[PATCH] -Wformat: fix nonsensical "wide character" message (PR c/84258)

[Jeff Law]

On 02/07/2018 06:53 PM, David Malcolm wrote:
> PR c/84258 reports that we issue this:
> 
>  warning: format is a wide character string [-Wformat=]
> 
> on this code:
> 
>   const unsigned char cuc[] = "%i";
>   sprintf(buf, (char *)cuc, 1);
> 
> despite the absence of wide characters.
> 
> This wording dates back 17.5 years to r36586:
> 
>  2000-09-24  Joseph S. Myers  <jsm28@cam.ac.uk>
> 
>       * c-common.c (check_format_info): Warn for a wide character string
>        used as a non-wide format argument.
> 
>        * gcc.dg/c90-printf-1.c: Add test for wide string format.
> 
> which handled this case:
> 
>   printf ((const char *)L"foo"); /* { dg-warning "wide" "wide string" } */
> 
> The code in question is currently just checking for something that isn't
> char_type_node.
> 
> digest_init converts the string literal to the type of the array, converting
> it from array of char to array of unsigned char, leading to the condition
> failing.
> 
> This patch fixes the nonsensical message by checking for the other string
> types that lex_string can emit - char16_array_type_node,
> char32_array_type_node, and wchar_array_type_node - and only referring
> to "wide character string" if it's one of them.  Otherwise it uses a new,
> more generic wording.
> 
> Successfully bootstrapped&regrtested on x86_64-pc-linux-gnu.
> 
> This isn't a regression, but is fairly low risk.
> 
> OK for trunk, either now, or for next stage 1?
> 
> gcc/c-family/ChangeLog:
> 	PR c/84258
> 	* c-format.c (struct format_check_results): Add field
> 	"number_non_char".
> 	(check_format_info): Initialize it, and warn if encountered.
> 	(check_format_arg): Distinguish between wide char and
> 	everything else when detecting arrays of non-char.
> 
> gcc/testsuite/ChangeLog:
> 	PR c/84258
> 	* c-c++-common/Wformat-pr84258.c: New test.
OK for the trunk.

Jeff