ng update / init broken on an existing project since beta.31

OS

Mac OSX Sierra

Versions.

@angular/cli: 1.0.0-beta.31
node: 6.9.5

Repro steps.

During the update from beta.30 to beta.31, running ng update or ng init exits after asking me to overwrite README.md… No matter what option i choose, it just quits straight away without continuing further.

The log given by the failure.

$ ng update
installing ng2
? Overwrite README.md? Diff
$
$ ng update
installing ng2
? Overwrite README.md? Diff
$
$ ng update
installing ng2
? Overwrite README.md? Yes, overwrite
$
$ ng init
installing ng2
? Overwrite README.md? Diff
$

[coryrylan]

With ng update if you had your project under source control it was easy to update. Let ng update replace all files then with a git gui compare the changes and take the new ones that you want or leave your unchanged files. Not sure if this was the correct workflow but seemed to work pretty well.

[Meligy]

Unless I am missing something, I think I tried ng init / ng update in a beta.30 CLI project on Windows, I didn’t have readme in that project so the file it asked to replace was different (I think it was index.html), then it just ended. I looked at my git and found that the files were changed, and I was able to upgrade my project.

In saying this, I’m agreeing with @splintercode, the way I update my projects is by running ng init, accepting all core files (tslint, tsconfig, editorconfig, test.ts, karma.conf.js, package.json, polyfills, main.ts, etc.) and then going to git to undo the changes I don’t want.

Now it seems that the command is misleading, because it brings default files even to app module / component etc. I agree, but I think this has largely increased by calling it ng update. It was more predictable and making more sense to have these effects when it was called ng init.

The reasoning I was given for removing the tool and resorting to changelog instead is that the team expects no changes to these core files (or very minimal) after v 1.0 final - I honestly don’t know what to think of this reasoning, as good or bad. It looks like this command is going away a tiny bit earlier than that anyway.