Closed stevengill closed 10 years ago
I thought we were still publishing tools to npm as RCs. It was platforms that we decided against publishing as RCs since it required too many steps to update if we found a bug during RC testing.
That way we can still test the RC with npm install cordova@rc
and cordova platform add android --usegit
I'd suggest that no RCs be pushed to the real npm registry. If there is a respin, then you can't reuse the version number. I think that is more problematic on tools than platforms. If we really want to test via npm install, I'd suggest a private registry.
I agree that having a private npm registry for testing in the long run is the better option.
For now though, I disagree with not publishing RCs to npm. Why add a barrier to testing RCs? If we publish no tool rcs to npm, we would expect our users/testers to grab the tags for lib/cli/plugman/js, npm link them and then they are able to test. I'm sure we have commiters & contributors who wouldn't be able to do it properly now!
Being able to tell our users npm install -g cordova@rc to test has been a great way to get them to try it before we do an official release. Even when we setup a private npm repo to do this, it will still be one more step to get people to test for us. Adding barriers!
Worst case, we find a issue with cordova-lib, we will have to fix the version and do a minor bump and then republish as an RC. CLI & Plugman will also need to be republished. This happens infrequently enough that I don't think it is a big deal. Definitely when you consider how easy it is for people to test.
On Tue, Oct 7, 2014 at 12:40 PM, Marcel Kinard notifications@github.com wrote:
I'd suggest that no RCs be pushed to the real npm registry. If there is a respin, then you can't reuse the version number. I think that is more problematic on tools than platforms. If we really want to test via npm install, I'd suggest a private registry.
Reply to this email directly or view it on GitHub https://github.com/apache/cordova-coho/pull/49#issuecomment-58247365.
Erg, was just going through PRs and assumed the reason this wasn't merged was that you didn't want it to be :P. I do also like publishing to npm before voting so that a publish is just a switching of tag. If we need to go through many version numbers (like just recently happened), that's fine.
If folks are OK with bumping versions numbers on a respin, I'm OK with publishing to npm with the rc tag.
Due to the number of respins I did, I felt burned. But that number of respins should be rare.
Note for the passer-by: this PR is on hold until we figure out if we want to use a private staging npm repo instead (I believe)