Run integration tests against hyperion #3334

mxstbr · 2018-06-15T13:15:44Z

Status

WIP
Ready for review
Needs testing

Related issues (delete if you don't know of any)
Replaces #2843, based on #3243, also based on #3382 (so until that's merged you'll also see those changes here)

This basically means we're running integration tests in the production setup, which is a big deal. ™️ We should (in a follow-up PR) also be able to configure Cypress to fail tests on SSR checksum mismatches (in case that doesn't happen yet) so we catch those as soon as they happen instead of months after the fact by chance!

spectrum-bot · 2018-06-15T13:16:14Z

	Warnings
⚠️	These modified files do not have Flow enabled: config-overrides.js shared/middlewares/thread-param.js

Generated by 🚫 dangerJS

mxstbr · 2018-06-15T16:00:11Z

They are runninnnng!!!

mxstbr · 2018-06-15T16:17:34Z

@brian-mann CircleCI errors out because "Too long with no output (exceeded 10m0s)", but there was Cypress output to the console the whole time 🤔 Any clues why that happens?

brian-mann · 2018-06-15T16:25:29Z

It means something happened when trying to spawn the browser to run thread/delete_spec.js...

Doesn't make sense considering all of the other specs spawned correctly. Rerun CI a few times to see if it continually happens.

You can also try setting the DEBUG logs which will output a massive stream of data but will give details as to where Cypress was when this happened.

DEBUG=cypress:* cypress run

Another option is to try Chrome although that won't record a video. We automatically detect things like Electron crashes, and we automatically handle when the browser doesn't connect (we will correctly time out). So what I'm thinking here is that the browser actually did launch, but that something failed to load related to Cypress, since it doesn't appear that your tests ever began.

Try rerunning so we can see if it's just that spec file.

mxstbr · 2018-06-16T09:39:07Z

Now they ran through, must've been a fluke! Thanks for the response 🆗

brian-mann · 2018-06-16T19:04:30Z

Other users are reporting the same problem such as this: cypress-io/cypress#1934

There is something we need to look at and fix.

@brianlovin

@brianlovin I changed your test-payments stuff to work with :3006 instead of :3001 so that it works when running integration tests against hyperion!

mxstbr · 2018-06-18T11:54:17Z

All of the remaining tests fail due to the same error: "SecurityError: Blocked a frame with origin 'http://localhost:3006' from accessing a cross-origin frame." Unfortunately, the stack trace is totally useless in debugging why this happens:

Interestingly, this happens to all the tests that load the inbox view. @brianlovin do you have a hunch what this could be? What is the inbox view loading that could cause this, are we redirecting somewhere?

@brian-mann have you seen this before/could that error be made to be a bit more helpful? Right now, I'm stuck here, not knowing why this is happening or what I have to do to fix it.

This was necessary when we didn't use `react-loadable` for SSR, but now that we do this is entirely unused so we can just get rid of it!

@brianlovin

@brianlovin this _should_ fix the ?t param even when loaded with SSR, right? The issue previously was that we'd mount with an active thread, but the dashboard would still switch to the first one that's loaded once the thread feed loads.

We unset the mounted thread ID before any threads were loaded, so when they finally did load in mountedThreadId was null and the app thought it had to select the first thread. This fixes it by only doing the "unset mounted thread id" logic after threads have already loaded! 🎉

Also refactor render code a bit to be nicer

…gainst-hyperion

This should fix the issues with the SecurityError's from #3334

mxstbr · 2018-06-25T13:44:47Z

I thought disabling chrome web security would fix it, but apparently it didn't... 🤔

EDIT: Nevermind, now only the navbar specs are failing!

mxstbr · 2018-06-25T14:12:04Z

I think the remaining tests are failing because the ServiceWorker is caching HTML pages instead of going to the server-side rendering server or serving the app shell.

@NekR we want to do the standard app shell pattern where the first visit to a page returns the server-side rendered HTML, then the SW caches an app shell (/index.html) and then all future visits return that app shell, which loads fresh data from the API. How can we tell the offline-plugin to not cache the server-side rendered HTML, and only the app shell? What are we doing wrong?

spectrum/config-overrides.js

Lines 106 to 120 in 760da30

    
           new OfflinePlugin({ 
        
             appShell: '/index.html', 
        
             caches: process.env.NODE_ENV === 'development' ? {} : 'all', 
        
             externals, 
        
             autoUpdate: true, 
        
             ServiceWorker: { 
        
               entry: './public/push-sw.js', 
        
               events: true, 
        
               prefetchRequest: { 
        
                 mode: 'cors', 
        
                 credentials: 'include', 
        
               }, 
        
             }, 
        
             AppCache: false, 
        
           })

@brianlovin

Omgggg @brianlovin @uberbryn I think I finally fixed that gosh darn ServiceWorker error we've been constantly getting AND this might also fix those cached notifications pages we've been seeing!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

mxstbr · 2018-06-25T14:30:27Z

@NekR that last commit fixes it: a4f345d

The issue was that our app shell lives at /index.html, which the default rewrites method turns into /—but the app shell doesn't get returned for /, only for index.html! I think the appShell option shouldn't be passed through rewrites, that was very very annoying to debug? Or at least document the default rewrites method and make it clear that appShell is being run through it.

mxstbr · 2018-06-25T14:59:37Z

Now I'm getting "Timed out waiting for the browser to connect." for all tests 🤔

This is cherry-picked from #3334 since that's not quite done yet, but this should go live asap.

mxstbr · 2018-06-25T15:06:03Z

Since it seems like it'll take a bit longer to figure out Cypress 3.0 I've moved the ServiceWorker fix to #3384, which we should ship asap.

NekR · 2018-06-25T19:12:33Z

Yeah, that makes sense. I'll do the explanation and add warning that appShell is passed through rewrites. "Not passing it through rewrites" is a question for a major release though. If you're interested in that, please file an issue in the repo 🙏 P.S. Glad you were able fix the issue and sorry for the poor docs.

…

On Mon, Jun 25, 2018, 17:30 Max Stoiber ***@***.***> wrote: Omg @NekR <https://github.com/NekR> that last commit fixes it: a4f345d <a4f345d> The issue was that our app shell lives at /index.html, which the default rewrites method turns into /—but the app shell doesn't get returned for /, only for index.html! I think the appShell option shouldn't be passed through rewrites, that was very very annoying to debug? Or at least document the default rewrites method and make it clear that appShell is being run through it. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3334 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABIlkeJmpT4jLPZej_r2Mq19epVDjaE2ks5uAPQHgaJpZM4Upi5f> .

mxstbr added 4 commits June 2, 2018 11:09

Upgrade to Cypress 3.0

d4ece4b

Install cypress extra step

a265301

Merge branch 'alpha' into upgrade-cypress

fcd0b31

Run integration tests against hyperion

bba2dae

mxstbr mentioned this pull request Jun 15, 2018

Run integration tests against hyperion #2843

Closed

6 tasks

spectrum-bot bot added the WIP: Needs Testing label Jun 15, 2018

mxstbr added 3 commits June 15, 2018 16:46

Fix wait-on condition

c2c6b2f

Wait for API first, then hyperion

e8cb15b

Im a dummy

3d696e5

mxstbr added 2 commits June 18, 2018 13:37

Fix community billing tests

60944fb

@brianlovin I changed your test-payments stuff to work with :3006 instead of :3001 so that it works when running integration tests against hyperion!

Fix thread error

b077641

mxstbr added 2 commits June 18, 2018 17:08

Fix websockets in e2e tests

56d6a15

Add logout option to cy.auth(null)`

7b09560

brian-mann mentioned this pull request Jun 20, 2018

Cypress hangs on empty tests files cypress-io/cypress#1905

Closed

mxstbr added 8 commits June 23, 2018 13:00

Dashboard code cleanup

d997fae

Remove unused extra reducer and middleware injection

c683516

This was necessary when we didn't use `react-loadable` for SSR, but now that we do this is entirely unused so we can just get rid of it!

Don't redirect ?t for authed users; src/index.js cleanup

e097555

Fix ?t param with SSR

63ad628

@brianlovin this _should_ fix the ?t param even when loaded with SSR, right? The issue previously was that we'd mount with an active thread, but the dashboard would still switch to the first one that's loaded once the thread feed loads.

Fix dashboard thread feed loading state

ddaa093

Also refactor render code a bit to be nicer

Merge branch 'fix-t-query-param-on-load' into run-integration-tests-a…

a8dc5d3

…gainst-hyperion

Temporarily disable chrome web security for e2e tests

c9d3fea

This should fix the issues with the SecurityError's from #3334

mxstbr added 3 commits June 25, 2018 16:27

Fix content type header being set for uncached pages

61149cd

Fix cy.auth(null)

bc01b26

FIX SERVICEWORKER CACHING

a4f345d

Omgggg @brianlovin @uberbryn I think I finally fixed that gosh darn ServiceWorker error we've been constantly getting AND this might also fix those cached notifications pages we've been seeing!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

mxstbr added a commit that referenced this pull request Jun 25, 2018

Fix ServiceWorker caching

33660d7

This is cherry-picked from #3334 since that's not quite done yet, but this should go live asap.

mxstbr mentioned this pull request Jun 25, 2018

Fix ServiceWorker caching #3384

Merged

3 tasks

brianlovin closed this Aug 24, 2021

Run integration tests against hyperion #3334

Run integration tests against hyperion #3334

Uh oh!

Conversation

mxstbr commented Jun 15, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

spectrum-bot bot commented Jun 15, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mxstbr commented Jun 15, 2018

Uh oh!

mxstbr commented Jun 15, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

brian-mann commented Jun 15, 2018

Uh oh!

mxstbr commented Jun 16, 2018

Uh oh!

brian-mann commented Jun 16, 2018

Uh oh!

mxstbr commented Jun 18, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mxstbr commented Jun 25, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mxstbr commented Jun 25, 2018

Uh oh!

mxstbr commented Jun 25, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mxstbr commented Jun 25, 2018

Uh oh!

mxstbr commented Jun 25, 2018

Uh oh!

NekR commented Jun 25, 2018 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

mxstbr commented Jun 15, 2018 •

edited

Loading

spectrum-bot bot commented Jun 15, 2018 •

edited

Loading

mxstbr commented Jun 15, 2018 •

edited

Loading

mxstbr commented Jun 18, 2018 •

edited

Loading

mxstbr commented Jun 25, 2018 •

edited

Loading

mxstbr commented Jun 25, 2018 •

edited

Loading