Re: [patch] [python] Use TRY_CATCH in some functions.

[Phil Muldoon <pmuldoon@redhat.com>]

Tom Tromey <tromey@redhat.com> writes:

> Phil> @@ -624,20 +625,31 @@ typy_lookup_type (struct demangle_component *demangled,
> [...]
>
> Phil> +      if (except.reason < 0)
> Phil> +	{
> Phil> +	  PyErr_Format (except.reason == RETURN_QUIT
> Phil> +			? PyExc_KeyboardInterrupt : PyExc_RuntimeError,
> Phil> +			"%s", except.message);
> Phil> +	  return NULL;
>
> Why not GDB_PY_HANDLE_EXCEPTION here?
> (And also fix up typy_lookup_typename to do the same.)

Because typy_lookup_typename and typy_lookup_type are helper functions
that return struct type, while GDB_PY_HANDLE_EXCEPTION returns a
PyObject (from gdbpy_convert_exception).  Should we have a macro that
builds an exception without returning one? We should, I think.  But I was not
going to do that in this patch context.

> Phil> @@ -990,8 +1002,16 @@ typy_richcompare (PyObject *self, PyObject *other, int op)
> [...]
> Phil> +	{
> Phil> +	  bcache_xfree (cache);
> Phil> +	  VEC_free (type_equality_entry_d, worklist);
> Phil> +	  GDB_PY_HANDLE_EXCEPTION (except);
>
> I think the freeing should be hoisted above the exception check, to
> consolidate the code.  With the current patch the code is duplicated on
> both branches.

I amended the logic to look like this:

@@ -1007,11 +1007,13 @@
 	{
 	  result = check_types_worklist (&worklist, cache);
 	}
-      if (except.reason < 0)
-	result = Py_NE;
-
+      /* check_types_worklist calls several nested Python helper
+	 functions, some of which can raise a GDB Exception, so we
+	 just check and convert here.  If there is a GDB exception, a
+	 comparison is not capable (or trusted), so exit.  */
       bcache_xfree (cache);
       VEC_free (type_equality_entry_d, worklist);
+      GDB_PY_HANDLE_EXCEPTION (except);
     }
 
   if (op == result)

Cheers,

Phil