On 5/28/19 9:11 AM, Jakub Jelinek wrote:
On Tue, May 28, 2019 at 08:59:57AM -0400, Jason Merrill wrote:
On 5/27/19 5:20 PM, Jakub Jelinek wrote:
As the testcase shows, we are silently accepting writes into const
variables, because the type generic builtins don't have a prototype.
Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
trunk?
2019-05-27 Jakub Jelinek <ja...@redhat.com>
PR c/90628
* c-common.c (check_builtin_function_arguments)
<case BUILTIN_*_OVERFLOW>: Diagnose pointer to const qualified integer
as last argument.
* c-c++-common/builtin-arith-overflow-3.c: New test.
--- gcc/c-family/c-common.c.jj 2019-05-21 16:16:48.068973678 +0200
+++ gcc/c-family/c-common.c 2019-05-27 10:46:25.525968739 +0200
@@ -5995,6 +5995,13 @@ check_builtin_function_arguments (locati
"has pointer to boolean type", fndecl);
return false;
}
+ else if (TYPE_READONLY (TREE_TYPE (TREE_TYPE (args[2]))))
+ {
+ error_at (ARG_LOCATION (2), "argument 3 in call to function %qE "
+ "has pointer type to %<const%> qualified integer",
+ fndecl);
Is there a reason not to also print the type with %qT?
So like:
+ error_at (ARG_LOCATION (2), "argument 3 in call to function %qE "
+ "has pointer type to %<const%> qualified integer "
+ "(%qT)", fndecl, TREE_TYPE (args[2]));
or some other wording?
I didn't want to say
"argument 3 in call to function %qE has type %qT" because then
users wouldn't know what the actual problem is.
Sure. Or "has pointer to %<const%> type (%qT)" as a terser alternative.
OK either way.
Jason
