So currently waitForDevice tries 21 times (but looks like it wants to only try 20 times). Up until now (with joel's patches to update our devicemanager), it was never actually retrying to connect past the initial attempt. Now that we do, we typically see it connect within the first or second attempt, unless the tegra is bad. In the bad cases, it retries 20 times, with a full minute between retries, giving us ~30 minutes wait when a device would reasonably fail out sooner. This patch addresses the in-IRC discussed plan, which is to retry merely 3 times (instead of 20) and fail out/RETRY if the device isn't back yet. This also addresses an issue we noticed where reboot.py had a delay of *10 minutes* between retries to connect, for 21 tries, which meant we were actually running the reboot.py at the *end* of a job for 2-3 hours, we can do much better and report job results sooner if we reduce that as well, so reduced to 5 minutes, meaning once we issue a reboot command, reboot.py for tegras shouldn't take longer than 15 minutes to return job results to developers. This patch addresses that, r? to joel since this greatly affects visibility in that area, and Joel was the one who helped come up with this plan, f? to armen for sake of sanity in having another relenger peek. This patch is built on top of joel's three in Bugs 795028, 795071, and 795074

Comment on attachment 667839 [details] [diff] [review] [tools] waitForDevice fixups Review of attachment 667839 [details] [diff] [review]: ----------------------------------------------------------------- actually I never modified config.py, so the waitForDevice inside of there won't try to reconnect. If you do want it to reconnect add in: if dm._sock: dm._sock.close() dm._sock = None dm.deviceRoot = None dm.retries = 0 just like the sut_lib.py version. Better yet, use sut_lib.py! r+ as this change does what it intends to do for this bug, accept my comment as a secondary bug.

Comment on attachment 667839 [details] [diff] [review] [tools] waitForDevice fixups Completely sane. This is great! Thanks to the two of you. I can't thank enough how much you guys are cleaning everything up. You rock!

(In reply to Joel Maher (:jmaher) from comment #1) > just like the sut_lib.py version. Better yet, use sut_lib.py! > > r+ as this change does what it intends to do for this bug, accept my comment > as a secondary bug. I didn't change this patch to address these comments (but they sure are good comments), we'll treat them as further cleanup/followup.