trace: fix SIGSEGV after pop

[pmhahn]

Calling trace_pop_target() invalidates the linked list of target_stack_node_ts by free()ing the node and its p_target.

But p_stack_top may still points at that now freed node. Entering the debugger will crash as it will start from p_stack_top.

1. Explicitly invalidate p_target by setting it to NULL.
2. Explicitly pop the top node from p_stack_top before calling trace_pop_target().

Closes: #158

[rocky]

Looks, good to me. Thanks!

[pmhahn]

Looks, good to me. Thanks!

Actually it created the next bug: p_stack_top can be NULL and will now crash on p_stack_top->p_parent 😢

Core was generated by `/home/phahn/REPOS/LINUX/remake/make -C /home/phahn/GU/gu/KERNEL_brcma9_build/sr'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0  0x00005f5d548d868a in update_file_1 (file=0x5f5d883c9240, depth=2, p_call_stack=0x5f5d883e3b30) at src/remake.c:880

warning: Source file is more recent than executable.
880	      p_stack_top = p_stack_top->p_parent;
(gdb) l
875	  if (file->command_state != cs_finished)
876	    {
877	      DBF (DB_VERBOSE, _("Recipe of '%s' is being run.\n"));
878	      if ( file->tracing & BRK_AFTER_CMD || i_debugger_stepping )
879		  enter_debugger(p_call_stack, file, 0, DEBUG_BRKPT_AFTER_CMD);
880	      p_stack_top = p_stack_top->p_parent;
881	      trace_pop_target(p_call_stack);
882	      return 0;
883	    }
884	
(gdb) p p_stack_top 
$1 = (target_stack_node_t *) 0x0