September 2026 - notifications

[xwiki/xwiki-dev-llm] 9aadf6: [Misc] Make the test-guidelines skill claim the ac...
by XWiki Notifications 17 Sep '26

17 Sep '26

Branch: refs/heads/okf-prchecker-grant Home: https://github.com/xwiki/xwiki-dev-llm Commit: 9aadf69ee9a6e8d64e9653f8ae5ea079f6b9e753 https://github.com/xwiki/xwiki-dev-llm/commit/9aadf69ee9a6e8d64e9653f8ae5ea… Author: Vincent Massol <vincent(a)massol.net> Date: 2026-09-17 (Thu, 17 Sep 2026) Changed paths: M xwiki/instructions/xwiki-org.md M xwiki/skills/xwiki-test-guidelines/SKILL.md Log Message: ----------- [Misc] Make the test-guidelines skill claim the act of writing a test * The skill's description described its contents instead of naming its trigger, and the always-on instruction called it "deeper testing guidance", so a small test change (one added @Test method in an existing class) routed to nothing and the rules were never loaded. Turn the description into a trigger that names the action, covers functional @UITest tests, and closes the "it's only one method / the class already exists / this one is simple" escape, and make the always-on line an obligation rather than an invitation. * Drop the *Test.java / *IT.java naming rule from the always-on file: it is stated verbatim in okf/testing/strategy.md, which is its home, and the budget is better spent on the routing rule. Co-Authored-By: Claude Opus 5 (1M context) <noreply(a)anthropic.com> To unsubscribe from these emails, change your notification settings at https://github.com/xwiki/xwiki-dev-llm/settings/notifications

1 0

[xwiki/xwiki-dev-llm] 57d4c0: [Misc] Cover granting Programming Rights to a func...
by XWiki Notifications 17 Sep '26

17 Sep '26

Branch: refs/heads/okf-prchecker-grant Home: https://github.com/xwiki/xwiki-dev-llm Commit: 57d4c005c8cbc04e971c3351e47d460caa47be54 https://github.com/xwiki/xwiki-dev-llm/commit/57d4c005c8cbc04e971c3351e47d4… Author: Vincent Massol <vincent(a)massol.net> Date: 2026-09-17 (Thu, 17 Sep 2026) Changed paths: M xwiki/okf/testing/strategy.md Log Message: ----------- [Misc] Cover granting Programming Rights to a functional test's own content * The PRChecker bullet only explained how to READ a `PRChecker: Block programming right` log line, so nothing in the OKF answered the authoring question: my test creates content that needs the right, how do I grant it? Add that direction to the same bullet, keyed on the symptoms actually seen (the rendered "You need Programming Rights to execute the script macro [groovy]" error), with the concrete `@UITest(properties = {"xwikiPropertiesAdditionalProperties=test.prchecker.excludePattern=..."})` form, the fact that the exclusion really grants the right rather than only silencing the log, and the whole-reference match trap. * Tighten the probe clause that the addition made longer. Co-Authored-By: Claude Opus 5 (1M context) <noreply(a)anthropic.com> To unsubscribe from these emails, change your notification settings at https://github.com/xwiki/xwiki-dev-llm/settings/notifications

1 0

[[email protected]][Issue] Updates for XWIKI-6688: Possible race condition when clicking on a tab at the bottom of a page in view mode
by XWiki Notifications 17 Sep '26

17 Sep '26

1 0

[[email protected]][Issue] Updates for XWIKI-6161: Reset action does not work anymore
by XWiki Notifications 17 Sep '26

17 Sep '26

1 0

[[email protected]][Issue] Updates for XWIKI-6688: Possible race condition when clicking on a tab at the bottom of a page in view mode
by XWiki Notifications 17 Sep '26

17 Sep '26

1 0

[[email protected]][Issue] Updates for XWIKI-25019: Acquiring a document lock can fail with a deadlock on MySQL
by XWiki Notifications 17 Sep '26

17 Sep '26

1 0

[[email protected]][Issue] Updates for XWIKI-6161: Reset action does not work anymore
by XWiki Notifications 17 Sep '26

17 Sep '26

1 0

[xwiki/xwiki-platform] b57b24: XWIKI-6161: Reset action does not work anymore
by XWiki Notifications 17 Sep '26

17 Sep '26

Branch: refs/heads/XWIKI-6161 Home: https://github.com/xwiki/xwiki-platform Commit: b57b2443b7b7dbe9bb78102f33efdd4db7c0b007 https://github.com/xwiki/xwiki-platform/commit/b57b2443b7b7dbe9bb78102f33ef… Author: Vincent Massol <vincent(a)massol.net> Date: 2026-09-17 (Thu, 17 Sep 2026) Changed paths: M xwiki-platform-core/xwiki-platform-flamingo/xwiki-platform-flamingo-skin/xwiki-platform-flamingo-skin-test/xwiki-platform-flamingo-skin-test-docker/src/test/it/org/xwiki/flamingo/test/docker/VersionIT.java M xwiki-platform-core/xwiki-platform-oldcore/src/main/java/com/xpn/xwiki/store/XWikiHibernateVersioningStore.java Log Message: ----------- XWIKI-6161: Reset action does not work anymore * Detach the document's history nodes from the Hibernate session before the bulk delete query that removes them, so that the single node the reset archive is rebuilt with no longer clashes with them * Add a functional test resetting the history of a document Co-Authored-By: Claude Opus 5 (1M context) <noreply(a)anthropic.com> To unsubscribe from these emails, change your notification settings at https://github.com/xwiki/xwiki-platform/settings/notifications

1 0

[xwiki/xwiki-platform] 73b9ba: XWIKI-24997: XClass is cached as non-existing or o...
by XWiki Notifications 17 Sep '26

17 Sep '26

Branch: refs/heads/stable-17.10.x Home: https://github.com/xwiki/xwiki-platform Commit: 73b9ba8461540813aa36acea8e2ab38ec775d15f https://github.com/xwiki/xwiki-platform/commit/73b9ba8461540813aa36acea8e2a… Author: Michael Hamann <michael.hamann(a)xwiki.com> Date: 2026-09-17 (Thu, 17 Sep 2026) Changed paths: M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/pom.xml A xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/main/java/org/xwiki/xclass/test/XClassVisibilityTestScriptService.java A xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/main/resources/META-INF/components.txt M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/test/it/org/xwiki/xclass/test/ui/AllIT.java A xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/test/it/org/xwiki/xclass/test/ui/XClassCacheVisibilityIT.java Log Message: ----------- XWIKI-24997: XClass is cached as non-existing or outdated when it is saved while a document using it is being loaded (MySQL/MariaDB) * add XClassCacheVisibilityIT, a functional test that reproduces the race: one thread loads a document holding one object of each of 400 distinct XClasses while a second thread recreates a 401st XClass part-way through that load. The delay before recreating it is swept across the measured load window so that some round lands after the loading transaction took its snapshot and before it resolves that XClass. A round fails when the recreated XClass is reported as non-existing although flushing the cache brings it back, which proves the row was in the database all along. * assert separately that the store's connections run at READ COMMITTED, so that losing that setting fails here with a clear message instead of showing up as unrelated tests failing at random. Co-Authored-By: Claude Opus 5 (1M context) <noreply(a)anthropic.com> (cherry picked from commit 5eadc17a50ec528f8e607cd69aa607a499323715) Commit: cac35d0845a7dcca5327d07d2bb6eb6820a9a45b https://github.com/xwiki/xwiki-platform/commit/cac35d0845a7dcca5327d07d2bb6… Author: Michael Hamann <michael.hamann(a)xwiki.com> Date: 2026-09-17 (Thu, 17 Sep 2026) Changed paths: M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/pom.xml M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/main/java/org/xwiki/xclass/test/XClassVisibilityTestScriptService.java M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/test/it/org/xwiki/xclass/test/ui/XClassCacheVisibilityIT.java Log Message: ----------- XWIKI-24997: XClass is cached as non-existing or outdated when it is saved while a document using it is being loaded (MySQL/MariaDB) * rewrite XClassCacheVisibilityIT to decide by ordering instead of by timing: the loading thread opens a store transaction and forces its database snapshot, the main thread then saves the XClass, and only then does the loading thread load a document using it. Latches order those steps, so the test reproduces the problem or doesn't, the same way on every machine. * cover both halves of the issue: an XClass created while the document loads, which ends up cached as non-existing, and one that gains a property while the document loads, which ends up cached without it. * wait for the asynchronous work a save triggers - the tasks consumer and the Solr indexer - and empty the document cache before loading, so that the loading thread is the only reader of the XClass. The previous version raced two threads and swept a delay across the measured duration of a document load, whose length it set with 400 filler XClasses. It reproduced the problem locally in 12 of 20 rounds but not at all on CI: there the sweep found nothing on MySQL with a 454 ms window, and on PostgreSQL the window guard failed the test for being too fast rather than for the bug. It had two timing dependencies rather than one: the save has to commit between the snapshot and the resolution of the class, and the class has to be read from the database rather than from the document cache - which the asynchronous work following the save re-fills within milliseconds, from its own newer transaction. Both are now preconditions the test establishes instead of races it has to win, which also makes it cheap: a few seconds rather than the minute the sweep needed, and no filler documents at all. Co-Authored-By: Claude Opus 5 (1M context) <noreply(a)anthropic.com> (cherry picked from commit c28327d554a177476c939354a5f968f73e8c44aa) Compare: https://github.com/xwiki/xwiki-platform/compare/60e25ea58d37...cac35d0845a7 To unsubscribe from these emails, change your notification settings at https://github.com/xwiki/xwiki-platform/settings/notifications

1 0

[xwiki/xwiki-platform] 60945a: XWIKI-24997: XClass is cached as non-existing or o...
by XWiki Notifications 17 Sep '26

17 Sep '26

Branch: refs/heads/stable-18.4.x Home: https://github.com/xwiki/xwiki-platform Commit: 60945a8ce994127145b8af93272309ffa79b138b https://github.com/xwiki/xwiki-platform/commit/60945a8ce994127145b8af932723… Author: Michael Hamann <michael.hamann(a)xwiki.com> Date: 2026-09-17 (Thu, 17 Sep 2026) Changed paths: M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/pom.xml A xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/main/java/org/xwiki/xclass/test/XClassVisibilityTestScriptService.java A xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/main/resources/META-INF/components.txt M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/test/it/org/xwiki/xclass/test/ui/AllIT.java A xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/test/it/org/xwiki/xclass/test/ui/XClassCacheVisibilityIT.java Log Message: ----------- XWIKI-24997: XClass is cached as non-existing or outdated when it is saved while a document using it is being loaded (MySQL/MariaDB) * add XClassCacheVisibilityIT, a functional test that reproduces the race: one thread loads a document holding one object of each of 400 distinct XClasses while a second thread recreates a 401st XClass part-way through that load. The delay before recreating it is swept across the measured load window so that some round lands after the loading transaction took its snapshot and before it resolves that XClass. A round fails when the recreated XClass is reported as non-existing although flushing the cache brings it back, which proves the row was in the database all along. * assert separately that the store's connections run at READ COMMITTED, so that losing that setting fails here with a clear message instead of showing up as unrelated tests failing at random. Co-Authored-By: Claude Opus 5 (1M context) <noreply(a)anthropic.com> (cherry picked from commit 5eadc17a50ec528f8e607cd69aa607a499323715) Commit: ca06993981f3e99ee10ebebadfedcc69934bf2ce https://github.com/xwiki/xwiki-platform/commit/ca06993981f3e99ee10ebebadfed… Author: Michael Hamann <michael.hamann(a)xwiki.com> Date: 2026-09-17 (Thu, 17 Sep 2026) Changed paths: M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/pom.xml M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/main/java/org/xwiki/xclass/test/XClassVisibilityTestScriptService.java M xwiki-platform-core/xwiki-platform-xclass/xwiki-platform-xclass-test/xwiki-platform-xclass-test-docker/src/test/it/org/xwiki/xclass/test/ui/XClassCacheVisibilityIT.java Log Message: ----------- XWIKI-24997: XClass is cached as non-existing or outdated when it is saved while a document using it is being loaded (MySQL/MariaDB) * rewrite XClassCacheVisibilityIT to decide by ordering instead of by timing: the loading thread opens a store transaction and forces its database snapshot, the main thread then saves the XClass, and only then does the loading thread load a document using it. Latches order those steps, so the test reproduces the problem or doesn't, the same way on every machine. * cover both halves of the issue: an XClass created while the document loads, which ends up cached as non-existing, and one that gains a property while the document loads, which ends up cached without it. * wait for the asynchronous work a save triggers - the tasks consumer and the Solr indexer - and empty the document cache before loading, so that the loading thread is the only reader of the XClass. The previous version raced two threads and swept a delay across the measured duration of a document load, whose length it set with 400 filler XClasses. It reproduced the problem locally in 12 of 20 rounds but not at all on CI: there the sweep found nothing on MySQL with a 454 ms window, and on PostgreSQL the window guard failed the test for being too fast rather than for the bug. It had two timing dependencies rather than one: the save has to commit between the snapshot and the resolution of the class, and the class has to be read from the database rather than from the document cache - which the asynchronous work following the save re-fills within milliseconds, from its own newer transaction. Both are now preconditions the test establishes instead of races it has to win, which also makes it cheap: a few seconds rather than the minute the sweep needed, and no filler documents at all. Co-Authored-By: Claude Opus 5 (1M context) <noreply(a)anthropic.com> (cherry picked from commit c28327d554a177476c939354a5f968f73e8c44aa) Compare: https://github.com/xwiki/xwiki-platform/compare/8896be50b6f9...ca06993981f3 To unsubscribe from these emails, change your notification settings at https://github.com/xwiki/xwiki-platform/settings/notifications

1 0