Should we lax JBang Resolver?

[cstamas]

Context: this issue #1890 and PR #1901

Latest release makes JBang ruthless, and will fail if you have a typo in your //DEPS, as it enforces that POM is resolvable.

But, I just figured one use case where this is not desirable: if you want to consume a JAR that has no POM deployed (as some very ancient artifacts are, or maybe some "ad hoc" remote repository). This effectively prevents consuming JARs without POMs.

Maven behavior: Maven in case of lack of POM emits a warning "no dependencies will be available" and errors out if artifact is not resolvable (ie. typo).

In short, both behaviors (strict and lenient) have their pros and cons, leniency kept Bazel isue hidden in #1890, while strictness prevents consuming artifacts without POMs. Personally, I'd keep it as is, maybe add some switch/config to make it lenient.

@quintesse @maxandersen WDYT?

[quintesse]

@cstamas would there be a way to detect that an error occurred because of this strictness (besides running it two times of course :-) )? That way we could tell the user what's happening and give them instructions on how to fix the issue or to use the --lenient flag.

Because otherwise I'd be leaning towards being lenient. JBang is a tool to make the user's life easier and I don't know how being strict gives us any advantages.

[quintesse]

Ok, I've read now more of the issues involved (eg enola-dev/enola#1040) and I think I understand it somewhat better now. In this case the POM involved didn't have a typo but was really completely wrong :-)

I'm guessing that somewhere somehow the POM dependency got turned into a "group:id:type:version" string and because the type was the version and the version was empty the result was a somewhat legal "group:id:version"?

In that case I think it's okay that we are more strict and fail.
I was afraid that we'd fail in situations that were perhaps incorrect but that people might expect to work (however wrongly), but in this case the people involved agree they made a major mistake.

[quintesse]

Btw, when you say "and will fail if you have a typo in your //DEPS" what are you referring to exactly? The issue and PR you refer to don't seem to be about typos in DEPS. Can you give an example?

[quintesse]

No matter, I've seen the discussion on enola-dev/enola#1040

[cstamas]

Yes, currently only issue JBang (may) have, is that it cannot utilize dependencies without POMs. This is the question, do we want to do anything about it?