-main: accept an optional classpath argument

[vemv]

By accepting a fixed user-provided string as the classpath, one can be sure that lein-nvd is not interfering in classpath computation and therefore one prevents false positives/negatives.

Fixes #46

Also gives a way to easily solve:

#50
#73
#74

[vemv]

PR ready

tests are failing, but so is master (locally and also shown in 849d739)

I did verify that the proposed command works, in both its Lein and deps.edn variants

[vemv]

Amended with some refinements and an assert (as a lightweight test)

[vemv]

I simplified things a bit here by using json/write-str, this way the Lein and deps.edn helpers don't have to construct escaped strings by hand

[rm-hull]

Looks fine 👍
I'll investigate the failure on master

[rm-hull]

Since travis has gone all shit, this project really could do with moving to github actions

ooh, so it was already moved across 😊

[vemv]

Alright!

Feel free to LMK when master is green again. Meanwhile I'll try to add a simple test (I think a bash script would be nice for exercising Lein + deps.edn + the new argument all in the sample place)

[rm-hull]

Main branch is now green

[vemv]

Looks like master went red again after f4a7187

[vemv]

Progressing...

The fix that would also fix master is:

    (is (== 0 (get-in project [:nvd :highest-score])))

Not really sure it's the right fix though. Originally that test had the opposite intent. But updating dependencies can change things.

[vemv]

Finally got the build green 😄 (with the caveat of #75 (comment))

Also, I used the occasion to introduce a JVM matrix in CI, so that we're sure lein-nvd runs correctly in all JVMs that Clojure officially supports (today: 8 and 11).

Similarly, the integration test script runs in a parallel job, for a faster feedback loop.

[rm-hull]

slightly puzzled why this dropped from 7.5 to 5.0 and now to zero .. happy to take this for now, but think i will create an issue to update self-test.json so we have something that is > 0