Based on the assertions it looks like this is getting through to toolbox.destroy but that the promise returned there isn't being resolved
At one point I was able to reproduce this locally with the old frontend and `./mach mochitest devtools/client/framework/test/browser_toolbox_custom_host.js --run-until-failure`. AFAICT the destroy() function in the old console isn't resolving but as soon as I started to debug it the problem went away and I can't reproduce anymore. This appears to be happening even without beta simulation, but just with the old console UI active: https://treeherder.mozilla.org/#/jobs?repo=try&revision=9c65d0c4114a0610ceece3c59d66fd94f584d389&selectedJob=120112607.
Component: Developer Tools: Framework → Developer Tools: Console
Not sure if it actually made a difference, but I've been able to reproduce again with: ./mach mochitest devtools/client/framework/test/browser_toolbox_custom_host.js --run-until-failure --setenv MOZ_CHAOSMODE=true Also, it's a bit surprising but it appears that the webconsole panel does get destroyed successfully and that we are instead hanging inside the target destroy.
I've been able to reproduce most consistently on a clean m-c with: ./mach mochitest devtools/client/framework/test/browser_toolbox_custom_host.js --run-until-failure --setpref devtools.webconsole.new-frontend-enabled=false Then right as the test window opens up and before the toolbox opens, switch focus away from the the browser and I can get the failure to happen most of the time. So far I've narrowed this down to WebConsoleClient.detach not calling back onResponse which prevents the client from closing, which prevents the target from destroying, which prevents the toolbox from destroying. The shutdown is something like: tabbrowserxml: removeTab -> target.js: destroy -> main.js: close -> webconsole/client.js: detach -> webconsole/client.js: stopListeners And the call to stopListeners is not returning. For whatever reason this only happens with the old frontend, and only in this particular shutdown sequence. I will try to debug more today but I'm short on time, so leaving needinfos to some folks who may be able to help track it down tomorrow. Since this is blocking beta we could force the new webconsole on for this test to get past the error. But I think this is surfacing a race in the toolbox shutdown sequence that leaves things in a very bad state (similar to Bug 1383997).
If we need to unblock the beta merge I can make a patch that changes a pref for this test to use the new frontend and file a new bug for tracking down the toolbox shutdown race that this is surfacing.
Or we could skip the test in beta and re-enable it once we fix it
Priority: -- → P3
Whiteboard: [console-html] [triage] → [reserve-console-html]
Per IRC discussion with bgrins, disabled on Beta. https://hg.mozilla.org/releases/mozilla-beta/rev/f9b4b766ab8e
status-firefox56: --- → disabled
status-firefox57: --- → affected
14 failures in 888 pushes (0.016 failures/push) were associated with this bug in the last 7 days. Repository breakdown: * try: 12 * mozilla-beta: 2 Platform breakdown: * linux64: 8 * linux32-nightly: 2 * windows8-64: 1 * osx-10-10: 1 * linux64-nightly: 1 * linux32: 1 For more details, see: https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1386410&startday=2017-07-31&endday=2017-08-06&tree=all
Still alive and well for 57 too.
2 failures in 901 pushes (0.002 failures/push) were associated with this bug in the last 7 days. Repository breakdown: * try: 2 Platform breakdown: * windows7-32: 1 * linux32-nightly: 1 For more details, see: https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1386410&startday=2017-08-07&endday=2017-08-13&tree=all
1 failures in 949 pushes (0.001 failures/push) were associated with this bug in the last 7 days. Repository breakdown: * try: 1 Platform breakdown: * linux64-stylo: 1 For more details, see: https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1386410&startday=2017-08-14&endday=2017-08-20&tree=all
Created attachment 8902458 [details] [diff] [review] workaround patch I've been using this patch to work around the issues at the moment. If we can't find anybody to prioritize investigating these issues, I'd rather just land it on m-c so I can stop juggling it in my patch queue.
Attachment #8902458 - Flags: feedback?(pbrosset)
Sure, let's skip the test for now. We still need to investigate the toolbox destroy race, but that can happen later in this bug or in another one.
Attachment #8902458 - Flags: feedback?(pbrosset) → feedback+
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/mozilla-inbound/rev/e9b91c1af369 Skip browser_toolbox_custom_host.js for permafailing on uplift to Beta. r=bgrins, f=pbro
Thanks for the feedback, Patrick.
status-firefox57: affected → disabled
Assignee: nobody → ryanvm
Status: NEW → ASSIGNED
Priority: P3 → P1
I am very-much not the assignee of this bug nor is it sounding likely to be fixed during the 57 cycle.
Assignee: ryanvm → nobody
Status: ASSIGNED → NEW
status-firefox56: disabled → verified disabled
status-firefox57: disabled → verified disabled
Target Milestone: Firefox 57 → ---
I believe this can be unskipped now as of Bug 1397425. Would you please be able to check un-skipping this on a beta simulation to confirm?
Depends on: 1397425
Looks that way indeed!
Assignee: nobody → bgrinstead
status-firefox57: verified disabled → affected
Pushed by email@example.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/a153979bd62f Re-enable browser_toolbox_custom_host.js now that the new console frontend is riding the trains. r=bgrins
Status: NEW → ASSIGNED
Iteration: --- → 57.3 - Sep 19
Priority: P3 → P1
https://hg.mozilla.org/mozilla-central/rev/a153979bd62f - patch landed 4 hours ago
Status: ASSIGNED → RESOLVED
Last Resolved: 11 months ago
status-firefox57: affected → fixed
Resolution: --- → FIXED
Target Milestone: --- → Firefox 57
You need to log in before you can comment on or make changes to this bug.