Upgrade JS package dependencies (yarn upgrade).

[creachadair]

This change is 

[creachadair]

Unfortunately, it looks like I'm going to have to split the Babel upgrade from the rest of this change; there are just too many AST changes and no clear documentation as to what the API guarantees.

[bzz]

it looks like I'm going to have to split the Babel upgrade from the rest of this change;

🙏 and there conveniently is a #47 for it

[creachadair]

I surveyed the diffs between the existing native fixtures and those generated by the new version of the parser. As far as I can tell, all of the differences are the addition/removal of one or more of the following properties:

extra: { parenthesizedArgument: bool }
inexact: bool
internalSlots: []
implements: []
inexact: bool
interpreter: ~
leadingComments: ~
method: bool
proto: bool
trailingComments: ~
typeArguments: ~

This seems to be a straightforward metadata extension of some kind, and I think it should be safe to update the fixtures on that basis. @dennwc do you think that is reasonable?

[creachadair]

it looks like I'm going to have to split the Babel upgrade from the rest of this change;

🙏 and there conveniently is a #47 for it

Unfortunately, while splitting that part of the change is appealing for reasons of organization, there is (it turns out) no easy way to cut that one piece out: Upgrading Babel also pulls in a whole bunch of other updates, which is more or less what this change does anyway. And those updates are the cause for the rest of the changes in this PR already. Worse, I found that if I leave the old version pinned in place I can't update the packages I need to update to fix the upstream bug. 🤷🏻‍♀️

So I think in practice it will not save anything to separate them.

[creachadair]

I have updated the fixtures for the output changes described above. The errors that remain are because the parser is now more strict than it used to be, and will complain for invalid structures like re-definitions of existing names, and exports of non-existing names.

Our existing parser logic has some custom hooks to "guess" which mode to run the parser in ("module" or "script"). An error in the first falls back to the second. This means when we have an invalid file in module mode, we also get an error from script mode. But since script mode cannot handle newer ES constructs like import and export, that will always fail. I believe the current version of Babel can handle this guesswork by itself.

My plan for now is to make the fixtures "valid", and to address the fallback weirdness in a separate PR.

[creachadair]

I believe this is now ready for a review. While a lot of files are involved, most of the changes are mechanical. The changes that should be reviewed closely are:

1. The files under the native directory, including the version locks, parser.js, and rollup config. This is the substance of the change, updating the dependencies.

2. The fixture changes to fixtures/import-declaration.js, fixtures/export-declaration.js, and fixtures/u2_import_subsymbol.js, together with their associated golden files. These files needed updates for increased strictness in the @babel/parser API.

The commits in this PR separate these changes as much as is practical, so you may find it easier to review commit-by-commit. I will do my best to keep that structure during the review.

[dennwc]

Adding fields may break existing annotations though. CI is green which is a good sight, but let me review all the fixtures just in case. For example, the disappearance of leadingComments is worrying - those are used in our annotations. But they might be optional.

Reviewed 5 of 5 files at r2, 72 of 381 files at r3.
Reviewable status: 77 of 386 files reviewed, all discussions resolved (waiting on @bzz and @dennwc)

[dennwc]

,leadingComments looks safe, we drop it on our side anyway.

Also, I expected something more disruptive in this update. They mostly wiped empty fields and added a few new ones. Also, I've noticed they now differentiate between "module" and "script" file types better.

It may make sense to add fixtures that cover the new fields later. Looking at the diff, all those fields are empty, so we are missing some exotic fixtures, it seems.

Reviewed 309 of 381 files at r3.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @bzz)