Bug 1416879 - Part 0d: Move browser_multie10s_update.js into its own directory. r=bkelly

This test is unfortunately a victim of a multi-e10s unregister() race
from browser_force_refresh and given the imminent multi-e10s cleanup
that's going to happen, the more complicated alternatives aren't worth
the effort versus just moving this test to its own directory.

..and now it's disabled too.  Bug 1429794 tracks re-enabling.

--HG--
rename : dom/workers/test/serviceworkers/browser_multie10s_update.js => dom/workers/test/serviceworkers/isolated/multi-e10s-update/browser_multie10s_update.js
rename : dom/workers/test/serviceworkers/file_multie10s_update.html => dom/workers/test/serviceworkers/isolated/multi-e10s-update/file_multie10s_update.html
rename : dom/workers/test/serviceworkers/server_multie10s_update.sjs => dom/workers/test/serviceworkers/isolated/multi-e10s-update/server_multie10s_update.sjs
extra : rebase_source : 2f93eda70120fff7f5b8ebac3db1f46c3aaf3070
extra : source : 3f059dbbf8ba6d42984a4a8c08836aec3ae37f47

dom/workers/moz.build

@@ -125,6 +125,7 @@ MOCHITEST_CHROME_MANIFESTS += [
 
 BROWSER_CHROME_MANIFESTS += [
     'test/serviceworkers/browser.ini',
+    'test/serviceworkers/isolated/multi-e10s-update/browser.ini',
 ]
 
 XPCSHELL_TESTS_MANIFESTS += ['test/xpcshell/xpcshell.ini']

dom/workers/test/serviceworkers/browser.ini

@@ -8,20 +8,16 @@ support-files =
   download_canceled/server-stream-download.sjs
   download_canceled/sw_download_canceled.js
   fetch.js
-  file_multie10s_update.html
   file_userContextId_openWindow.js
   force_refresh_browser_worker.js
   empty.html
   empty.js
-  server_multie10s_update.sjs
   utils.js
 
 [browser_devtools_serviceworker_interception.js]
 [browser_force_refresh.js]
 [browser_download.js]
 [browser_download_canceled.js]
-[browser_multie10s_update.js]
-skip-if = !e10s || os != "win" # Bug 1404914
 [browser_storage_permission.js]
 [browser_unregister_with_containers.js]
 [browser_userContextId_openWindow.js]

dom/workers/test/serviceworkers/isolated/README.md

@@ -0,0 +1,19 @@
+This directory contains tests that are flaky when run with other tests
+but that we don't want to disable and where it's not trivial to make
+the tests not flaky at this time, but we have a plan to fix them via
+systemic fixes that are improving the codebase rather than hacking a
+test until it works.
+
+This directory and ugly hack structure needs to exist because of
+multi-e10s propagation races that will go away when we finish
+implementing the multi-e10s overhaul for ServiceWorkers.  Most
+specifically, unregister() calls need to propagate across all
+content processes.  There are fixes on bug 1318142, but they're
+ugly and complicate things.
+
+Specific test notes and rationalizations:
+- multi-e10s-update: This test relies on there being no registrations
+  existing at its start.  The preceding test that induces the breakage
+  (`browser_force_refresh.js`) was made to clean itself up, but the
+  unregister() race issue is not easily/cleanly hacked around and this
+  test will itself become moot when the multi-e10s changes land.

dom/workers/test/serviceworkers/isolated/multi-e10s-update/browser.ini

@@ -0,0 +1,8 @@
+[DEFAULT]
+support-files =
+  file_multie10s_update.html
+  server_multie10s_update.sjs
+
+[browser_multie10s_update.js]
+skip-if = true # bug 1429794 is to re-enable, then un-comment the below.
+#skip-if = !e10s # this is an e10s-only test.

dom/workers/test/serviceworkers/isolated/multi-e10s-update/browser_multie10s_update.js

@@ -5,7 +5,8 @@ const { classes: Cc, interfaces: Ci, results: Cr } = Components;
 // Testing if 2 child processes are correctly managed when they both try to do
 // an SW update.
 
-const BASE_URI = "http://mochi.test:8888/browser/dom/workers/test/serviceworkers/";
+const BASE_URI =
+ "http://mochi.test:8888/browser/dom/workers/test/serviceworkers/isolated/multi-e10s-update/";
 
 add_task(async function test_update() {
   info("Setting the prefs to having multi-e10s enabled");