CLI accepts invalid increment specifiers with -i/--increment and treats them as 'patch'

[mklement0]

E.g.:

semver -i nonesuch 1.0.0

outputs 1.0.1 and sets exit code to 0; i.e., it acts as if patch (or no argument) had been specified.

[lukekarrys]

This is working by design. The default value of -i is patch and invalid versions are filtered out of the input.

[mklement0]

Quietly accepting invalid input is problematic.

Another way of putting it: if this is by design, please reconsider the design.

Consider reporting an error if an invalid -i argument is specified.

[lukekarrys]

@mklement0 Thanks for the feedback. I closed this issue as a first pass through this repo, so maybe "by design" was the wrong choice of words. The code makes it clear that any invalid values passed to -i are ignored but I agree that erroring is a good idea.

For this project I believe in a very cautious approach to versioning, so I think this would be a breaking change so I'm going to tag it as such. I don't have any immediate plans to release a major version but surfacing this as a breaking change will help to batch this in when it's time.

[pjohnmeyer]

@lukekarrys would it be acceptable in the current v7 to add a console.warn about what is happening, and that this behavior might go away in a future version? This preserves the current behavior (including the ability to redirect or pipe the output) whilst still explaining what is happening.

[pjohnmeyer]

@owlstronaut are you sure we want this closed? The changes I made in #859 are only a stepping stone to the v8 changes being tracked.

[owlstronaut]

@owlstronaut are you sure we want this closed? The changes I made in #859 are only a stepping stone to the v8 changes being tracked.

oh thanks :) things are hectic 😆