[kernel] Make scheduler run queue's prev be an id, not a pointer

This would lead to errors in GDB's j6threads when the previous thread
had already exited.

assets/debugging/jsix.elf-gdb.py

@@ -237,11 +237,7 @@ class GetThreadsCommand(gdb.Command):
             self.print_cpudata(cpu)
 
             previous = int(gdb.parse_and_eval(f"{runlist}.prev"))
-            if previous != 0:
+            print(f"            prev: {previous:x}")
-                tcb = f"((TCB*){previous:#x})"
-                thread = f"({tcb}->thread)"
-                koid = int(gdb.parse_and_eval(f"{thread}->m_obj_id"))
-                print(f"            prev: {koid:x}")
             print()
 
             for pri in range(8):
@@ -332,8 +328,13 @@ class DumpLogCommand(gdb.Command):
         end += self.base_addr
         while addr < end:
             entry = self.get_entry(addr)
+            if entry.bytes < 8:
+                print(f"Bad log header size: {entry.bytes}")
+                break
             addr += entry.bytes
-            area = self.areas[entry.area]
+            area = "??"
+            if entry.area < len(self.areas):
+                area = self.areas[entry.area]
             level = self.level_names[entry.severity]
             print(f"{area:>7}:{level:7}  {entry.message}")
 

src/kernel/scheduler.cpp

@@ -28,7 +28,7 @@ scheduler *scheduler::s_instance = nullptr;
 struct run_queue
 {
     tcb_node *current = nullptr;
-    tcb_node *prev = nullptr;
+    uint64_t prev = 0;
     tcb_list ready[scheduler::num_priorities];
     tcb_list blocked;
 
@@ -310,7 +310,7 @@ scheduler::schedule()
         return;
     }
 
-    queue.prev = queue.current;
+    queue.prev = queue.current->thread->obj_id();
     thread *next_thread = next->thread;
 
     cpu.thread = next_thread;