Upgrade web3.js for latest node.js (<11) support

[schmidsi]

The scrypt packages which is a dependency of web3-eth (and others) does not work with node.js >=11.15. This issue is already solved in web3/web3.js#2938, so theoretically it should be sufficient to just update web3 to the latest 1.x version (which happens in this pull-request). Unfortunately, after this update, some of the tests fail :(

I did not have time to find the root cause of this, but just wanted to inform about my failing attempt.

Maybe in the meantime, add something like this to package.json?

    "engines": {
        "node": ">=8.0.0 <=11.15.0"
    },

I can confirm that it works with [email protected] (LTS)

[CLAassistant]

All committers have signed the CLA.

[sohkai]

Thanks @schmidsi, I'll look into why the tests are failing!

I think with web3.js having a proper v1 we can also remove the pin, since that was a fix for more unpredictable releases of web3.js previously :).

[schmidsi]

I think with web3.js having a proper v1 we can also remove the pin, since that was a fix for more unpredictable releases of web3.js previously :).

As far as I know, web3 v1 is pretty stable: https://medium.com/@samuel_91690/1-0-release-web3-js-ddd23d3c8f62

[sohkai]

Yep, exactly :).

[coveralls]

Coverage increased (+0.01%) to 90.85% when pulling 7b906f8 on schmidsi:upgrade-web3 into 1dc05b6 on aragon:master.

[sohkai]

@schmidsi I've tracked down the issue to web3's ABIEncoderV2 implementation (circa beta.36 IIRC), where it now assumes all non-named inputs in a function are tuples.

I'm not sure where the rationale came about, and it does seem to be fixed in web3.js@2, but setting a non-empty name value fixes the issue for now.

I've also unpinned the web3.js version, since it has more stable versioning now than before :).

Thanks for trying this out @schmidsi!

[sohkai]

Will release as a release candidate and test with a few existing examples to make sure nothing else was broken :).