Failing jdk_time test on MacOS x64 JDK18+36 release triage

[Haroon-Khel]

java/time/test/java/time/format/TestZoneTextPrinterParser.java.TestZoneTextPrinterParser from jdk_time in the extended openjdk test suite is failing on MacOS

[2022-03-23T12:40:00.263Z] java.time.format.DateTimeParseException: Text '2021-11-07 01:30:00.000 EST' could not be parsed at index 24
[2022-03-23T12:40:00.263Z] 	at java.base/java.time.format.DateTimeFormatter.parseResolved0(DateTimeFormatter.java:2056)
[2022-03-23T12:40:00.263Z] 	at java.base/java.time.format.DateTimeFormatter.parse(DateTimeFormatter.java:1958)

Test job link https://ci.adoptopenjdk.net/job/Test_openjdk18_hs_extended.openjdk_x86-64_mac_testList_1/26/testReport/
Trss link https://trss.adoptium.net/output/test?id=623b273c72761a4c5cf83036

I reran the job on test-macstadium-macos1014-x64-4, https://ci.adoptopenjdk.net/job/Grinder/3917/console, it still failed. I am rerunning it on a build machine to double check https://ci.adoptopenjdk.net/job/Grinder/3932/console

[Haroon-Khel]

Many upstream bugs concerning timeout failures can be found, but none with this error

[smlambert]

Adding the jtr file (from the test output archive) for reference: TestZoneTextPrinterParser.jtr.txt

[smlambert]

Also noting that this test consistently passes in jdk17 (https://ci.adoptopenjdk.net/job/Test_openjdk17_hs_extended.openjdk_x86-64_mac_testList_1/38/testReport/java_time_test_java_time_format_TestZoneTextPrinterParser/).

Run against Oracle's jdk18 build here: https://ci.adoptopenjdk.net/job/Grinder/3937/
also fails in same way. We will need to dig into test code and understand what is being tested and why its failing. Likely will not block as also failing against Oracle build.

[smlambert]

We should also double-check that there is no 'test skew'... that the test material we are using, matches up with the source build tag being tested. (believe this to be the case, but will confirm).

[smlambert]

@sophia-guo - you mentioned this test case, DateTimeParseException is also failing on win32?

[sophia-guo]

Not the same tests, the underlying failure is the same java.time.format.DateTimeParseException. Test : test.java.time.format.TestUnicodeExtension.java

[sophia-guo]

test.java.time.format.TestUnicodeExtension.java against azul on win32 got the same failure https://ci.adoptopenjdk.net/view/Test_grinder/job/Grinder/3962/consoleFull

[smlambert]

Ok, so we are going to treat it as 'non-blocking', but I'd like to figure out if this is a test case that needs to be improved, or a product problem that should be fixed (or skew/mismatch of test material).

[sophia-guo]

This should not be skew/mismatch of test material. Test builds are using commit 0f2113cee79b9645105b4753c7d7eacb83b872c2, which is exactly same as jdk-18+36

12:04:32       [echo] git clone --depth 1 -q -b jdk-18+36 https://github.com/adoptium/jdk18u.git openjdktemp
12:05:17       [exec] Note: switching to '0f2113cee79b9645105b4753c7d7eacb83b872c2'.

https://github.com/adoptium/jdk18u/releases/tag/jdk-18%2B36

[sophia-guo]

test.java.time.format.TestUnicodeExtension has not been updated for two years.

java/time/test/java/time/format/TestZoneTextPrinterParser.java did have updated recently with jdk source code update. Also it is related with ZonedDateTime parse

adoptium/jdk18u@a363b7b.

adoptium/jdk18u@064ee6a
Need to take a look at those changes.

or https://bugs.openjdk.java.net/browse/JDK-8177819

Also https://bugs.openjdk.java.net/browse/JDK-8177819 only delivered to jdk18, no backport to jdk17, which explains why tests passed on jdk17.

[sophia-guo]

@jiekang , @andrew-m-leonard Based on #3498 (comment) it might be helpful to open a JBS issue?

[jiekang]

@sophia-guo to clarify, you suspect changes from JDK-8177819 are causing the failure in 18, and since it's not backported to 17, it's still passing there. Is that right?

[sophia-guo]

Yes, correct. @jiekang

[jiekang]

Just noting, the test that fails in TestZoneTextPrinterParser didn't exist in 17, it was added by https://bugs.openjdk.java.net/browse/JDK-8177819 (openjdk/jdk@a363b7b#diff-d4fb442a77ecfe0a5c7ab7dbc30acdeb8aee1d001f4f0555b5e0d50bfd5c9047R261)

[andrew-m-leonard]

22:28:09  test test.java.time.format.TestZoneTextPrinterParser.test_roundTripAtOverlap("yyyy-MM-dd HH:mm:ss.SSS z", "2021-11-07 01:30:00.000 EST"): failure
22:28:09  java.time.format.DateTimeParseException: Text '2021-11-07 01:30:00.000 EST' could not be parsed at index 24
22:28:09  	at java.base/java.time.format.DateTimeFormatter.parseResolved0(DateTimeFormatter.java:2056)
22:28:09  	at java.base/java.time.format.DateTimeFormatter.parse(DateTimeFormatter.java:1958)
22:28:09  	at java.base/java.time.ZonedDateTime.parse(ZonedDateTime.java:600)
22:28:09  	at test.java.time.format.TestZoneTextPrinterParser.test_roundTripAtOverlap(TestZoneTextPrinterParser.java:266)

[andrew-m-leonard]

This is the same problem as https://bugs.openjdk.java.net/browse/JDK-8282023
needs same fix in this testcase here: https://github.com/openjdk/jdk18u/blob/1ab6f9395f784293deb37ea0186db59d1166ae40/test/jdk/java/time/test/java/time/format/TestZoneTextPrinterParser.java#L266

To recreate:

        String pattern = "yyyy-MM-dd HH:mm:ss.SSS z";
        String input = "2021-11-07 01:30:00.000 EST";

        var dtf = DateTimeFormatter.ofPattern(pattern);
        String parsed = dtf.format(ZonedDateTime.parse(input, dtf));

Exception in thread "main" java.time.format.DateTimeParseException: Text '2021-11-07 01:30:00.000 EST' could not be parsed at index 24
	at java.base/java.time.format.DateTimeFormatter.parseResolved0(DateTimeFormatter.java:2056)
	at java.base/java.time.format.DateTimeFormatter.parse(DateTimeFormatter.java:1958)
	at java.base/java.time.ZonedDateTime.parse(ZonedDateTime.java:600)
	at datetime.main(datetime.java:13)

Using Locale.ROOT fix:

        String pattern = "yyyy-MM-dd HH:mm:ss.SSS z";
        String input = "2021-11-07 01:30:00.000 EST";

        var dtf = DateTimeFormatter.ofPattern(pattern, Locale.ROOT);
        String parsed = dtf.format(ZonedDateTime.parse(input, dtf));

Parsed = 2021-11-07 01:30:00.000 EST

[andrew-m-leonard]

Opened JBS bug: https://bugs.openjdk.java.net/browse/JDK-8283842

[sophia-guo]

@andrew-m-leonard #3494 has same issue, which is on win32. Noticed https://bugs.openjdk.java.net/browse/JDK-8283842 only mentioned Macos.

[jiekang]

@sophia-guo I added a comment on seeing it on win32.

[andrew-m-leonard]

Fixed upstream: openjdk/jdk@a41550b

[sophia-guo]

Any plan to back port to jdk18? @andrew-m-leonard

[andrew-m-leonard]

I've asked upstream..

[jiekang]

I don't expect this to be backported to 18, but I don't think it matters to us. @sophia-guo can this issue be closed?

[sophia-guo]

Yes, no need to backport to 18 any more. Close it.