Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Add support for -unstable gopkg.in imports #587
dep-test - master! ❯ dep ensure -v gopkg.in/mgo.v2-unstable dep: No constraint or alternate source specified for "gopkg.in/mgo.v2-unstable", omitting from manifest Root project is "scratch/dep-test" 1 transitively valid internal packages 2 external packages imported from 2 projects (0) ✓ select (root) (1) ? attempt github.com/pkg/errors with 1 pkgs; at least 1 versions to try (1) try firstname.lastname@example.org (1) ✓ select email@example.com w/1 pkgs (2) ? attempt gopkg.in/mgo.v2-unstable with 1 pkgs; 1 versions to try (2) try gopkg.in/mgo.v2-unstable@v2-unstable (2) ✓ select gopkg.in/mgo.v2-unstable@v2-unstable w/5 pkgs ✓ found solution with 6 packages from 2 projects Solver wall times by segment: b-list-pkgs: 637.471037ms b-gmal: 398.518303ms b-source-exists: 156.043538ms satisfy: 550.358µs select-atom: 358.518µs new-atom: 240.444µs select-root: 83.551µs b-deduce-proj-root: 15.934µs other: 14.055µs b-list-versions: 11.312µs TOTAL: 1.19330705s dep-test - master! ❯ cat Gopkg.lock memo = "842e27492a249b7020fca265003afa2ef25e260ebaae9e0242ccb02decac74ef" [[projects]] name = "github.com/pkg/errors" packages = ["."] revision = "645ef00459ed84a119197bfb8d8205042c6df63d" version = "v0.8.0" [[projects]] branch = "v2-unstable" name = "gopkg.in/mgo.v2-unstable" packages = [".","bson","internal/json","internal/sasl","internal/scram"] revision = "9a2573d4ae52a2bf9f5b7900a50e2f8bcceeb774"
If this is the right approach, I'll have @sdboyer push an -unstable version to the gopkt testbed instead of using mgo, which was just an example from the wild.
I'm not really sure of the semantics of "unstable", so this could be totally off. It does, however, fetch the appropriate code and allow you to
It does, however, seem in line with how we treat "normal" gopkg.in imports:
May 16, 2017
I think you just need to push an -unstable branch to
Yep, I think we're good to go here.
We may eventually want a follow-up where we DO allow semver tags when unstable is set, but only e.g. if they have a non-empty prerelease property. I'd rather not do that until an actual human runs into that problem, though.
May 22, 2017
i think they've all been useful
It's not, but only because it has just the two numerical components.