vsapsai added inline comments.
================
Comment at: Parse/ParseTemplate.cpp:492
+ // Is there just a typo in the input code? ('typedef' instead of 'typename')
+ if (Tok.is(tok::kw_typedef)) {
+ Diag(Tok.getLocation(), diag::err_expected_template_parameter);
----------------
How does it work when you have `typedef` for the first template parameter?
================
Comment at: Parser/typedef-instead-of-typename-typo.hpp:3
+
+template <typename A, typedef B> struct Foo {
+ a
----------------
Maybe put this test in clang/test/FixIt ?
Also please check what file extensions are used for testing templates. .hpp
reflects real-life usage but most tests are .cpp. Or maybe I wasn't paying
attention.
================
Comment at: Parser/typedef-instead-of-typename-typo.hpp:5
+ a
+}; // CHECK: expected-error{{expected template parameter}} \
+// CHECK: expected-note{{Did you mean to use 'typename'?}} \
----------------
It is a little bit confusing to what lines the messages would be attributed to.
Need to check locally because not sure I interpret all those backslashes the
same way lit does.
Also idea for the test. To check that the fix-it was applied properly you can
add a member like `B b;` and it shouldn't trigger any errors.
================
Comment at: clang/Basic/DiagnosticParseKinds.td:1167
+def note_meant_to_use_typename : Note<
+ "Did you mean to use 'typename'?">;
}
----------------
Looks like other diagnostic messages "did you mean to use …" have lowercase "d"
in "did". Though I haven't checked how it looks in various situations.
https://reviews.llvm.org/D42170
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
