Run E2E tests from correct working directory

[rmunn]

Description

E2E tests assume they are being run from root of repo, not from docker directory, so they need working_directory: . in the GHA workflow file.

Fixes #1369.

Type of Change

Only keep lines below that describe this change, then delete the rest.

• Bug fix (non-breaking change which fixes an issue)

[github-actions]

Unit Test Results

    1 files      1 suites   14s ⏱️
373 tests 373 ✔️ 0 💤 0 ❌

Results for commit 39a814e.

♻️ This comment has been updated with latest results.

[megahirt]

Great thanks, @rmunn This should fix the issue that @longrunningprocess identified with E2E tests not running on integrate-and-deploy

[longrunningprocess]

Thanks for taking a look at this and I'll pull this into my branch on my next rebase. I'd like to note however, had this followed convention and put those commands behind a make target like make e2e or something, we could've avoided this. Most likely in the make target there would've been a cd .. and then all the necessary commands needed to make it work.

[longrunningprocess]

actually looking at the Makefile, I see there are still targets in there... maybe we should've tweaked those or added a new one make e2e-tests-new if the old one is still being used anywhere?

[megahirt]

Yes, I still want to keep the old ones in make e2e-tests as I can still run those on my machine locally. Yes, we should have a make target, probably make playwright-tests. Or we could rename the old ones to make protractor-tests and the new ones be make e2e-tests.

Yes, it would be better to use Make in the build script. There's a number of places where we aren't doing that since there are a few small differences - but I think those could be ironed out.

My goal after the lfnext branch is merged, is to move the Makefile to the repo root. No more cd docker for Make commands.

[longrunningprocess]

ok, that sounds good

[rmunn]

Thanks for taking a look at this and I'll pull this into my branch on my next rebase. I'd like to note however, had this followed convention and put those commands behind a make target like make e2e or something, we could've avoided this. Most likely in the make target there would've been a cd .. and then all the necessary commands needed to make it work.

Unfortunately, the way Playwright's webServer config works is that it registers the process ID of the command that runs, and kills that process ID when it's done — which means that I couldn't use make e2e-tests (I tried that first). I'm not always good at concisely explaining all the technical details behind my decisions, which means that sometimes I decide not to write an essay and to just do things; sorry about that.

[longrunningprocess]

gotcha, thanks for at least trying!