Message ID | 20101209031007.GF25904@nightcrawler |
---|---|
State | New |
Headers | show |
On 12/08/2010 10:10 PM, Nathan Froyd wrote: > The patch below fixes PR 46852: before we inspect the type that we just > parsed for a class/struct/union definition, we should make sure that > it's actually a type and not something bogus from a parse error. > > Tested on x86_64-unknown-linux-gnu. OK to commit? Yes. Fixes like this qualify as obvious. Jason
diff --git a/gcc/cp/parser.c b/gcc/cp/parser.c index ab533f4..089361a 100644 --- a/gcc/cp/parser.c +++ b/gcc/cp/parser.c @@ -16937,7 +16937,9 @@ cp_parser_class_specifier (cp_parser* parser) break; } - if (want_semicolon) + /* If we don't have a type, then something is very wrong and we + shouldn't try to do anything clever. */ + if (TYPE_P (type) && want_semicolon) { cp_token_position prev = cp_lexer_previous_token_position (parser->lexer); diff --git a/gcc/testsuite/g++.dg/pr46852.C b/gcc/testsuite/g++.dg/pr46852.C new file mode 100644 index 0000000..2c9d8dd --- /dev/null +++ b/gcc/testsuite/g++.dg/pr46852.C @@ -0,0 +1,7 @@ +// PR c++/46852 +// { dg-do compile } + +template +< +class +{ // { dg-error "" }