Comment 13 for bug 1470265

> 1. we should make snappy build fail instead of producing packages that cannot be installed.
...
> on 1. i would suggest that review tools clearly separate tests that check if package is valid to install from those that implement our store policy (e.g. apps with changes to sandbox don't get auto approved).

In general, the tests are separated in this manner-- there are 'lint' tests, 'security' tests, etc in different files. For example, in this case, the test that failed is a lint test (eg, from John's example, "lint_hooks_env_what_valid malformed application name: 'env_what'").

Also note, I think it is vitally important for snappy build to produce correct binaries. I also think snappy install should guard against malformed input (especially since it is run as root! :), but I also think the review tools are an important gateway to block crafted snaps from ever hitting devices. Regardless of if '_' should be valid or not, IMO the review tools should be re-enabled in snappy build and there should be a commit policy for snappy that does a 'snappy build' for any changes to yaml and a corresponding MP to the review tools as needed. I discussed this previously in meetings on IRC-- there is agreement on the approach and this work is already started and in a trello card.