[
https://issues.apache.org/jira/browse/UNOMI-1027?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18136138#comment-18136138
]
Serge Huber commented on UNOMI-1027:
------------------------------------
Root cause found: this is UNOMI-1028 (fixed in master 1285eb670, PR #933,
2026-10-29).
* All 22 failing CI jobs (2026-08-11 → 2026-10-26, ES and OS) log
{{graphql-condition-type-null typeId=userListPropertyCondition
resolvableUnderSystemTenant=false}}: the type is missing even from the system
tenant, so {{findLists}} matches nothing.
* The failing master run's Karaf log (run 37733062661) registered only 29
condition types, none from graphql cxs-impl: the startup-scan race from
UNOMI-1028 lost that bundle's types. There have been no occurrences since the
fix.
PR https://github.com/apache/unomi/pull/939 (tenant fallback plus writing list
membership directly) was closed: it would not have prevented any of these
failures. A separate real bug found while reviewing it is filed as UNOMI-1030.
> GraphQLListIT.testCRUD intermittently fails waiting for a profile in a list
> ---------------------------------------------------------------------------
>
> Key: UNOMI-1027
> URL: https://issues.apache.org/jira/browse/UNOMI-1027
> Project: Apache Unomi
> Issue Type: Test
> Affects Versions: unomi-4.0.0
> Reporter: Serge Huber
> Priority: Major
> Fix For: unomi-4.0.0
>
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> h3. In short
> A GraphQL integration test fails at random on both search engines, which
> blocks unrelated changes until CI is re-run.
> h3. Details
> * {{GraphQLListIT.testCRUD}} fails at "Failed waiting for profile in list
> query" after {{addProfileToList}} (seen on PR 843 and PR 928, both unrelated;
> passes on re-run).
> * The test already refreshes the indices on every polling attempt, so a stale
> refresh is not the cause; the list membership is sometimes not visible to
> {{findLists}} at all within the wait.
> h3. Fix and test
> Find why the profile's list membership is not visible (asynchronous save, or
> a different index or tenant), fix it in the code or the test, and confirm
> with repeated runs on both engines.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
