bug#67141: 30.0.50; Missing element in the backtrace

[Mattias Engdegård]

> B. ELisp primitives with their own bytecode.

Some time ago I made a feeble attempt to remedy this problem but only got as 
far as dealing with a few byte ops that open-coded their respective Lisp 
operations (car, aref, and a few more).

The straightforward solution would be to access the PC and SP from a bytecode 
frame when a backtrace is required but it will be a bit messy and I feared 
there might be performance consequences, and it's possible that the stack frame 
format would need changing as well. Still worth a try, though.

(Actually I'd like the option to include a byte code offset to source location 
map in the compiled bytecode as well. It would be the next logical step. Not 
sure how best to store it but a byte string is probably what I'd use, possibly 
loaded dynamically like doc strings.)

Not sure if the native compiler would prefer building on that kind of source 
location structure but why not.