A DESCRIPTION OF THE PROBLEM :
Provided `DateTimeFormatter` are all rather lenient. Say that I want to have something specific. I can use:

DateTimeFormatter strictDateTimeFormatter = new DateTimeFormatterBuilder()
                .parseStrict()
                .parseCaseSensitive()
                .appendValue(YEAR, 4, 10, SignStyle.EXCEEDS_PAD)
                .appendLiteral('-')
                .appendValue(MONTH_OF_YEAR, 2)
                .appendLiteral('-')
                .appendValue(DAY_OF_MONTH, 2)

                .appendLiteral('T')

                .appendValue(HOUR_OF_DAY, 2)
                .appendLiteral(':')
                .appendValue(MINUTE_OF_HOUR, 2)
                .appendLiteral(':')
                .appendValue(SECOND_OF_MINUTE, 2)
                .appendFraction(MILLI_OF_SECOND, 3, 3, true)
                .appendOffsetId()
                .optionalStart()
                .appendLiteral('[')
                .parseCaseSensitive()
                .appendZoneRegionId()
                .appendLiteral(']')
                .toFormatter();

mostly it's just inlined formatters into this big one, removing all optionality, case insensitivity etc.

The interesting part is:

.appendFraction(MILLI_OF_SECOND, 3, 3, true)

so I want 3 digit millis, but with leading decimal separator. It works, but problem is, that it should not accept 1 or 2 digit fractions in input, only 3 digits should succeed. However in following samples:

    "2017-12-01T14:12:45.1+01:00[Europe/Prague]"
    "2017-12-01T14:12:45.12+01:00[Europe/Prague]"
    "2017-12-01T14:12:45.123+01:00[Europe/Prague]"
    "2017-12-01T14:12:45.1234+01:00[Europe/Prague]"

.12 and .123 succeeds.

I'm pretty convinced, that this is a bug, and in following code:

if (digit < 0) {
                    if (pos < minEndPos) {
                        return ~position; // need at least min width digits
                    }
                    pos--;
                    break;
                }

the `pos--;` should happen before `if` or if should also test equality.

STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
see description.

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
If I specify I want min 3 digits and max 3 digits, then only 3 digit frations should be OK.
ACTUAL -
2 and 3 digit fractions are accepted.

CUSTOMER SUBMITTED WORKAROUND :
probably only not using constructs leading to this method.

FREQUENCY : always