Stash the bytecode interpreter's internal structures when OCaml's run…

…time is processing a signal.

Not doing so would cause the garbage collector to silently move the
carpet from under our feet, thus causing a memory corruption. The
chance it might have happened in the past is extremely low, because
it would have required a memory shortage at the exact time a signal
was processed. Something might have changed in the runtime of OCaml
5.1 that makes this conjunction of events more likely to occur,
e.g., systematically performing a collection before or after
processing a signal.

Note that using Setup_for_caml_call rather than Setup_for_gc might
be sufficient here, since "accu" is almost certainly always
irrelevant here (function application). But since performance does
not matter on this path, better safe than sorry.

kernel/byterun/coq_interp.c

@@ -571,20 +571,26 @@ value coq_interprete
       /* We also check for signals */
 #if OCAML_VERSION >= 50000
       if (Caml_check_gc_interrupt(Caml_state)) {
+        Setup_for_gc;
         value res = caml_process_pending_actions_exn();
         Handle_potential_exception(res);
+        Restore_after_gc;
       }
 #elif OCAML_VERSION >= 41000
       if (caml_something_to_do) {
+        Setup_for_gc;
         value res = caml_process_pending_actions_exn();
         Handle_potential_exception(res);
+        Restore_after_gc;
       }
 #else
       if (caml_signals_are_pending) {
         /* If there's a Ctrl-C, we reset the vm */
         intnat sigint = caml_pending_signals[SIGINT];
         if (sigint) { coq_sp = coq_stack_high; }
+        Setup_for_gc;
         caml_process_pending_signals();
+        Restore_after_gc;
         if (sigint) {
           caml_failwith("Coq VM: Fatal error: SIGINT signal detected "
                         "but no exception was raised");