Possible regression caught by semver's test suite

[steveklabnik]

I ported semver to HEAD of semver-parser, and I get these test failures:

---- version::tests::test_from_str_errors stdout ----
        thread 'version::tests::test_from_str_errors' panicked at 'assertion failed: `(left == right)`
  left: `Err(ParseError("expected more input"))`,
 right: `Err(ParseError("Error parsing major identifier"))`', src\version.rs:850:8
note: Run with `RUST_BACKTRACE=1` for a backtrace.

---- version::tests::test_parse stdout ----
        thread 'version::tests::test_parse' panicked at 'assertion failed: `(left == right)`
  left: `Err(ParseError("expected more input"))`,
 right: `Err(ParseError("Error parsing major identifier"))`', src\version.rs:395:8

---- version_req::test::test_ordering stdout ----
        thread 'version_req::test::test_ordering' panicked at 'assertion failed: req("=1") < req("*")', src\version_req.rs:953:8

cc @udoprog

The first two are because I'm testing the literal string output; I don't think that's actually part of the contract, and besides semver is also pre-1.0 so they don't matter. Additionally, they're actually two tests of the same thing. So really one failure.

The last one is curious though. I haven't actually debugged this one, so it might be a bug in my small port, or it might be here. needs investigation.

[steveklabnik]

If it's here, #29 would be the cause, I'd imagine.

[steveklabnik]

ah ha! so, I think I've narrowed this down; we rely on the derived implementations of PartialOrd, and so I bet #29's change changed how those comparisons work. gah.

I think that means that the bug is in semver, not here!

[udoprog]

This is fixed in dtolnay/semver#151. Ordering relative to the wildcard changes because it's modeled as an empty set of predicates instead of a predicate WildcardVersion::Major variant.

[steveklabnik]

ah ha! You're the best.