[DO NOT MERGE] Compass web e2e [PoC] WRITING-15921 #5355
Conversation
| (process.env.MONGODB_VERSION?.endsWith('-enterprise') && 'yes') ?? 'no'; | ||
|
|
||
| // should we test compass-web (true) or compass electron (false)? | ||
| export const TEST_COMPASS_WEB = process.argv.includes('--test-compass-web'); |
There was a problem hiding this comment.
I don't know what to call this thing. We can easily find/replace it with a better name. Not sure if it is best to calculate it based off an env var or a cli parameter. We need to use this both in mocha tests and in helpers and having it as a const you can import seems like the cleanest of the options I can think of.
| browser: CompassBrowser | ||
| ): Promise<void> { | ||
| const connectScreenElement = await browser.$(Selectors.ConnectSection); | ||
| // TODO: can we not just make this the same selector in both cases somehow? |
There was a problem hiding this comment.
Good point, we might be able to use connection-form in the sandbox, this should make this stuff more unified even if connection form will not end up in data explorer
| ): Promise<void> { | ||
| const tab = browser.$(Selectors.collectionSubTab(tabName)); | ||
| const selectedTab = browser.$(Selectors.collectionSubTab(tabName, true)); | ||
| const tab = Selectors.collectionSubTab(tabName); |
There was a problem hiding this comment.
We have a lot of cases of this bug. in this case tab is a selector (ie. a string) and browser.clickVisible(selector) takes a selector, not an element. And in quite a few places we mistakenly sent it either an element or a promise that will resolve to an element. I'm wondering if we can't just make browser.clickVisible() take either a selector OR an element. OR better yet maybe we can have element custom commands and not just browser ones.
| const sidebarNavigationItem = browser.$( | ||
| Selectors.sidebarInstanceNavigationItem(tabName) | ||
| ); | ||
| const sidebarNavigationItem = |
There was a problem hiding this comment.
Again - sidebarNavigationItem was a promise that would resolve to an element which is probably not what we want to pass to clickVisible.
There was a problem hiding this comment.
This is specifically not awaited so that clickVisible can re-request element from a selector instead of trying to click an element that was potentially removed / reinserted in DOM. Like an element matching .foo is in the DOM, but awaiting here would keep a reference to a specific DOM node that is gone already. In webdriver types this is referred to as ChainablePromiseElement. Awaiting a specific element and passing it around was causing issues in a few places, so this is why clickVisible was extended to support this. If it's not an issue anymore, feel free to remove it, but then you'd probably want to remove the ability to pass ChainablePromiseElement to the method itself too
There was a problem hiding this comment.
I relented and just added ChainalePromiseElement support everywhere.
There was a problem hiding this comment.
Btw this is usually why I just pass a selector and would re-select based on the selector. Otherwise you inevitably end up with a stale element eventually. I'm not 100% sure if ChainablePromiseElement won't end up caching the element id internally somehow anyway.
Also we have sooo many webdriver warnings flying past about stale elements in version 8 now.. 😬
| Selectors.DuplicateViewModalConfirmButton | ||
| ); | ||
| confirmDuplicateButton.waitForEnabled(); | ||
| await confirmDuplicateButton.waitForEnabled(); |
There was a problem hiding this comment.
Something about webdriverio's old types caused typescript to not notice that this is a promise that wasn't being awaited. Now it works as expected with the latest webdriver.
| capabilities: { | ||
| browserName: 'chrome', | ||
| browserName: 'chromium', | ||
| browserVersion: '120.0.6099.216', // TODO: this must always be the corresponding chromium version for the electron version |
There was a problem hiding this comment.
This kinda sucks. We have to set this otherwise webdriver now goes down a path where it tries to use the electron version as the chromium version and then gets the chrome driver version for that version.. Electron always includes the chromium version in their release notes. Maybe there's a way to programmatically get this? https://releases.electronjs.org/release/v28.1.4
There was a problem hiding this comment.
$ echo 'console.log("CHROME_VERSION=" + process.versions.chrome)' | npx electron -i 2>&- | grep -o 'CHROME_VERSION=.*'
CHROME_VERSION=120.0.6099.216could probably adapt this into a JS script fairly easily
| await browser.waitUntil(async () => { | ||
| const element = await browser.$(selector); | ||
| await element.waitForDisplayed(); | ||
| await element.click(); // focus |
There was a problem hiding this comment.
Looks like who knows how many subtle things like this broke in the webdriverio upgrade. await element.clearValue() now does nothing and await element.setValue() appends to the text that's already in the form field's value...
lerouxb
force-pushed
the
compass-web-e2e
branch
from
c4a996d to
828d099
Compare
lerouxb
force-pushed
the
compass-web-e2e
branch
from
08d064b to
bdc20ce
Compare
lerouxb
changed the title
3 participants
Skipping everything but the "time to first query" tests, left lots of TODOs.