It is hard to tell if this is an actual bug or an artifact of the beta cluster test environment, but in recently time the browser tests for MMV in Chrome have been encountering sporadic errors on well-known test images "Error: Could not load image data. http"
Examples:
I suspect this is still beta being unreliable because of HHVM hiccups. We've had sporadic errors like that ever since HHVM was rolled out on beta.
Probably a HTTP 500 on the imageinfo API call. The last test run was Oct 17, 2014 7:29:05 AM. The Apache log for deployment-mediawiki03 doesn't contain anything for that date; no clue if that's the right place to look, though.
It's actually gotten a lot worse now that I've split the large scenarios into smaller ones following Dan's advice. Just because there are statistically more chances of running into the problem, I guess.
Last successful puppet run on all the cache instances on betalabs was on:
The last Puppet run was at Fri Oct 17 10:07:24 UTC 2014 (37931 minutes ago).
So this *might* be related?
See bug 73263 for info on the puppet issue.
Might be, but there are occurrences of "Could not load image data http" that are older than Oct 17th.
Today I ran into a similar message while I was working locally, that happened because I navigated away from the page as it was loading an image. It might be what's happening here and the fact that this error is displayed at all in situations where an XHR is aborted by the browser is undesirable. Glancing at media viewer's code where that error comes from, I think we could narrow it down a little to understand what's happening in the browser tests.
I see these every now and then on beta, even with normal browsing behavior (no aborted requests). I think beta is just aborting requests (maybe request involving storing a thumbnail locally?) from time to time.
At any rate, that error message could be improved.
Mass-removing the Multimedia tag from MediaViewer tasks, as this is now being worked on by the Reading department, not Editing's Multimedia team.
I do not remember seeing this problem in a long time. Please reopen if this is still a problem.