Re: [PATCH] scriptreplay: fscanf portability fix

[Karel Zak <kzak@redhat.com>]

On Fri, Jul 22, 2011 at 10:15:40PM +0200, Sami Kerola wrote:
> This is a fix to an issue with llvm clang which I reported
> earlier.
> 
> http://www.spinics.net/lists/util-linux-ng/msg04331.html
> 
> When compiling with clang fscanf format string %[\n] will not
> work. 

 Sure, read scanf man page. The %[] format expects string +
 null byte. The bug is in the scriptreplay.c code where 
 we don't have room for the null byte :-( 
 
 I really don't think that this a llvm bug.

> --- a/term-utils/scriptreplay.c
> +++ b/term-utils/scriptreplay.c
> @@ -203,7 +203,7 @@ main(int argc, char *argv[])
>  		size_t blk;
>  		char nl;
        ^^^^^^^^
>  
> -		if ((fscanf(tfile, "%lf %zd%[\n]\n", &delay, &blk, &nl) != 3) ||
> +		if ((fscanf(tfile, "%lf %zd%c\n", &delay, &blk, &nl) != 3) ||
>  							(nl != '\n')) {

 Yep, it seems usable.

 Anyway, it's usually better to use fgets() + sscanf() to read and
 parse lines from files. The fscanf() could be tricky if your format
 depends on whitespaces.

 Applied, thanks.

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com