ses-ava vs install-ses import order matters when it should not.

[erights]

The three line import sequence seen at

agoric-sdk/packages/eventual-send/test/test-ses-ava-reject-deep-stacks.js

Line 1 in bf407ae

import { wrapTest } from '@agoric/ses-ava';

import { wrapTest } from '@agoric/ses-ava';
import '@agoric/install-ses';
import rawTest from 'ava';

happens to work. However, the more sensible import order

import '@agoric/install-ses';
import { wrapTest } from '@agoric/ses-ava';
import rawTest from 'ava';

does not. For reasons I do not understand, the latter import order causes the esm initializing logic at https://github.com/feross/buffer/blob/795bbb5bda1b39f1370ebd784bea6107b087e3a7/index.js#L816 to happen after lockdown, at which point it encounters the override mistake and fails. By importing ses-ava first, that initialization happens before lockdown successfully.

All of this happens before any of the code in ses-ava is actually executed. It seems to happen as a result of how the import order affects the initialization of other things. We did not see this before ses-ava which seems to suggest that there's something about ses-ava that is triggering this. But maybe we've just been getting lucky on import order till now. More experiments needed.

[Tartuffo]

@erights This is in the "Up Next" pipeline, but does not have a MN-1 label. If it is needed for MN-1, please label, otherwise move from "Up Next" to "Product Backlog".

[erights]

Hi @turadg , added you as it relates to other things you want to change about importing ses-ava.

[erights]

See endojs/endo#2133