Run CI for master branch push and PRs.

[simi]

No description provided.

[voxik]

Looking into the CI failures, specifically to rack_3 (3.2, spec_rack), I believe this is due to sinatra/sinatra@cd3e00d introduced by (Sinatra /) Rack::Protection 4.1.0, where the www.example.com is not on the list of enabled hosts.

[simi]

Thanks @voxik, I seen also some date related issues. Locally CI looks much better. I'll keep exploring, feel free to share patches (or open PR to merge into this branch) to fix any test you're aware of why is failing.

[simi]

Any idea what to use instead of example.com or better to re-configure to enable?

[voxik]

Any idea what to use instead of example.com or better to re-configure to enable?

diff --git a/lib/capybara/spec/test_app.rb b/lib/capybara/spec/test_app.rb
index b29a1523..eabb2a8e 100644
--- a/lib/capybara/spec/test_app.rb
+++ b/lib/capybara/spec/test_app.rb
@@ -19,6 +19,7 @@ class TestApp < Sinatra::Base
   set :static, true
   set :raise_errors, true
   set :show_exceptions, false
+  set :host_authorization, { permitted_hosts: [] }
 
   # Also check lib/capybara/spec/views/*.erb for pages not listed here
 

This is the most simple workaround I have been able to come up with (working on PR right now). This just disables the check.

Or the check can be more specific, e.g. set :host_authorization, { permitted_hosts: [".example.com", "127.0.0.1"] }.

Just FTR, the 127.0.0.1 is there to pass the test in https://github.com/teamcapybara/capybara/blob/master/lib/capybara/spec/session/current_url_spec.rb while it seems weird to me 🤷

[voxik]

Already looks better 👏 At least something is passing 🥳

[voxik]

This is likely issue for the rack_smoke (3.0, cucumber)

[simi]

Ahh, I think we can remove 3.0. WDYT? Or worth to keep?

[voxik]

Ahh, I think we can remove 3.0. WDYT? Or worth to keep?

For me personally (and looking into this due to Fedora), I don't mind removing the 3.0 support. Nevertheless I can't say the xpath code would be more readable with just & instead of &block, so I wonder if the change is worth of it.

However, wearing my Red Hat fedora, RHEL users would likely not mind longer support for Ruby 3.0 which is default in RHEL 9 and will be supported for years. But I don't think I can stop the train (after all, RHEL 10 is out fresh) 😉

[simi]

OK, I'll revert in xpath for now. I just blindly followed rubocop in there.

[simi]

@voxik revert is not that simple, since it doesn't work with latest JRuby. I'll find out how to fix and report back.