http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46705
--- Comment #3 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> 2010-12-11
15:54:38 UTC ---
If you look at the test case you will see that the '&' issue is outside of the
quotes in the format statement. Right away this told me we should not be
getting the warning at all since it is inside the "in_string" condition in
scanner.c: gfc_next_char_literal.
I found the problem here:
Index: io.c
===================================================================
--- io.c (revision 167624)
+++ io.c (working copy)
@@ -383,7 +383,7 @@ format_lex (void)
if (c == delim)
{
- c = next_char (1);
+ c = next_char (0);
if (c == '\0')
{
Just above this code in io.c, we have just detected the closing quote so we
should not call next_char with the in_string flag. I have regression tested
this and all our existing -Wampersand tests still work fine with this change.
Then we correctly get no warning with the first test case.
Now, if we do this:
1 FORMAT (' &
' )
END
We get a double warning. I think this is because in the scanner we may scan a
line more than once looking for matches. I am not decided yet what to do with
this bit.
