On 12/27/2012 01:44 AM, Christoph Hellwig wrote:
> On Wed, Dec 26, 2012 at 04:25:19PM -0600, Alex Elder wrote:
>> Nigel Tamplin reported getting a seg fault in xfsrestore when a path
>> name was too long.
>> Based on the surrounding code, I'm sure strerror(errno) was the
>> intended final argument to this call. This bug has been there
>> since the code was first committed.
> Seems like mlog needs an attribute marking it as printf-like.
> I'll see if I can come up with a patch for that.
That patch is easy. Fixing all the problems that points out
leads to something quite a bit bigger. I started on it but
have put it aside for now. In case anyone wants to take on
that challenge I've included the patch for activating the
[PATCH] xfsdump: annotate mlog() as printf-like
The mlog() function is basically a printf() wrapper. Take advantage
of the gcc extension that will verify compatibility between the
format argument and the arguments that follow.
Define a __printf() macro like Linux does so the declaration line
doesn't get too long.
Signed-off-by: Alex Elder <elder@xxxxxxxxxxx>
common/mlog.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
@@ -109,7 +109,8 @@ void mlog_override_level( intgen_t level
/* vprintf-based message format
-extern void mlog( intgen_t level, char *fmt, ... );
+#define __printf(a, b) __attribute__((format(printf, a, b)))
+extern void __printf(2, 3) mlog( intgen_t level, char *fmt, ... );
extern void mlog_va( intgen_t levelarg, char *fmt, va_list args );
#define mlog_exit( e, r ) _mlog_exit( __FILE__, __LINE__, (e), (r) )
extern int _mlog_exit( const char *file, int line, int exit_code, rv_t