Use ganache server (instead of provider)

[cgewecke]

Begins using ganache.server (instead of provider) so other processes can connect to the coverage run. This should make it possible to use the Oraclize/Provable testing client, as well as simplify Zeppelin's installation of 0.7.0. (See OZ 1918)

• Also adds support for --network <name> cli flag and tests for all the edge cases introduced by that.
• Seems to take ~5-7 minutes off the zeppelin run (2nd time have seen this), suggesting that there is a big perf difference between the docker (zeppelin's CI) and linux environments (CircleCI machine) for provider.

[codecov-io]

Codecov Report

Merging #402 into truffle-plugin will increase coverage by 0.71%.
The diff coverage is 97.26%.

@@                Coverage Diff                 @@
##           truffle-plugin     #402      +/-   ##
==================================================
+ Coverage              96%   96.72%   +0.71%     
==================================================
  Files                  11       11              
  Lines                 576      610      +34     
==================================================
+ Hits                  553      590      +37     
+ Misses                 23       20       -3

Impacted Files	Coverage Δ
dist/plugin-assets/truffle.ui.js	100% <ø> (ø)	⬆️
lib/ui.js	100% <ø> (ø)	⬆️
lib/collector.js	100% <100%> (ø)	⬆️
dist/truffle.plugin.js	100% <100%> (ø)	⬆️
lib/app.js	96.15% <91.66%> (+2.23%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 77628b0...04149dc. Read the comment docs.

[yxliang01]

@cgewecke It's interesting that you observe huge performance difference between dockerized environment and without docker environment. May I know what the reason you think is?

[cgewecke]

@yxliang01 I am very surprised by this myself.

There are additional notes on this topic at:

• ganache-core 481 asking about why ganache might be slower when it's run in the same memory space as mocha vs. in a separate process connecting over http.
• solidity-coverage #397

The performance difference noted above can be seen in the comparison of these three jobs:

• Zeppelin w/ in-process provider (docker: circleci/node:10): 11:07
• Zeppelin w/ in-process provider (ubuntu: circle/machine): 23:04* and 19:10*
• Zeppelin w/ in-process server (ubuntu: circle/machine): 14:40* and 15:28*

* Using the raw unix time numbers for the last two items (e.g minus the installation times) so the comparison with Zeppelin's docker is the same.

Circle's docs on the default machine specs. I've also seen ganache.provider run much slower than expected on Travis so suspect those two are the same kind of environments.

@yxliang01 Do you have any thoughts about this? If you think of a reason or find out please lmk.

[yxliang01]

Sounds very interesting. I'll have a look and share my thoughts if any time using my free time later :)