Electrolysis/Meetings/2016-09-22: Difference between revisions

Electrolysis/Meetings/2016-09-22 (view source)

1,351 bytes added , 22 September 2016

105

edits

@@ Line 20: / Line 20: @@
 ==chutten==
+{{bug|1218576}} landed and finally stuck. This brings content-process telemetry to the parent process so we can start reporting it on non-saved-session pings (primarily, on "main" pings). This also paves the way for things like {{bug|1304494}} to add telemetry from non-content child processes (in this case, a GPU process).
+Unfortunately, not all is amazing. First was an intermittent assert failure in {{bug|1304009}} which means that, though the child is accumulating to valid histograms, somehow on the parent it is trying to accumulate to an invalid histogram id. Then was a leak {{bug|1304660}} which I don't understand, as I'm doing the same thing as is done elsewhere without issue.
+Finally, and most troubling, there is an infrequent crash in {{bug|1304519}} which is pointing at memory corruption of a to-be-deallocated block as an potential culprit.
+I'm at a loss of how to handle this. This is my first rodeo of this kind, so any pointers others can provide are greatly appreciated.
+* In other news, I've been tracking release ER, which has shown a decent bump post-e10s-enabling: [https://sql.telemetry.mozilla.org/queries/1165/source#2038]. Hard to say for sure, as 48's release was a bit of a nightmare, and all releases have variable e10s over time... but there's no evidence that e10s is causing a decrease in release engagement.
 ==bytesized==