Lint for unused package defined in pubspec

[kevmoo]

No description provided.

[seaneagan]

There may be false positives with dependencies solely depended on for their barback input assets.

[srawlins]

pub publish does this now. Of course, it'd be nice to know sooner than pub publish... Come to think of it, why doesn't pub get do this?

[bwilkerson]

I would prefer to see this as a hint (we are now analyzing pubspec files). It's analogous to the dead code hints and I think it should be treated the same way.

[pq]

@bwilkerson : shall I move this to the SDK repo?

[kevmoo]

More important is to have a lint to warn about used packages that are unreferenced in the pubspec. This can cause real problems. Be careful, about the other way around. Folks can reference resources in packages via package URI or (for things like build_runner and build_web_compilers) have good reason for a dependency without referencing it in code!

…

On Sat, May 19, 2018 at 3:39 PM Phil Quitslund ***@***.***> wrote: @bwilkerson <https://github.com/bwilkerson> : shall I move this to the SDK repo? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#57174>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABCirYIvJqQ22T2-peNXwvH3aiIu6Lzks5t0J87gaJpZM4Dgjli> .

[kevmoo]

CC @nex3, who has thought about this a bit...

[nex3]

We definitely shouldn't warn people for packages that are in their pubspec but aren't imported. We want users to declare explicit dependencies on packages whose assets they use, whose executables they invoke, that they load dynamically via isolates, and so on.

[bwilkerson]

It sounds like the request for this lint has been withdrawn. A request for a hint for imports of packages that are not explicitly listed in the pubspec should be opened in the sdk repo (if it isn't there already), but this issue should be closed.