Page MenuHomePhabricator

Diffusion tag cache not updating correctly when tags move?
Closed, ResolvedPublic


See IRC. @mbishopim3 reports that when tags move across commits, the Diffusion UI isn't pointing at the new commit. This may be an error in the tag/branch ref cache.

Event Timeline

epriestley claimed this task.
epriestley raised the priority of this task from to Normal.
epriestley updated the task description. (Show Details)
epriestley added a project: Diffusion.
epriestley added subscribers: epriestley, mbishopim3.

I wasn't immediately able to reproduce this. Here's what I did:

>>> orbital ~/repos/POEMS $ git tag -fm bloop quack HEAD
Updated tag 'quack' (was ae1456f)


>>> orbital ~/repos/POEMS $ git push --tags --force
Counting objects: 1, done.
Writing objects: 100% (1/1), 153 bytes | 0 bytes/s, done.
Total 1 (delta 0), reused 0 (delta 0)
To ssh://dweller@localhost/diffusion/POEMS/
 + ae1456f...0254e64 quack -> quack (forced update)

Once parsing completed, the new commit was reflected in the UI:

Screen_Shot_2014-02-24_at_12.55.39_PM.png (113×1 px, 17 KB)

(Note that the commit is not rPOEMSae1456f.)

Can you browse via that tag now? We see the hash update but then get a "Failed to look up tag" exception.

Yeah, works fine locally:

Screen_Shot_2014-02-24_at_1.06.18_PM.png (1×1 px, 165 KB)

Maybe I could make a test repo on this server and you could try breaking it by pushing tags/commits? I can play around locally a bit more too but I'm sort of just fishing at this point.

Let's do it with a test repo, I'll execute the commands and see if it gets crazy.

Still seeing this issue on a hosted phabrictor instance (hosted by phoreplay, so I don't know if it's up to date). Phabricator version 64c87770266444ff6e4a7156a97c6236c1ff090b, Arcanist Version 46ce8a5a35f9057b30151b5cb99a80642d913d22, libphutil Version 00e1f2da28291027b369f363e82c12f0b70de374. No exceptions generated, just out of date tag information.

64c877702 is about 3 months out of date (April 29).

Fair enough. I am still seeing this bug at that point. I don't know if it's been fixed later...

This is like two years old now, and I'm pretty sure it works now since no one else has reported it and it would likely have come up when clustering was written a few months ago if it was still an issue. If anyone is still seeing issues, please feel free to file a new report with reproduction steps.