cvs commit: src/usr.sbin/setfmac setfmac.c

[Yar Tikhiy]

On Sun, Jan 27, 2008 at 01:19:47AM +0000, John Birrell wrote:
> jb          2008-01-27 01:19:47 UTC
> 
>   FreeBSD src repository
> 
>   Modified files:
>     usr.sbin/setfmac     setfmac.c 
>   Log:
>   fts_pathlen is now a size_t rather than an int so a cast is needed.
>   I'm not sure why warn() and err() string formatted variables need
>   to be right-justified.
>   
>   Revision  Changes    Path
>   1.10      +6 -6      src/usr.sbin/setfmac/setfmac.c

Thank you for fixing this!

The format strings in question might just have a bug in them.
My guess is that the author of the format strings feared that
fts_path might not be NUL-terminated, so he explicitly specified
variable precision of fts_pathlen; but he forgot to left-justify
the field as well.

In fact, fts_path in a fresh FTSENT just returned by fts_read() is
always NUL-terminated.  (See fts(3).)  Tricks with fts_pathlen are
necessary only if referring to old, saved FTSENTs, but this is
dangerous and unportable anyway.  setfmac seems to refer to fresh
FTSENTs only, so the format can be changed safely from "%.*s" to
just "%s".  It will eliminate the fts_pathlen arguments to the
printf-like functions completely.

It was Robert Watson who committed the lines to setfmac.c, according
to cvs annotate, so I added him to Cc:.

Robert, do you think it's OK to simplify the format strings by
changing "%.*s" to "%s" along with removing the fts_pathlen arguments?
E.g.:

                case FTS_NS:
-                       err(1, "traversing %.*s", ftsent->fts_pathlen,
-                           ftsent->fts_path);
+                       err(1, "traversing %s", ftsent->fts_path);
                default:

Thanks!

Yar