Re: -Wformat-overflow works strangely with struct members

[Martin Sebor <msebor at gmail dot com>]

On 07/07/2017 10:27 AM, Xi Ruoyao wrote:
> On 2017-07-07 09:35 -0600, Martin Sebor wrote:
> > On 07/07/2017 07:05 AM, Basin Ilya wrote:
> > > 	#include <stdio.h>
> > >
> > > 	extern struct {
> > > 		char s[142];
> > > 	} x;
> > >
> > > 	extern char s[142];
> > >
> > > 	int main(int argc, char *argv[]) {
> > > 		char buf[12];
> > > 		sprintf(buf, "%c%s", 'x', x.s); // why warning?
> > > 		sprintf(buf, "%s", x.s); // why no warning then?
> > > 		sprintf(buf, "%c%s", 'x', s); // why no warning then?
> > > 		return 0;
> > > 	}
> > With GCC 8 I get a warning for the second call to sprintf for
> > the same reason at both -Wformat-overflow levels.  I'm not sure
> > what prevents it in your case.  Please open a new bug for it and
> > either attach the translation unit for your test case (obtained
> > by preprocessing it with with -E) or remove the #include and
> > declare sprintf or use __builtin_sprintf.  (This should tell
> > us if the sprintf declaration in <stdio.h> is causing it.)
>
> No.  Using __builtin_sprintf still produces no warning with
> GCC 7.  I think it's a GCC 7 bug.

Ah, yes, I forgot.  GCC 7 doesn't handle format strings consisting
solely of a single %s directive because it transforms such calls
to strcpy very early on, before the warning pass has a chance to
see them.  It's fixed in GCC 8 under bug 77671.

Martin