On 11/21/23 17:51, Jakub Jelinek wrote:
On Tue, Nov 21, 2023 at 11:19:56PM +0100, Jakub Jelinek wrote:
+ error_at (location, "%<static_assert%> message must be a string "
+ "literal or object with %<size%> and "
+ "%<data%> members");
Let's print the type of the message as well.
so add " while it has type %qT", TREE_TYPE (message) or something else?
Now in patch form (except for the removal of warning_at for now):
+ if (TREE_CODE (message) != STRING_CST
+ && !type_dependent_expression_p (message))
+ {
+ message_sz
+ = finish_class_member_access_expr (message,
+ get_identifier ("size"),
+ false, tf_none);
+ message_data
+ = finish_class_member_access_expr (message,
+ get_identifier ("data"),
+ false, tf_none);
+ if (message_sz == error_mark_node || message_data == error_mark_node)
+ {
+ error_at (location, "%<static_assert%> message must be a string "
+ "literal or object with %<size%> and "
+ "%<data%> members while it has type %qT",
Actually, let's go back to the previous message, but change the tf_nones
above to 'complain' so that we see those errors and then this
explanation. Likewise with the conversion checks later in the function.
+ TREE_TYPE (message));
+ return;
+ }
+ releasing_vec size_args, data_args;
+ message_sz = finish_call_expr (message_sz, &size_args, false, false,
+ tf_warning_or_error);
+ message_data = finish_call_expr (message_data, &data_args, false, false,
+ tf_warning_or_error);
Can use 'complain' instead of tf_warning_or_error here, too.
+ if (message_sz == error_mark_node || message_data == error_mark_node)
+ return;
+ if (tree s
+ = cp_get_callee_fndecl_nofold (extract_call_expr (message_sz)))
+ if (!DECL_DECLARED_CONSTEXPR_P (s))
+ warning_at (location, 0, "%qD used in %<static_assert%> message "
+ "is not %<constexpr%>", s);
I don't think we need this check, it should be covered by the later
constant-expression checks.
If the static_assert condition is true, we won't diagnose anything then.
clang++ there incorrectly errors, but I thought a warning could be useful
to users. Perhaps it could warn only if the condition is true?
I don't think the extra warning is that useful, especially with no flag
to suppress it; we specifically decided not to require the message to be
constant if the condition is true, and involving a non-constexpr
function is just one example of how it might not be constant.
Jason
