Fix race conditions(?)

[wout]

I'm mildly optimistic about this update. I haven't seen any errors so far.

SESSION_RETRY_LIMIT

The biggest issue I think was the @retry_limit instance variable in the selenium driver class. Any delays in the setup, after initialization and before the actual tests were run, would eat into the time span. Now the full time span is considered.

The wait_for_ready method

Waits for the DOM to be ready. That's just a precaution,

Navigate to about:blank on reset

Just to make sure no remnants of the previous test are left. Essentially always starting with a blank slate.

LuckyFlow.wait_for_server

This is a small addition to apps to make sure the app's http server is ready to receive requests. So the spec/setup/start_app_server.cr file now looks like:

app_server = AppServer.new

spawn do
  app_server.listen
end

LuckyFlow.wait_for_server

Spec.after_suite do
  LuckyFlow.shutdown
  app_server.close
end

[jwoertink]

😱 omg, will this fix the issues I was having that I mentioned in luckyframework/lucky_cli#890 ? That would be amazing 🙌

[wout]

I don't know, that's why I added the question mark to the title 😄. This could fix it, but you have to see for yourself. So pull in my branch before merging it in. At least this makes sure that everything is ready and set up before actually testing, and that there's nothing left in the DOM from previous tests, which could also be a source for weird errors.

[jwoertink]

May potentially (hopefully) fix luckyframework/lucky_cli#883 .. 🤞

[jwoertink]

I can't tell if this made it better or not... but there was still a CI failure

https://github.com/luckyframework/lucky_cli/actions/runs/24057314088/job/71300227677?pr=925#step:10:133

However, there was only 1 failure, so it's possible this made a difference. I'm just not sure how to tell 🤔

[wout]

Hmm, I haven't seen any so far but this looks like a timing issue again. I'll have a look at it today.

Are you seeing less errors than before? Or about the same?

[wout]

Another tiny fix based on the error you linked, particularly this line:

https://github.com/luckyframework/lucky_cli/actions/runs/24057314088/job/71300227677?pr=925#step:10:140

It may be that the references from the driver.find_css(selector) call got stale by the time element.text is called.

This is a bit like shooting in the dark 😄, but if we keep tracing them down I suppose we'll get there eventually. I know it's not the prettiest fix, but at least it rules out the timing issue here.

[jwoertink]

oh snap! luckyframework/lucky_cli#925 It's green! I think this may have worked :D

[jwoertink]

I say we give it a shot. Can't be worse than what exists now where Flow specs don't even run 😂 The fact that I was able to take existing specs and just run them and get all green is very promising 🙌