Issue Bug Closed

#110244Linkvalidator aborts with "Data too long for column 'headline'"

Martina Kratky 07/21/26, 09:56 updated 08/11/26, 08:51

Description

Summary

The “Check for broken links” task can crash entirely with Data too long for column ‘headline’ (tx_linkvalidator_link.headline is varchar(255)). LinkAnalyzer::checkLinks() writes BackendUtility::getRecordTitle()’s result into it unchecked. Reproducible with core’s own default tt_content TCA (label_alt => ‘subheader,bodytext’): any content element with an empty header (e.g. hidden headline layout) and a long bodytext containing a broken link takes down the whole run, not just that one record.

Steps to reproduce

  1. tt_content record: empty header; bodytext > ~500 chars incl. a broken link.
  2. Run the “Check for broken links” scheduler task.

Root cause

Regression from #108581 (v14.2, record-type-specific labels). getRecordTitle() used to stop at the first non-empty label_alt field when the primary label is empty (see the break in the 12.4/13.4 branches). Since the rewrite, it always implodes all non-empty alt fields in that case – effectively label_alt_force is now always on, so unrelated long fields like bodytext get concatenated into the title. headline (and field/table_name/element_type) are still varchar(255); url/link_title/ url_response are already text, suggesting a related length issue (#98328, for url) was fixed by widening the column – headline seems to have been missed.

Suggested fix

Either restore the “first non-empty wins” behavior in getRecordTitle(), or cap headline before insert in linkvalidator.

Reviews 2

History 58

  1. Niklas Grieger commented and changed status Under Review Resolved done 0 % 100 % <span data-processor="relativetime" data-value="2026-08-07T07:10:13+00:00" > 08/07/26, 07:10 </span>

    Applied in changeset commit:47eac8ba460210ccb6a5586478ee3b47e9b80279.