On Wed, Feb 28, 2018 at 11:14:23PM -0500, Peter Eisentraut wrote:
> AFAICT, the issues addressed here either can't really happen without
> trying very hard, or would cause harmless output truncation. Still, it
> seems good to clean this up properly and not rely on made-up buffer size
> guesses that turn out to be wrong, even if we don't want to adopt the
> warning options by default.
Good idea.
> One issue that is of external interest is that I increase BGW_MAXLEN
> from 64 to 96. Apparently, the old value would cause the bgw_name of
> logical replication workers to be truncated in some circumstances. I
> have also seen truncated background worker names with third-party
> packages, so giving some more room here would be useful.
OK, no complains about that.
@@ -89,7 +89,7 @@ static Datum
build_pgstattuple_type(pgstattuple_type *stat, FunctionCallInfo fcinfo)
{
#define NCOLUMNS 9
-#define NCHARS 32
+#define NCHARS 314
So this one is caused by the output of %.2f...
Enabling them by default would generate some useless noise if the patch
is let as-is as a couple of them are not addressed. Please see the full
report attached. Is that intentional? I am using GCC 7.3 here.
interval.c: In function ‘AppendSeconds’:
interval.c:759:22: warning: ‘%0*d’ directive output between 1 and
2147483648 bytes may exceed minimum required size of 4095
[-Wformat-overflow=]
sprintf(cp, "%02d.%0*d", abs(sec), precision, (int) Abs(fsec));
pg_rusage.c:64:5: note: in expansion of macro ‘_’
_("CPU: user: %d.%02d s, system: %d.%02d s, elapsed: %d.%02d s"),
^
pg_rusage.c:63:2: note: ‘snprintf’ output between 51 and 108
bytes into a destination of size 100
snprintf(result, sizeof(result),
--
Michael