Comment 55 for bug 488354

Created an attachment (id=438688)
Patch v3

D'oh, the problem was so obvious and so problematic that I wonder how it could work without crashes on my builds, or with Mark's m-c builds O_o
The problem was that the stack buffer was too small in the paramCount == 1 case, and invoke_copy_to_stack would copy outside the buffer boundaries (it would write *before* the buffer, where the saved pc is for function return, in non optimized builds at least). The added assertion would catch this if this ever happens again.
Mark, can you check this patch ? I'll also try it on my end.