different results than tape itself

[ljharb]

I've started a branch to try out using this package here.

However, the tests fail when they should pass: https://travis-ci.com/github/ljharb/es-abstract/builds/176444693

Oddly enough, they pass in node < 8. Any ideas?

[Raynos]

Maybe i did not implement throws properly ? it proxies to node core throws()

[ljharb]

it doesn’t call into tape’s “throws”? altho they should be the same I’m sure there’s differences.

[Raynos]

it uses assert.throws to check whether it should report a failure to tape.throws ;

[Raynos]

This avoids a bunch of ok foobar lines in TAP output by not calling into tape for assertions that passed.

[ljharb]

that seems a bit more complex than var threw = false; try { maybeThrows(); } catch (e) { threw = true; }; would that work?

[Raynos]

Yeah but not with the regex parameter.

[ljharb]

Although #4 fixes it for throws/doesNotThrow, I'm seeing a similar issue with t.equal, since node's assert.equal changed from non-strict by default to strict by default at some point. Also tape 4 uses deep-equal v1 which defaults to loose, and tape 5 uses deep-equal v2 which defaults to strict.

It seems like it'd be better to use tape's API rather than node's assert. is that something you're interested in?

[Raynos]

It would be good to use tape's API but I dont know what public or private API there available to use.

We want to run the assertions but not report them unless they fail in which case we report all of them.

I'd rather not add a direct dependency on tape though as currently the library has zero dependencies.

[ljharb]

Without a direct dependency, I'm not sure how it could be injected in a way that would work for tape, and tap, and "insert tap-producer here". If it's meant to be tied to tape's API, then it should have a dependency on tape.

[Raynos]

We take t as an argument, the current design is meant to mirror assert in node core.

If assert made a breaking change i didnt pay attention to then this module should make the same breaking change :D

[ljharb]

It's made a number.

If it's meant to mirror assert, then that's unfortunately much less useful, since assert throws and doesn't produce TAP.

[Raynos]

I didnt mean it like that, it should implement assert like methods like tap/tape do.

We can update the equal method to loose or not loose deep equal, either is fine.

[ljharb]

Right, but the way tap and tape do that varies by version, hence the need for a peer dep.