change(web): lm-worker bundling + wrapper script rework 📜

[jahorton]

This PR exists as prep-work (on the lm-worker level) toward having alternate compile targets, such as ES6 artifacts, from Web and Web-related builds. In particular, note the goal of #10257.

Notable differences within this PR:

1. All "wrapping" functionality is handled by the build-wrapper.js script.
   • Sourcemaps are now always emitted by build-polyfiller.js; before, the adjustment was made here, rather than when wrapping.
   • Regardless of compile target, inclusion or exclusion of sourcemaps is now specified at the "wrapping" stage.
2. All lm-worker .js build scripts are now parameterized, rather than having specific source and output files hardcoded.
   • This means that all paths are now specified and editable within the main build script, rather than in the helpers.
   • The resulting generalization will greatly facilitate code re-use for alternate compile targets.
   • I've given two of the three scripts a builder-like interface, including help documentation.
     • The excluded one? It gets overhauled in the next PR.

@keymanapp-test-bot skip

[keymanapp-test-bot]

User Test Results

Test specification and instructions

User tests are not required

[mcdurdin]

LGTM, suggestions are only minor, I don't mind the wetness of the doHelp and parameter parsing at this point. Could consider using commander or some other command line parsing library but it hardly seems worth it at present.

[mcdurdin]

This looks like it might be flipped in 18.0? Do we need an issue to track this?

[jahorton]

Assuming I'm properly understanding you here, we will not need an issue to track this. From #10257:

keyman/common/web/es-bundling/src/configuration.mts

Lines 24 to 35 in dc7bfc1

	export function forES6(config: esbuild.BuildOptions): esbuild.BuildOptions {
	return {
	...config,
	target: "es6",
	conditions: ['es6-bundling'],
	// Even if they'd be tree-shaken out, esbuild will fall over from their presence
	// within an imported `@keymanapp/common-types`.
	external: ['timers', 'buffer', 'events'],
	// Available with ES6, but not necessarily with ES5.
	keepNames: true
	};
	}

The ES6 target's configuration will re-enable this.

[mcdurdin]

Will we ever want keepNames: false?

[jahorton]

It's not viable on Android 5.0 devices, so yes.

https://caniuse.com/mdn-javascript_builtins_function_name_configurable_true

[jahorton]

Prepping additional code-suggestions to apply with the others...

[jahorton]

Assuming I'm properly understanding you here, we will not need an issue to track this. From #10257:

keyman/common/web/es-bundling/src/configuration.mts

Lines 24 to 35 in dc7bfc1

	export function forES6(config: esbuild.BuildOptions): esbuild.BuildOptions {
	return {
	...config,
	target: "es6",
	conditions: ['es6-bundling'],
	// Even if they'd be tree-shaken out, esbuild will fall over from their presence
	// within an imported `@keymanapp/common-types`.
	external: ['timers', 'buffer', 'events'],
	// Available with ES6, but not necessarily with ES5.
	keepNames: true
	};
	}

The ES6 target's configuration will re-enable this.

[jahorton]

iOS build failure was due to the need to sign a license agreement. No need to hold this up for that.

There was a temporary merge conflict upon merging #10255, but it was auto-resolvable on desktop.

[keyman-server]

Changes in this pull request will be available for download in Keyman version 17.0.239-alpha