Hi, I wonder how __return_loc sections work. In general it would be nice to avoid direct output of assembler code in favour of adding instructions to rtl stream which should be possible to insert the call and return. __return_loc seems harder, also I wonder if it makes unwind information possible?
Honza
