"No route matched for" error when using engine routes about rspec-openapi

[k0kubun]

Thank you for reporting this. Since you already have a small reproduction code, could you also turn the issue directory into a working test case first that can be merged as is? Being able to see it as a draft PR would be even more helpful.

[k0kubun]

It looks like none of your currently open pull requests has a test case at this point. If you accidentally dropped it, could you resurrect it? I think this repository should have a test that runs a Rails engine.

[blocknotes]

It looks like none of your currently open pull requests has a test case at this point. If you accidentally dropped it, could you resurrect it? I think this repository should have a test that runs a Rails engine.

You are right, I have just updated the PR with an engine test.
Please check the PR carefully when you have some time 🙏
It looks correct to me but it's a tricky part.

[blocknotes]

Thank you for reporting this. Since you already have a small reproduction code, could you also turn the issue directory into a working test case first that can be merged as is? Being able to see it as a draft PR would be even more helpful.

Sure => #38
It contains also a change to support the case where there's no controller (ex. a route with only a block) and the case of 404 routes (without raising an exception).

BTW, I'm not 100% convinced right now of the third commit (the engine fix) because if the relevant route is the first one it could be ignored in the iterations of the recognize block.

Also, at the moment, I don't know why recognize catch more routes than expected.
Investigating a bit:

app.routes.recognize_path(request.path_info, method: request.method)
# => {:controller=>"hello", :action=>"world"}

routes = []; app.routes.router.recognize(request) { |route| routes << route }; routes.count
# => 2
routes.map(&:name)
# => ["my_engine", "hello_world"]
routes.map(&:defaults)
# => [{}, {:controller=>"hello", :action=>"world"}]

[k0kubun]

Thank you for filing it. You're still not so sure about your current approach and you expect me to think about it, right? I don't have time to do that right away, so please allow me to find time to work on it.

[blocknotes]

Thank you for filing it. You're still not so sure about your current approach and you expect me to think about it, right? I don't have time to do that right away, so please allow me to find time to work on it.

Yeah, I think that the proposed fix is not optimal.
Perhaps I already found a failing edge case with it (I'm making my tests on a project with a lot of different requests).

But I will probably look for a better solution and keep you posted.

[blocknotes]

Update: I think I found a better solution - PR updated.
I made some trials and it looks good now.

Please let me know your opinion.