Message ID | 20131118020746.GX16018@ringworld.MIT.EDU |
---|---|
State | New, archived |
Headers | show |
On Sun, 2013-11-17 at 21:07 -0500, Greg Price wrote: > [+linux-sparse and Chris] > > On Mon, Nov 18, 2013 at 01:33:49AM +0000, Al Viro wrote: > > On Sun, Nov 17, 2013 at 02:45:05PM -0800, Joe Perches wrote: > > > Yes. I think it's a defect in how sparse > > > treats string concatenation. > > > > > > That style [... with printk ...] is pretty common in the kernel sources. > > > > ... and it's perfectly fine, until somebody starts playing in nasal > > daemon country and do that in *macro* arguments. And a nasal daemon > > country it is - it's an undefined behaviour. See 6.10.3p11 in C99. > > And trying to define a semantics for that gets real ugly real fast. > > sparse matches gcc behaviour (I hope), but it warns about such abuses. > > It's a defect, all right - one being reported by sparse. > > Perhaps the following tweak to the error message would make this > subtlety clearer? Maybe, but this case isn't a macro. It's a function. Dunno if differentiating when it's a macro or a function is difficult though.
On Mon, Nov 18, 2013 at 12:15 AM, Joe Perches <joe@perches.com> wrote: > On Sun, 2013-11-17 at 21:07 -0500, Greg Price wrote: > > Maybe, but this case isn't a macro. It's a function. > Dunno if differentiating when it's a macro or a > function is difficult though. > > The case which was initially reported by sparse at http://www.spinics.net/lists/kernel/msg1636974.html was using pr_info (which is a macro) and not printk.
On Sun, Nov 17, 2013 at 06:15:32PM -0800, Joe Perches wrote: > > > sparse matches gcc behaviour (I hope), but it warns about such abuses. > > > It's a defect, all right - one being reported by sparse. > > > > Perhaps the following tweak to the error message would make this > > subtlety clearer? > > Maybe, but this case isn't a macro. It's a function. > Dunno if differentiating when it's a macro or a > function is difficult though. In which case? With printk() it's perfectly fine and sparse will not complain at all. With pr_info(), OTOH, we have #define pr_info(fmt, ...) \ printk(KERN_INFO pr_fmt(fmt), ##__VA_ARGS__) and when you start playing that kind of games with it, you get warnings.
On Sun, Nov 17, 2013 at 06:15:32PM -0800, Joe Perches wrote: > On Sun, 2013-11-17 at 21:07 -0500, Greg Price wrote: > > Perhaps the following tweak to the error message would make this > > subtlety clearer? > > Maybe, but this case isn't a macro. It's a function. > Dunno if differentiating when it's a macro or a > function is difficult though. Yeah, this error message is already only emitted for directives in macro arguments -- in this case, pr_info. It's in sparse's preprocessor code; the error arises when a directive is spotted while parsing a macro's arguments. By the time sparse (or an idealized C compiler) parses the arguments of a real function, the token stream is already the output of the preprocessor and any directives are gone. Cheers, Greg
diff --git a/pre-process.c b/pre-process.c index d521318..db58a97 100644 --- a/pre-process.c +++ b/pre-process.c @@ -204,7 +204,7 @@ static struct token *collect_arg(struct token *prev, int vararg, struct position if (next->pos.newline && match_op(next, '#')) { if (!next->pos.noexpand) { sparse_error(next->pos, - "directive in argument list"); + "directive in macro argument list"); preprocessor_line(stream, p); __free_token(next); /* Free the '#' token */ continue;