Under a no-external connections regime, these sorts of things will crash. With the patches in bug 995417 applied, during crashtests: 12:57:48 INFO - REFTEST TEST-LOAD | file:///builds/slave/test/build/tests/reftest/tests/layout/base/crashtests/813372-1.html | 1183 / 2655 (44%) 12:57:51 INFO - BAD CONNECT: connecting to safebrowsing.google.com 0 12:57:52 INFO - TEST-INFO | Main app process: killed by SIGSEGV 12:57:52 WARNING - TEST-UNEXPECTED-FAIL | file:///builds/slave/test/build/tests/reftest/tests/layout/base/crashtests/813372-1.html | Exited with code 11 during test run 12:57:52 INFO - INFO | automation.py | Application ran for: 0:03:11.148802 12:57:52 INFO - INFO | zombiecheck | Reading PID log: /tmp/tmpC0iNBxpidlog 12:57:52 INFO - mozcrash INFO | Downloading symbols from: https://ftp-ssl.mozilla.org/pub/mozilla.org/firefox/try-builds/nfroyd@mozilla.com-8a38620b1782/try-linux/firefox-32.0a1.en-US.linux-i686.crashreporter-symbols.zip 12:58:29 WARNING - PROCESS-CRASH | file:///builds/slave/test/build/tests/reftest/tests/layout/base/crashtests/813372-1.html | application crashed [@ nsSocketTransport::InitiateSocket()] 12:58:29 INFO - Crash dump filename: /tmp/tmpHCBuJu.mozrunner/minidumps/586d7143-9d97-5431-3aaf852f-3a8b8a0e.dmp It doesn't seem to happen consistently; took a number of try runs before seeing this. Looks like mochitests just set the relevant URLs to local servers; we should probably be turning the service off entirely for crashtests/reftests.

Just like the telemetry bits in bug 997820.

I snuck in a change to do this for jsreftests too; I expect that doing so was OK: https://hg.mozilla.org/integration/mozilla-inbound/rev/74d53a60db06

https://hg.mozilla.org/releases/mozilla-beta/rev/eada024fd5e0 https://hg.mozilla.org/releases/mozilla-b2g30_v1_4/rev/a17650929832 https://hg.mozilla.org/releases/mozilla-b2g28_v1_3/rev/e866dd05cb12 https://hg.mozilla.org/releases/mozilla-esr24/rev/2b8f4139e84b