Closed realityking closed 3 years ago
Thanks -- looks good to me!
The test failure in Node v15 is due to a change in node how peer dependencies are handled. I think this could be fixed by adding react-dom
as a peer dependency (which I think should already have been made). You could also use npm install --legacy-peer-deps
as a workaround (although I'm not sure how to configure that for Travis).
Yeah npm 7 is an pain in the ass. I wonder how tests passed on master - none of the changes in either PR really touched the React dependency. I'll have a look what can be done.
What's your take on the webpack question? Good to go with webpack 4?
I think what happened is that the node version named "node" (!) on Travis got updated from 14 to 15 at some point.
Yep, webpack 4 is fine. 👍
I thought so too but the last passing build already run with npm 7.0.8: https://travis-ci.org/github/civiccc/react-waypoint/jobs/743936673#L196
Quickly tested locally
Deviously, it works when the node_modules
folder already exists but it breaks when doing a clean install. I think this might be an npm bug.
The issue has been already been reported to npm: https://github.com/npm/cli/issues/2514
I see two possible ways out of this until this is fixed:
--legacy-peer-deps
for npmWhich one do you prefer?
@trotzig The npm bug has been fixed and CI passes. This is good to go from my perspective.
@trotzig It'd be great if you could have a look at this one :)
This one was interesting as quite a few things broke when upgrading stuff.
jasmine-core
to 2.99 was a relatively simple change. I believe the test has never worked, jasmine-core 2.7 just exposed thatjasmine-core
3I think most of these changes are no brainers except perhaps the upgrade to webpack. I think it's valuable to test with older webpack versions to assure compatibility. However webpack 4 is now almost 3 years old. I think it's fine to ignore these older versions. Especially since it uncovered tests weren't actually executed in strict mode, potentially hiding bugs.