Fix CI by using ganache from docker container

[dtebbs]

It appears that something has happened in a dependency which breaks the ganache build, so that needs to be checked and addressed. AFAICT, the issue seems to be related to the @types/bn.js package, although after some investigation it's still not clear to me what could have changed.

[AntoineRondelet]

It appears that something has happened in a dependency which breaks the ganache build, so that needs to be checked and addressed. AFAICT, the issue seems to be related to the @types/bn.js package, although after some investigation it's still not clear to me what could have changed.

Thanks for fixing this @dtebbs. Going forward, please be sure to either link to the CI run/job failing, and/or provide an excerpt of the error faced so we can have a better idea of what's going on under the hood and how the proposed change resolves the issue. It's good to know why from one day to another the CI stars to fail on some code that hasn't been touched, especially as such issue may appear again in the future. Without much context, this smells like an issue with managing dependencies (either a dependency release pushing a breaking change disguised in a minor, or we use loose versioning for some of the tools that run in the CI, i.e. don't constrain a specific node JS version or something - though I know this one is properly constrained as we specify node 10...).

I can see that some CI runs failed on the update-libsnark branch (PR #407) around 9 days ago (time a which this PR was opened), e.g. https://github.com/clearmatics/zeth/runs/3872675602?check_suite_focus=true - is that related at all, or did you get another error?

[AntoineRondelet]

Without much context, this smells like an issue with managing dependencies (either a dependency release pushing a breaking change disguised in a minor, or we use loose versioning for some of the tools that run in the CI, i.e. don't constrain a specific node JS version or something).

Some docker base images will come with specific packages versions (e.g. a Python base image) or won't be supporting the latest versions of some packages etc, which may (or not) be why moving things in Docker fixes the issue.

[AntoineRondelet]

LGTM