[
https://issues.apache.org/jira/browse/LUCENE-10168?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17427701#comment-17427701
]
Michael Sokolov commented on LUCENE-10168:
------------------------------------------
This is what I see:
{{> Task :lucene:core:test}}
{{:lucene:core:test (SUCCESS): 5044 test(s), 159 skipped}}
{{The slowest tests (exceeding 500 ms) during this run:}}
{{ 19.12s TestIndexWriterThreadsToSegments.testDocsStuckInRAMForever
(:lucene:core)}}
{{ 12.90s TestSortOptimization.testRandomLong (:lucene:core)}}
{{ 9.55s TestSimpleTextTermVectorsFormat.testMergeWithIndexSort
(:lucene:codecs)}}
{{ 8.39s TestSimpleTextTermVectorsFormat.testMergeWithoutIndexSort
(:lucene:codecs)}}
{{ 6.93s TestNRTReplication.testCrashReplica (:lucene:replicator)}}
{{ 5.54s TestTieredMergePolicy.testSimulateAppendOnly (:lucene:core)}}
{{ 5.41s TestCompressingTermVectorsFormat.testMergeWithoutIndexSort
(:lucene:test-framework)}}
{{ 5.32s TestCompressingTermVectorsFormat.testMergeWithIndexSort
(:lucene:test-framework)}}
{{ 5.26s TestLucene50TermVectorsFormat.testMergeWithIndexSort
(:lucene:backward-codecs)}}
{{ 5.05s TestAssertingTermVectorsFormat.testMergeWithIndexSort
(:lucene:test-framework)}}
{{The slowest suites (exceeding 1s) during this run:}}
{{ 36.81s TestBackwardsCompatibility (:lucene:backward-codecs)}}
{{ 26.76s TestSimpleTextTermVectorsFormat (:lucene:codecs)}}
{{ 20.04s TestIndexWriterThreadsToSegments (:lucene:core)}}
{{ 15.90s TestCompressingTermVectorsFormat (:lucene:test-framework)}}
{{ 13.60s TestSortOptimization (:lucene:core)}}
{{ 13.37s TestBestSpeedLucene80DocValuesFormat (:lucene:backward-codecs)}}
{{ 13.01s TestSimpleTextDocValuesFormat (:lucene:codecs)}}
{{ 12.99s TestLucene50TermVectorsFormat (:lucene:backward-codecs)}}
{{ 12.90s TestAssertingTermVectorsFormat (:lucene:test-framework)}}
{{ 12.87s TestIndexWriter (:lucene:core)}}
in the case of backward-codecs it seems more like many test cases than slow
test cases? Indeed looking at TestBackwardsCompatibility, its various tests get
repeated for every index version we are supporting, and there are 32 of them.
So I think that there might be limited scope for making the individual tests
faster, and the urge to remove some of these from the "supported" list makes
sense. I wonder if we could choose a random sample from the list?
> drop support for 7.0 indexes in 9.0 (master)
> --------------------------------------------
>
> Key: LUCENE-10168
> URL: https://issues.apache.org/jira/browse/LUCENE-10168
> Project: Lucene - Core
> Issue Type: Task
> Reporter: Robert Muir
> Priority: Blocker
>
> Recently it was decided to extend back compat *two* major releases, as a best
> effort.
> The effort has not been the best, though.
> Test times have doubled (10 minutes vs 5 minutes on my computer), all the
> slowness is in backwards codecs tests. And 9.0 isn't even released! Imagine
> how slow the tests would be if we somehow made it to e.g. 9.10 release or
> something.
> It is unsustainable.
> This issue is to drop support for 7.0 indexes. I realize some ppl won't be
> happy about that, if you don't like it, then suggest some alternative fix for
> the tests and put in your time. Otherwise bye bye.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
