different execution time behaviour for sec or min lowest field

[dllx]

trafficstars

I'm experiencing different next execution time behaviour depending if the first field is seconds or minutes. I think that without any explicit kind of configuration, the seconds field should not behave differently than the minutes field. It's an unexpected inconsistency.

Given

        CronDefinition cronDefinition = CronDefinitionBuilder.defineCron()
                .withMinutes().and()
                .withHours().and()
                .withDayOfMonth().and()
                .withMonth().and()
                .instance();
        Cron cron = new CronParser(cronDefinition).parse("0,1,3/1 * * * *");
        ExecutionTime executionTime = ExecutionTime.forCron(cron);

I expect the following, and indeed that's what I observe:

        ZonedDateTime now0 = ZonedDateTime.of(2024, 2, 7, 13, 0, 7, 123456789, ZoneOffset.UTC);
        ZonedDateTime expected0 = ZonedDateTime.of(2024, 2, 7, 13, 1, 0, 0, ZoneOffset.UTC);
        ZonedDateTime nextTime0 = executionTime.nextExecution(now0).get();
        assertEquals(expected0, nextTime0);

        ZonedDateTime now1 = ZonedDateTime.of(2024, 2, 7, 13, 1, 7, 123456789, ZoneOffset.UTC);
        ZonedDateTime expected1 = ZonedDateTime.of(2024, 2, 7, 13, 3, 0, 0, ZoneOffset.UTC);
        ZonedDateTime nextTime1 = executionTime.nextExecution(now1).get();
        assertEquals(expected1, nextTime1);

        ZonedDateTime now2 = ZonedDateTime.of(2024, 2, 7, 13, 2, 7, 123456789, ZoneOffset.UTC);
        ZonedDateTime expected2 = ZonedDateTime.of(2024, 2, 7, 13, 3, 0, 0, ZoneOffset.UTC);
        ZonedDateTime nextTime2 = executionTime.nextExecution(now2).get();
        assertEquals(expected2, nextTime2);

On the other hand, if I add a seconds field into the definition like

        CronDefinition cronDefinition = CronDefinitionBuilder.defineCron()
                .withSeconds().and()
                .withMinutes().and()
                .withHours().and()
                .withDayOfMonth().and()
                .withMonth().and()
                .instance();
        Cron cron = new CronParser(cronDefinition).parse("0,1,3/1 * * * * *");
        ExecutionTime executionTime = ExecutionTime.forCron(cron);

I expect the following, but the result is different if the seconds field matches and the next second would also match the seconds field! In that case, the nanos field is not set to zero.

        ZonedDateTime now0 = ZonedDateTime.of(2024, 2, 7, 13, 46, 0, 999999999, ZoneOffset.UTC);
        ZonedDateTime expected0 = ZonedDateTime.of(2024, 2, 7, 13, 46, 1, 0, ZoneOffset.UTC);
        ZonedDateTime nextTime0 = executionTime.nextExecution(now0).get();
        System.out.println(nextTime0); // 2024-02-07T13:46:01.999999999Z
        //assertEquals(expected0, nextTime0);

        ZonedDateTime now1 = ZonedDateTime.of(2024, 2, 7, 13, 46, 1, 123456789, ZoneOffset.UTC);
        ZonedDateTime expected1 = ZonedDateTime.of(2024, 2, 7, 13, 46, 3, 0, ZoneOffset.UTC);
        ZonedDateTime nextTime1 = executionTime.nextExecution(now1).get();
        assertEquals(expected1, nextTime1);

        ZonedDateTime now2 = ZonedDateTime.of(2024, 2, 7, 13, 46, 2, 123456789, ZoneOffset.UTC);
        ZonedDateTime expected2 = ZonedDateTime.of(2024, 2, 7, 13, 46, 3, 0, ZoneOffset.UTC);
        ZonedDateTime nextTime2 = executionTime.nextExecution(now2).get();
        assertEquals(expected2, nextTime2);

By the way, to possibly counter the "strange" expression 3/1 used above, the behaviour can also be seen when using a plain * for the leftmost cron field. If it's "seconds", then the nano part of the nextExecution time is not set to zero, if it's "minutes", then the nano (and the second) part of the nextExecution time is indeed set to zero.

[jmrozanec]

@dllx thank you for reporting this and for contributing a fix! We will soon merge and release a new version! Thanks! 😎

[dllx]

There is also something similarly wrong in the ExecutionTime.isMatch method if the nanos are not zero. I will provide a better test.