I've noticed this especially during testing when I disable/usercancel jobs submitted to staging treeherder. If you clear logcat -c then issue a logcat -d -b main, it will hang until there is more data written to the logcat. This was caused when we added buffers to adb's logcat and defaulted to -b main if none were specified in Bug 1156812. If all we want is the main logcat buffer, there is no need to specify it explicitly and removing the -b main from the command solves the hang problem. I am fixing this locally here in Autophone but this could/should also be changed in mozdevice's copy as well.

Comment on attachment 8667432 [details] [diff] [review] bug-1209653-do-not-default-to-logcat-main-v1.patch Review of attachment 8667432 [details] [diff] [review]: ----------------------------------------------------------------- do we need to resolve this in the clear_logcat as well?

Probably, I noticed that when I was explaining in irc as well. I'll make that change locally before I push. Thanks!

Bug 1156812 introduced the use of logcat buffers. This appears to be the cause of the previously unexplained "apparent" improvement on 8/21 noticed by mfinkle on 8/25 See http://phonedash.mozilla.org/#/org.mozilla.fennec/throbberstart/local-blank/norejected/2015-08-19/2015-08-24/notcached/noerrorbars/standarderror/notry This patch will revert that change and will introduce an "apparent" regression which is not real. The difference appears to be due to the change in the "start" time used to convert the raw Throbber measurements into a value relative to when Gecko/Fennec starts. Looking at this, I do see some opportunity to improve analyze_logcat in s1s2 at least, but I will treat that as a separate bug.