On Wed, 20 Jul 2016, Daniel Bristot de Oliveira wrote:
> I am having the following warning while compiling cyclictest on
> Federa 24 (GCC 6.1.1 20160621):
>
> --------------------------%<--------------------------------------------
> [root@f24desk rt-tests]# make
> cc -D VERSION=1.1 -c src/cyclictest/cyclictest.c -Wall -Wno-nonnull -O2 -DNUMA -DHAVE_PARSE_CPUSTRING_ALL -D_GNU_SOURCE -Isrc/include -o bld/cyclictest.o
> src/cyclictest/cyclictest.c: In function ‘timerthread’:
> src/cyclictest/cyclictest.c:427:30: warning: ‘*((void *)&stop+8)’ may be used uninitialized in this function [-Wmaybe-uninitialized]
> diff += ((int) t1.tv_nsec - (int) t2.tv_nsec) / 1000;
> ^~~~~~~~~~~~~~~~
> src/cyclictest/cyclictest.c:980:39: note: ‘*((void *)&stop+8)’ was declared here
> struct timespec now, next, interval, stop;
> ^~~~
> src/cyclictest/cyclictest.c:426:54: warning: ‘stop.tv_sec’ may be used uninitialized in this function [-Wmaybe-uninitialized]
> diff = USEC_PER_SEC * (long long)((int) t1.tv_sec - (int) t2.tv_sec);
> ^~~~~~~~~~~~~~~
> src/cyclictest/cyclictest.c:980:39: note: ‘stop.tv_sec’ was declared here
> struct timespec now, next, interval, stop;
> ^~~~
> -------------------------->%--------------------------------------------
>
> This is a false positive, but rather than workaround it using GCC pragma
> option, I think it is better to use stop variable itself to define
> whether it should be used of not.
>
> Cc: John Kacur <jkacur@xxxxxxxxxx>
> Cc: Clark Williams <williams@xxxxxxxxxx>
> Signed-off-by: Daniel Bristot de Oliveira <bristot@xxxxxxxxxx>
> ---
> src/cyclictest/cyclictest.c | 13 ++++++++-----
> 1 file changed, 8 insertions(+), 5 deletions(-)
>
> diff --git a/src/cyclictest/cyclictest.c b/src/cyclictest/cyclictest.c
> index 5e23fc5..ad11b1b 100644
> --- a/src/cyclictest/cyclictest.c
> +++ b/src/cyclictest/cyclictest.c
> @@ -977,7 +977,8 @@ static void *timerthread(void *param)
> struct sigevent sigev;
> sigset_t sigset;
> timer_t timer;
> - struct timespec now, next, interval, stop;
> + struct timespec now, next, interval;
> + struct timespec *stop = NULL;
> struct itimerval itimer;
> struct itimerspec tspec;
> struct thread_stat *stat = par->stats;
> @@ -1066,9 +1067,11 @@ static void *timerthread(void *param)
> tsnorm(&next);
>
> if (duration) {
> - memset(&stop, 0, sizeof(stop)); /* grrr */
> - stop = now;
> - stop.tv_sec += duration;
> + stop = malloc(sizeof(struct timespec));
> + if (!stop)
> + fatal("error allocating space for duration option\n");
> + stop->tv_sec = now.tv_sec + duration;
> + stop->tv_nsec = now.tv_nsec;
> }
> if (par->mode == MODE_CYCLIC) {
> if (par->timermode == TIMER_ABSTIME)
> @@ -1185,7 +1188,7 @@ static void *timerthread(void *param)
> }
>
>
> - if (duration && (calcdiff(now, stop) >= 0))
> + if (stop && (calcdiff(now, *stop) >= 0))
> shutdown++;
>
> if (!stopped && tracelimit && (diff > tracelimit)) {
> --
> 2.7.4
>
I'm not sure about this one. Do we really need to change our code for a
false positive, that is really a compiler regression? And then we're
adding a malloc to a rt program for that? I'd rather ignore the warning if
we can't do this in a simpler manner.
Thanks
John
