Local testing fails with KeyError: 'development'

[trentmc]

Background / motivation

When I follow the usual flow to (a) get barge going, then (b) run any test on df-py, in (b) I immediately get KeyError: 'development'.

Digging further, it seems that barge is the culprit, not df-py. I'm putting here in df-py because it's blocking me in df-py work.

By default, barge is supposed to...

1. compile contracts
2. create address.json in ~/.ocean/ocean-contracts/artifacts/
3. deploy Ocean contracts to ganache
4. update address.json with info from (3), with top-level key "development"

It does (1) and (2), good. But not (3) and (4). Even if I wait 15+ minutes.

It's easy to inspect: just run barge, and see that it's missing "development" in address.json.

Here's my output, from when I run barge via ./start_ocean.sh --with-thegraph 1>out.txt 2>&1 &).

Here's a log from a second run - out2.txt. I'd done all the cleaning, including rebooting my laptop.

Here's some interesting output in both logs.

• Note that right after "Deploying OceanToken" it says "NomicLabsHardhatPluginError: You are trying to create a contract factory from an artifact, but you have not passed a valid artifact parameter".
• Also note what's missing: if it had deployed the addresses, it would have outptu them to the log (iirc)

ocean-contracts_1  | deploy contracts is true
...
ocean-contracts_1  | Using RPC: http://172.15.0.3:8545
ocean-contracts_1  | Using chain 8996
ocean-contracts_1  | Use existing addresses:{}
ocean-contracts_1  | Deploying OceanToken
ocean-contracts_1  | NomicLabsHardhatPluginError: You are trying to create a contract factory from an artifact, but you have not passed a valid artifact parameter.
ocean-contracts_1  |     at getContractFactoryFromArtifact (/ocean-contracts/node_modules/@nomiclabs/hardhat-ethers/internal/helpers.js:73:15)
ocean-contracts_1  |     at getContractFactory (/ocean-contracts/node_modules/@nomiclabs/hardhat-ethers/internal/helpers.js:56:16)
ocean-contracts_1  |     at async main (/ocean-contracts/scripts/deploy-contracts.js:237:19)
subgraph_1         | Using custom ADDRESS_FILE instead of ocean-contracts npm dep

[trentmc]

Related: #547 is a similar issue but for CI. It seems to be sporadic, though. (Whereas this issue hits me all the time.)

[trentmc]

I tried locally checking out the old barge (and doing related cleanup). It didn't help :(

To use a "branch" for old barge:

git checkout 9cc3a9342a47857a0495309e787462d6c60979e6

This time I also ran with more components turned off (now identical to github actions in df-py):

./start_ocean.sh --no-aquarius --no-elasticsearch --no-provider --no-dashboard --with-thegraph > out.txt 2>

Result: log.

[trentmc]

The fix: delete ~/.ocean

:)