Changes to remove extraneous HandledPromise shim dependencies

[michaelfig]

Closes #1469

We import the HandledPromise shim once-and-for-all in @agoric/install-ses. Then, using a version of SES with endojs/endo#416, it is propagated as a global to all child compartments.

Removing the extraneous HandledPromise plumbing from the code will make it easier to migrate to the new Promise.delegate API once @agoric/eventual-send/shim implements it.

[warner]

I'm surprised the tests pass if all the Compartments are now missing their HandledPromise endowment entirely. Does that work because install-ses is now changing the global Promise to really be HandledPromise?

Take a second look at install-metering-and-ses and check that the ordering of (modify globalThis.Promise) and (instrument globalThis) doesn't leave us with a non-instrumented Promise.. I suspect there might be a problem there but I haven't dived in yet.

I also noticed this downgrades the harden(E) to Object.freeze(E). Does that open up a communication channel via something like E.when.message = 'hello' ?

[michaelfig]

I'm surprised the tests pass if all the Compartments are now missing their HandledPromise endowment entirely. Does that work because install-ses is now changing the global Promise to really be HandledPromise?

It will work with endojs/endo#416

Take a second look at install-metering-and-ses and check that the ordering of (modify globalThis.Promise) and (instrument globalThis) doesn't leave us with a non-instrumented Promise.. I suspect there might be a problem there but I haven't dived in yet.

I've fixed this by ensuring the shim is installed before metering is turned on.

I also noticed this downgrades the harden(E) to Object.freeze(E). Does that open up a communication channel via something like E.when.message = 'hello' ?

I was hoping to split the restructuring changes (needed to disentangle harden from the vetted shim before lockdown) to eventual-send to a different PR, but I could do them in the current PR if that would be better.

[michaelfig]

@erights, this is of lower priority than Zoe RC. I only expect a detailed review from you on the changes to @agoric/eventual-send that were necessary in order to make it a vetted shim. Notably, we couldn't early-bind some global properties since the metering transform replaced them in the course of wrapping all the globals (including HandledPromise).

[erights]

LGTM

[michaelfig]

@FUDCo or @warner, would you be able to review the other changes (not packages/eventual-send)? The criterion is whether this is indeed a good cleanup.

[FUDCo]

Three cheers for taking out the trash!