On Sat, Dec 18, 2021 at 07:43:55PM -0500, Eric Gallager wrote:
> On Fri, Dec 17, 2021 at 5:59 PM Marek Polacek via Gcc-patches
> <gcc-patches@gcc.gnu.org> wrote:
> >
> > This fixes
> >
> > gcc/cp/parser.c:4618:41: warning: narrowing conversion of '(char)(*(str +
> > ((sizetype)i)))' from 'char' to 'unsigned char' [-Wnarrowing]
> > 4618 | unsigned char s[3] = { '\'', str[i], '\'' };
> > | ~~~~~^
> >
> > Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
> >
>
> Hi, I thought that GCC was built with -Wno-narrowing; has that changed
> since I last checked?
No, I noticed that warning when compiling parser.i manually.
Here's a thread I found which led to -Wno-narrowing:
<https://gcc.gnu.org/pipermail/gcc-patches/2011-October/326369.html>.
> I'd support a move to officially switch from
> disabling -Wnarrowing to enabling it instead if that's possible and
> hasn't been done yet.
I don't know if the issues above have been fixed. Maybe they have; it's
been a while...
> Also the '(char)(*(str + ((sizetype)i)))' looks
> like some implementation details leaking; is there a bug against
> -Wnarrowing (or the diagnostics system in general) open about that?
Dunno.
> > gcc/cp/ChangeLog:
> >
> > * parser.c (make_char_string_pack): Add a cast to const unsigned
> > char *.
> > ---
> > gcc/cp/parser.c | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/gcc/cp/parser.c b/gcc/cp/parser.c
> > index 44eed7ea638..56232ab029f 100644
> > --- a/gcc/cp/parser.c
> > +++ b/gcc/cp/parser.c
> > @@ -4607,7 +4607,8 @@ make_char_string_pack (tree value)
> > {
> > tree charvec;
> > tree argpack = make_node (NONTYPE_ARGUMENT_PACK);
> > - const char *str = TREE_STRING_POINTER (value);
> > + const unsigned char *str
> > + = (const unsigned char *) TREE_STRING_POINTER (value);
> > int i, len = TREE_STRING_LENGTH (value) - 1;
> > tree argvec = make_tree_vec (1);
> >
> >
> > base-commit: d7ca2a79b82c6500ead6ab983d14c609e2124eee
> > --
> > 2.33.1
> >
>
Marek
