Page MenuHomePhabricator

SLF bot should have +2 rights and automatically merge own patches
Open, Stalled, Needs TriagePublic

Description

Hi, on Gerrit we have SLF bot. It updates domain.php of MediaWiki-extensions-SecureLinkFixer. I think it should have +2 for the extension repository and automatically set it on its own patches as they don't need review.

Event Timeline

Legoktm changed the task status from Open to Stalled.Nov 24 2019, 7:07 AM

This is currently on purpose. The following things would need to be resolved for me to feel comfortable enabling automatic +2 for SLF:

  • Moving SLF bot to some shared maintenance and higher trust infrastructure and out of Toolforge (maybe CI? or at the very least its own Cloud VPS project/instance like libup)
  • Sanity check the generated file is exactly what should have been generated (verify reproducibility)
  • Some sanity check on the diffstat, if every single domain is removed from the list, then we know something went wrong on Mozilla's end.

In general I think we can trust Mozilla's infrastructure/codebase because if we can't do that, then we're all already in big trouble - but we should have some kind of sanity check just in case the scraper fails and returns no results or something, OR they change their code structure around.

Legoktm renamed this task from SLF bot should have +2 right and automaticly merge own patches to SLF bot should have +2 rights and automatically merge own patches.Dec 14 2019, 7:48 PM

As an update:

This is currently on purpose. The following things would need to be resolved for me to feel comfortable enabling automatic +2 for SLF:

  • Moving SLF bot to some shared maintenance and higher trust infrastructure and out of Toolforge (maybe CI? or at the very least its own Cloud VPS project/instance like libup)

Still needs to be done. I think moving it into CI is the best option.

  • Sanity check the generated file is exactly what should have been generated (verify reproducibility)

Implemented.

  • Some sanity check on the diffstat, if every single domain is removed from the list, then we know something went wrong on Mozilla's end.

Implemented.

I also think using CI is better.

Yes, also.