[libc-commits] [libc] [libc] Scanf shouldn't match just "0x" for hex int (PR #112440)

[Nick Desaulniers via libc-commits]

================
@@ -124,13 +124,24 @@ int convert_int(Reader *reader, const FormatSection &to_conv) {
 
       if (to_lower(cur_char) == 'x') {
         // This is a valid hex prefix.
+
+        is_number = false;
+        // A valid hex prefix is not necessarily a valid number. For the
+        // conversion to be valid it needs to use all of the characters it
+        // consumes. From the standard:
+        // 7.23.6.2 paragraph 9: "An input item is defined as the longest
+        // sequence of input characters which does not exceed any specified
+        // field width and which is, or is a prefix of, a matching input
+        // sequence."
+        // 7.23.6.2 paragraph 10: "If the input item is not a matching sequence,
+        // the execution of the directive fails: this condition is a matching
+        // failure"
         base = 16;
         if (max_width > 1) {
           --max_width;
           cur_char = reader->getc();
         } else {
-          write_int_with_length(0, to_conv);
-          return READ_OK;
+          return MATCHING_FAILURE;
----------------
nickdesaulniers wrote:

Consider making this an early return if you reverse the conditional.

https://github.com/llvm/llvm-project/pull/112440