Page MenuHomePhabricator

Revisit handling of `about` ids in parser tests
Open, Needs TriagePublic

Description

At some point in the distant past, we started normalizing away (removing, really) about ids in parser test output. The rationale at the time was that parsing was asynchronous and we couldn't guarantee that the exact same about IDs would be generated in two different runs, so they were causing spurious test failures and a bunch of update noise in tests.

Since that time parsing has changed to be almost entirely synchronous, but then another source of variance emerged because Parsoid would parse in multiple different DOM documents and in the process of cloning fragments from one document to another the about IDs would get incremented in unpredictable ways. I believe that issue has been mostly resolved now as well.

We should revisit our handling of about ids and stop stripping them from parser test results. If they are still inconsistent between runs, we can do a renumbering pass as part of normalization, but we shouldn't remove them entirely.

Event Timeline

Change #1190727 had a related patch set uploaded (by Arlolra; author: Arlolra):

[mediawiki/extensions/Cite@master] Revert "Fix re-serialization of incomplete follow in <ref> tags"

https://gerrit.wikimedia.org/r/1190727

Change #1192131 had a related patch set uploaded (by Thiemo Kreuz (WMDE); author: Thiemo Kreuz (WMDE)):

[mediawiki/extensions/Cite@master] Remove problematic fallback behavior for Follow nodes

https://gerrit.wikimedia.org/r/1192131

Change #1192132 had a related patch set uploaded (by Thiemo Kreuz (WMDE); author: Thiemo Kreuz (WMDE)):

[mediawiki/extensions/Cite@master] Add test cases for multiple & incomplete Follow nodes

https://gerrit.wikimedia.org/r/1192132

Change #1192131 merged by jenkins-bot:

[mediawiki/extensions/Cite@master] Remove problematic fallback behavior for Follow nodes

https://gerrit.wikimedia.org/r/1192131

Change #1192132 abandoned by Thiemo Kreuz (WMDE):

[mediawiki/extensions/Cite@master] Add test cases for multiple & incomplete Follow nodes

Reason:

One test appears to be a duplicate, the other should be unreachable by users. Personally I think it's not worth keeping these around now that Ia11f56e is merged. Please restore if you disagree.

https://gerrit.wikimedia.org/r/1192132

Change #1190727 merged by jenkins-bot:

[mediawiki/extensions/Cite@master] Replace ref content fail-safe with less aggressive one

https://gerrit.wikimedia.org/r/1190727