Feel free to reject as wontfix if needed :) I hit several jobs on production where the TC task ran long, was killed and the artifact contained no time data to extract. It produces a console errors in Logviewer (per screen grabs). Maybe we can provide a clearer error so the user viewing doesn't think there is something wrong with Logviewer itself. To reproduce: o load https://treeherder.mozilla.org/logviewer.html#?job_id=11255405&repo=mozilla-inbound o open the console Expected: Some sort of "Task Cluster header is invalid" console error. Or perhaps a non-sticky thNotify blue info notifier instead during the logviewer load, with the same message? I kind of like that idea since the initial page load looks fine and you don't see the current error unless you open the console. See also bug 1147867 and garndt's comments below from IRC: "Yea we have a bug for reporting errors to tree herder when a task runs too long. If a task consistently runs long or always comes close then we should increase the max runtime in the in tree task definition"

Console on Firefox.

Console on Chrome.

Ultimately this is another example of lack of json schema validation. The artifact is: https://treeherder.mozilla.org/api/project/mozilla-inbound/artifact/?job_id=11255405&name=text_log_summary The second step is (unsurprisingly given the job failed and TC current sucky implementation wrt this) missing the "finished" property. We should either: 1) Enforce in the log parser that "finished" is present (or make up a value if it's not) 2) Just handle "finished" being missing in the log viewer, to avoid the JS exception (eg say "Unknown" for the duration). Mauro/Cameron, thoughts on which approach we should take? I'm guessing we need to do #2 either way, since when people start submitting their own text_log_summary artifacts (vs relying on the log parser) they could easily be missing the properties themselves too.

This is fixed by one of my commits in bug 1188444 :-)