uweigand commented on Issue #2357:
Hi @yurydelendik, just a quick question on the new RestoreFramePointer: shouldn't this reset the CFA to SP + initial_sp_offset, instead of whatever the current sp_offset is?
yurydelendik commented on Issue #2357:
new RestoreFramePointer: shouldn't this reset the CFA to SP + initial_sp_offset, instead of whatever the current sp_offset is?
IMHO it shall be SP + current sp_offset. In reality, it is initial_sp_offset = sp_offset, and we can add assert there.
At this moment, the input::UnwindCode values a chosen as subset to be able to produce Windows UNWIND_INFO and SystemV FDE. I suggest we will keep it this way until old existing backends implement unwind info (w/tests). We can refactor later.
yurydelendik edited a comment on Issue #2357:
new RestoreFramePointer: shouldn't this reset the CFA to SP + initial_sp_offset, instead of whatever the current sp_offset is?
IMHO it shall be SP + current sp_offset. In reality, it is initial_sp_offset = sp_offset, and we can add assert there.
At this moment, the input::UnwindCode values a chosen as superset to be able to produce Windows UNWIND_INFO and SystemV FDE. I suggest we will keep it this way until old existing backends implement unwind info (w/tests). We can refactor later.
yurydelendik edited a comment on Issue #2357:
new RestoreFramePointer: shouldn't this reset the CFA to SP + initial_sp_offset, instead of whatever the current sp_offset is?
IMHO it shall be SP + current sp_offset. In reality, it is initial_sp_offset = sp_offset, and we can add assert there.
At this moment, the input::UnwindCode values a chosen as subset/superset to be able to produce Windows UNWIND_INFO and SystemV FDE. I suggest we will keep it this way until old existing backends implement unwind info (w/tests). We can refactor later.
Last updated: Dec 23 2024 at 12:05 UTC