Failing random test seed/culture: en-IE with seed 0x9a2b7430d6d33f0d

[paulirwin]

Is there an existing issue for this?

• I have searched the existing issues

Describe the bug

Found a random failing test seed/culture pair. Set the following in a lucene.testsettings.json file in your repo root (or somewhere between the test executable and the root of your drive):

{
  "tests": {
     "seed": "0x9a2b7430d6d33f0d",
     "culture": "en-IE"
  }
}

This causes the following tests to fail on these frameworks:

Lucene.Net.Index.TestIndexReaderClose.TestCloseUnderException() (net5.0, net48, net472)

Message:

System.ArgumentException : The supplied DateTime represents an invalid time. For example, when the clock is adjusted forward, any time in the period that is skipped is invalid. (Parameter 'dateTime')

Lucene.Net.Util.Events.TestPubSubEvent.CanAddSubscriptionWhileEventIsFiringNonGeneric() (net472)

Message:

Expected: True Actual: False

Lucene.Net.Analysis.Icu.TestICUNormalizer2CharFilter.TestRandomStrings() (net5.0, net48)

Message:

Lucene.Net.Diagnostics.AssertionException : End() called before IncrementToken() returned false!

Expected Behavior

The tests all pass.

Steps To Reproduce

No response

Exceptions (if any)

No response

Lucene.NET Version

4.8.0-beta00017

.NET Version

Various (listed above)

Operating System

Windows

Anything else?

x64 platform, running ./build.ps1 -t

[NightOwl888]

Lucene.Net.Util.Events.TestPubSubEvent.CanAddSubscriptionWhileEventIsFiringNonGeneric()

I think we should [Ignore] this one. This is a test we inherited when we brought over the code from Prism.Core (which went commercial), and it has always been flakey. But we haven't had any complaints from .NET Framework or .NET Standard 2.0 users that these weak events are failing or causing any other issues.

It is a workaround for when ConditionalWeakTable<TKey, TValue> had no enumerator, so we use weak events to fake it.

[NightOwl888]

Lucene.Net.Index.TestIndexReaderClose.TestCloseUnderException() (net5.0, net48, net472)

I was unable to reproduce this on net48 with this seed. A stack trace might have been helpful. But, after looking up the exception, it doesn't seem like a coincidence that this error was reported very close to the time of the daylight savings time change in the EU (October 27, 2024).

Lucene.Net.Analysis.Icu.TestICUNormalizer2CharFilter.TestRandomStrings() (net5.0, net48)

This is repeatable, but only if running all of the tests in the class with the provided seed. Running the test explicitly will pass the test. I set numThreads = 0 in CheckRandomData() (being careful not to comment out TestUtil.NextInt32(random, 2, 4);) to disable the concurrency checking and the test passes. So, this seems like a concurrency problem with the latest changes to ICU4N.Text.Normalizer2 and/or the changes to the analyzer to accommodate the new design.