1587994 - 40% more GV crashes on Fenix 2.0 vs 1.4

Status: RESOLVED DUPLICATE of bug 1557449

(GeckoView :: General, defect)

Description

[Kevin Brosnan [Ex-Mozilla]]

The Fenix 2.0 release has an increase of GV crashes. This shows up in the ARMv7 and aarch64 builds.

Below is the crashes per build ID for the last 3 aarch64 Fenix releases. The number of installs has not changed dramatically between the two releases. I don't have any information about usage hours. I am also seeing an uptick in the daily crash count on my spreadsheet. It contains raw install counts so is not available for public distribution.

12341950 (1.3)   114 
12531634 (1.4)   129 
12610151 (2.0)   246 

I don't see a dramatic difference between the GV 69 and GV 70 crash data on crash-stats. There are a few signatures that move around but I don't see something big and new for GV 70.

Comment 1

[Emily Toop (:fluffyemily)]

Did the crash ping land in GV70?

Comment 2

[Kevin Brosnan [Ex-Mozilla]]

I don't know. The numbers I have been tracking are independent of the crash ping AFIK.

Comment 3

[Chris Peterson [:cpeterson]]

Our GV top crash is still CreateEndpoints bug 1557449 (java.lang.RuntimeException: at org.mozilla.gecko.process.GeckoProcessManager.start(GeckoProcessManager.java)). CreateEndpoint crashes make up 65% (924) of GV 70.0b crash reports from the last two weeks. Unfortunately, we don't what is causing this crash. We're hoping the e10s-multi rewrite of GV's process launching code will make these crashes "go away".

Comment 4

[Chris Peterson [:cpeterson]]

Did Fenix or A-C make a change in Fenix 2.0 that caused a race condition to become more likely? Kevin said he will check with the Fenix team.

Didn't we make some change to GV or Fenix around process warming to reduce race conditions?

Comment 5

[Chris Peterson [:cpeterson]]

Duping to CreateEndpoint bug 1557449 because that crash signature is the cause of this spike in Fenix crashes.