop before the crash in
ResumeDataDirectReader_consume_vable_info
break jit_metainterp_resume.c:825 if num_entered == 4989

The following watch point shows the problematic write to fail_boxes_ptr
watch pypy_g_array_525.b->items[0] == 0x25f5830

________________________________________
PyPy bug tracker <tracker at bugs.pypy.org>
<https://bugs.pypy.org/issue978>
________________________________________