Bug 138303 – 2 date tests fail on 2014.11.01 - the last day of daylight saving

• Component: JavaScriptCore
• Status: NEW

[Csaba Osztrogonác]

Maybe it is related to bug130967 somehow.

- Nov 01 03:43 - r175450 - 0 failures:
https://build.webkit.org/builders/Apple%20Mavericks%20Release%20WK1%20%28Tests%29/builds/9652
- Nov 01 11:27 - r175453 - 12 failures:
https://build.webkit.org/builders/Apple%20Mavericks%20Release%20WK1%20%28Tests%29/builds/9653
...
- Nov 02 00:48 - r175460 - 12 failures:
https://build.webkit.org/builders/Apple%20Mavericks%20Release%20WK1%20%28Tests%29/builds/9661
- Nov 02 09:08 - r175462 - 0 failures:
https://build.webkit.org/builders/Apple%20Mavericks%20Release%20WK1%20%28Tests%29/builds/9662

[Csaba Osztrogonác]

Only 2 different test ( with all configuration 2*6 ):
- mozilla-tests.yaml/ecma/Date/15.9.5.14.js.mozilla
new Date(1414918204891)).getHours() = 1 FAILED! expected: 0

- mozilla-tests.yaml/ecma/Date/15.9.5.34-1.js.mozilla
TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getTime() = 1420015811536 FAILED! expected: 1420098611536
TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.valueOf() = 1420015811536 FAILED! expected: 1420098611536
TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getUTCFullYear() = 2014 FAILED! expected: 2015
TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getUTCMonth() = 11 FAILED! expected: 0
TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getUTCDate() = 31 FAILED! expected: 1
TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getUTCDay() = 3 FAILED! expected: 4
TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getUTCHours() = 8 FAILED! expected: 7
TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getHours() = 0 FAILED! expected: 23

[Radar WebKit Bug Importer]

<rdar://problem/18859472>

[Byungseon(Sun) Shin]

This year's PDT ends at Sun, Nov 2, 2:00 AM.
So, the error time boundary seems a little bit mis-matched with DST.
But I will double check it by reverting the changes locally.

[Byungseon(Sun) Shin]

@Csaba, could you please specify timezone information?
In PST(PDT) timezone, it works same with/without patch of bug130967.

[Csaba Osztrogonác]

mozilla-tests.yaml/ecma/Date/15.9.5.34-1.js.mozilla failed again on all bots:

Apple Mavericks bot:
Mar 08 14:51 	181246 	success 	#12794 	Build successful
Mar 08 10:24 	181245 	failure 	#12793 	Failed 6 jsc tests failed
Mar 08 00:21 	181222 	failure 	#12792 	Failed 6 jsc tests failed
Mar 07 21:18 	181220 	failure 	#12791 	Failed 6 jsc tests failed
Mar 07 19:53 	181219 	success 	#12790 	Build successful

Apple Yosemite bot:
Mar 08 14:50 	181246 	success 	#161 	Build successful
Mar 08 10:15 	181245 	failure 	#160 	Failed 6 jsc tests failed
Mar 08 00:18 	181222 	failure 	#159 	Failed 6 jsc tests failed
Mar 07 21:14 	181220 	failure 	#158 	Failed 6 jsc tests failed
Mar 07 19:50 	181219 	success 	#157 	Build successful

EFL bot:
Mar 08 14:33 	181246 	failure 	#20280 	success
Mar 08 10:03 	181245 	failure 	#20279 	Failed 4 jsc tests failed
Mar 08 03:28 	181242 	failure 	#20278 	Failed 8 jsc tests failed
Mar 08 00:11 	181222 	failure 	#20277 	Failed 4 jsc tests failed
Mar 07 22:52 	181221 	failure 	#20276 	Failed 4 jsc tests failed
Mar 07 20:58 	181220 	failure 	#20275 	Failed 4 jsc tests failed 
Mar 07 19:38 	181219 	failure 	#20274 	success

...

mozilla-tests.yaml/ecma/Date/15.9.5.34-1.js.mozilla-llint: TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getTime() = 1451586369509 FAILED! expected: 1451582769509
mozilla-tests.yaml/ecma/Date/15.9.5.34-1.js.mozilla-llint: TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.valueOf() = 1451586369509 FAILED! expected: 1451582769509
mozilla-tests.yaml/ecma/Date/15.9.5.34-1.js.mozilla-llint: TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getUTCHours() = 18 FAILED! expected: 17
mozilla-tests.yaml/ecma/Date/15.9.5.34-1.js.mozilla-llint: TDATE = new Date(now); (TDATE).setMonth(11,31); TDATE.getHours() = 10 FAILED! expected: 9

[Csaba Osztrogonác]

similar bug: bug139123

[Byungseon(Sun) Shin]

Just for sharing, I have verified that this issue does not related with bug130967 since event after I manually reverted the patch, same issue still existed.

I am looking into this issue what make this error regardless of the relationship.

[Byungseon(Sun) Shin]

This issue re-introduced from https://bugs.webkit.org/show_bug.cgi?id=136363 which was originally fixed by https://bugs.webkit.org/show_bug.cgi?id=71371.

Mozilla also has same issue: <https://bugzilla.mozilla.org/show_bug.cgi?id=610185>

We need to re-block those test cases to prevent another investigation in DST period.

[Byungseon(Sun) Shin]

Created attachment 249195 [details]
Patch

[Byungseon(Sun) Shin]

Created attachment 249196 [details]
Patch

[Byungseon(Sun) Shin]

Created attachment 249197 [details]
Patch

[Csaba Osztrogonác]

Comment on attachment 249197 [details]
Patch

Skipping tests doesn't fix the issue. And run-javascriptcore-tests run stress tests by default, which ignores this kind of skipping.

[Byungseon(Sun) Shin]

If so, what is your purpose to create this issue? Do we need to fix the mozilla date test cases to adapt DST changing? It fails on Mozilla also.

[Csaba Osztrogonác]

(In reply to comment #13)
> If so, what is your purpose to create this issue? Do we need to fix the
> mozilla date test cases to adapt DST changing? It fails on Mozilla also.

I have 2 problems with the proposed patch:
- Adding a test to @testsToSkip list doesn't skip it in the real life,
because run-javascriptcore-tests run stress tests by default with the
run-jsc-stres-tests harness. (Can be orverriden manually with --no-jsc-stress)
This list is for the old mozilla test harnes, which isn't used by default.
If we really want to skip a test, it should be done in
Source/JavaScriptCore/tests/mozilla/mozilla-tests.yaml too.
- I don't think if we should skip a full test just because one test
case in it fails on 1-2 days in a year.

15.9.5.14.js: (only the "now" case fails only when DST changes)
    addTestCase( now );
    addTestCase( TIME_YEAR_0 );
    addTestCase( TIME_1970 );
    addTestCase( TIME_1900 );
    addTestCase( TIME_2000 );
    addTestCase( UTC_FEB_29_2000 );
    addTestCase( UTC_JAN_1_2005 );

I think we should skip only the "now" case it this test after adding
DST specific tests too and adding crash-only tests for the "now" case.
See bug139123 too.

[Byungseon(Sun) Shin]

Now I can understand your intention. Thanks for the explanation.

[Csaba Osztrogonác]

Still valid bug, bots are red again due to PST->PDT change.

mozilla-tests.yaml/ecma/Date/15.9.5.14.js.mozilla-baseline: 15.9.5.14 Date.prototype.getHours()
mozilla-tests.yaml/ecma/Date/15.9.5.14.js.mozilla-baseline: (new Date(1478388956518)).getHours() = 16 FAILED! expected: 15
mozilla-tests.yaml/ecma/Date/15.9.5.14.js.mozilla-baseline: (new Date(1478392556518)).getHours() = 17 FAILED! expected: 16
mozilla-tests.yaml/ecma/Date/15.9.5.14.js.mozilla-baseline: (new Date(1478396156518)).getHours() = 18 FAILED! expected: 17
mozilla-tests.yaml/ecma/Date/15.9.5.14.js.mozilla-baseline: (new Date(1478399756518)).getHours() = 19 FAILED! expected: 18
mozilla-tests.yaml/ecma/Date/15.9.5.14.js.mozilla-baseline: (new Date(1478403356518)).getHours() = 20 FAILED! expected: 19
mozilla-tests.yaml/ecma/Date/15.9.5.14.js.mozilla-baseline: (new Date(1478406956518)).getHours() = 21 FAILED! expected: 20