feat: use aegir for build/testing/release

[RangerMauve]

This is part of the overall ESM migration. ipld/ipld#224

We're removing building cjs on publish and are replacing a bunch of our existing build tooling to publish just ESM with typescript declarations using aegir

TODO:

• Check that release scripts will still work
• Update github actions with new changes (standardize actions accross repos?)

[RangerMauve]

I'm getting the following errors when I try to lint:

/home/mauve/programming/js-multiformats/src/block.js:25:11: Blocks are nested too deeply (5). Maximum allowed is 4. [Error/max-depth]
/home/mauve/programming/js-multiformats/src/block.js:59:11: Blocks are nested too deeply (5). Maximum allowed is 4. [Error/max-depth]
/home/mauve/programming/js-multiformats/test/test-cid.spec.js:168:5: Unexpected 'todo' comment: 'TODO: after i have a keccak hash for the...'. [Warning/no-warning-comments]
/home/mauve/programming/js-multiformats/test/ts-use/src/main.ts:0:0: Parsing error: "parserOptions.project" has been set for @typescript-eslint/parser.
The file does not match your project config: test/ts-use/src/main.ts.
The file must be included in at least one of the projects provided. [Error]

Any tips for what I should do about the ts-use section?

For the "blocks are nested too deeply" bit, I was thinking of splitting out some code into separate functions, but the use of yield makes it harder. Should I just disable the rule for that file?

What should I do about the TODO comment?

[RangerMauve]

TODO: Compare the types published from aegir with old data.

[rvagg]

such a hard pill to swallow, and one of the primary reasons I've resisted aegir to date:

added 478 packages, and audited 479 packages in 11s

$ du -sh node_modules/
162M    node_modules/

which is bad enough

to:

added 1402 packages, and audited 1613 packages in 1m

$ du -sh node_modules/
367M    node_modules/

[rvagg]

Fixed up the linting complaints and pushed to this branch, @RangerMauve you may want to pull rebase for this.

A quick look at the types differences:

• Lots of replacements for Object to `object
• Lots of minor comment linting
• BlockEncoder's encode goes from encode(data: T): ByteView<T>; to encode: (data: T) => ByteView<T>;
• BlockDecoder's decode goes from decode(bytes: ByteView<T>): T; to decode: (bytes: ByteView<T>) => T;
• MultihashHasher's digest goes from digest(input: Uint8Array): Promise<MultihashDigest> | MultihashDigest; to digest: (input: Uint8Array) => Promise<MultihashDigest> | MultihashDigest;
• SyncMultihashHasher's digest goes from digest(input: Uint8Array): MultihashDigest; to digest: (input: Uint8Array) => MultihashDigest;

I think those method definition changes are fine. The Object -> object thing might be worth digging in to further since I think that's a narrowing of the types that can be fed in as arguments.

Also, pushing to this branch and making the tests pass resulted in a release! So something's wonky and I'll need to push a patch release to overwrite it.

[rvagg]

Fixed up the release thing, there was an error in the release block in Actions:

diff --git a/.github/workflows/js-test-and-release.yml b/.github/workflows/js-test-and-release.yml
index 4c19afa..76f568f 100644
--- a/.github/workflows/js-test-and-release.yml
+++ b/.github/workflows/js-test-and-release.yml
@@ -126,7 +126,7 @@ jobs:
   release:
     needs: [test-node, test-chrome, test-chrome-webworker, test-firefox, test-firefox-webworker, test-electron-main, test-electron-renderer]
     runs-on: ubuntu-latest
-    if: github.event_name == 'push' && github.ref == 'refs/heads/esm-migration' # with #262 - 'refs/heads/${{{ github.default_branch }}}'
+    if: github.event_name == 'push' && github.ref == 'refs/heads/${{{ github.default_branch }}}'
     steps:
       - uses: actions/checkout@v3
         with:

I've pushed a 9.8.1 to restore the 9.7.1 code.

[rvagg]

So, a couple of things to note as I drive by this:

• npm test doesn't lint for me, I have to run this separately. Is this the pattern the other aegir projects are taking? I personally quite like having lint & test bundled in to one but am happy to stick to pattern if that's how it needs to be.
• When we briefly had the 9.8.0 in the wild, it came back in as a dependency for master (because a few packages used for test pull in multiformats using ^ semver ranges) but I was getting loader errors about not finding the ./basics export when trying to run npm test. It could just be the way things were arranged but seemed suspicious to me because I think npm test should be doing ESM imports and the package.json here has a "./basics" which has an "import", so I'm scratching my head about that one and wondering if it's going to be a problem when we publish. Maybe a pre-release before we push this out would be a good thing to do so we can properly test integration with other packages instead of just simulating it (although simulating would certainly be a good idea too).
• The .github/workflows/js-test-and-release.yml seems to suggest that tests are only going to be run on ${{{ github.default_branch }}} (when the hardwired esm-migration is removed)—is this correct? Shoulnd't we be running them on all branches? There's already a clause (fixed above) for not doing release unless it's on master so that shouldn't be a problem.

I'm going to try and get #197 and this one bundled together for a v10.0.0 release; somehow, still think through that process.

[rvagg]

OK, I've "merged" this with #197 into a proposal/v10.0.0 branch, PR @ #199. So I'll close this one and we can PR incremental changes against proposal/v10.0.0.

I just have a weird and annoying browser failure now, it doesn't seem to be reading the "browser" export for hashes/sha2 but I'm not sure what would have stopped that from working. It works on this branch but not on proposal/v10.0.0.