Fix stripping local debug code so can that dead-code-elimination produces smaller bundles with fewer function calls (part 1)

[NullVoxPopuli]

Goals:

• fix prod build
  • import.meta.env behavior was based on an env var (MODE), which rollup wasn't using
• inline @glimmer/debug so we can strip away via dead-code-elimination the calls to check(...)
  Turns out, this wasn't needed, and we can keep @glimmer/debug separate -- terser seems to have figured out now that functions from @glimmer/debug are not needed in the consuming packages.

---

Of note, in ember's build, the output of @glimmer/debug is changed to things like:

function CheckInterface(obj) {
  return isDevelopingApp() ? new PropertyChecker(obj) : new NoopChecker();
}

which is great!

But this keeps the check call in source, unless production mode is smart enough to see that the NoopChecker can be removed.

Maybe we should move the production checks to the check call sites if we determine they remain in production ember apps.

---

There are a number of flags that are meant to be internal to glimmer-vm and not exposed to ember.js.

There are also flags that are meant to be exposed to ember.js (which are swapped out with DEBUG from '@glimmer/env' during the build)

Output size (using dust)

	Dev	Prod
This PR	cd packages/\@glimmer du --ignore_hidden \ --reverse --apparent-size \ --filter ".+\/dist\/dev\/index.js$" \ --no-percent-bars --only-dir --depth 1 618K └─┬ . 170K ├── runtime 168K ├── syntax 104K ├── compiler 63K ├── opcode-compiler 23K ├── manager 19K ├── validator 17K ├── util 13K ├── program 8.9K ├── reference 6.0K ├── debug 6.0K ├── destroyable 4.5K ├── vm 4.1K ├── node 3.7K ├── global-context 2.5K ├── wire-format 966B ├── encoder 844B ├── vm-babel-plugins 502B └── owner	cd packages/\@glimmer du --ignore_hidden \ --reverse --apparent-size \ --filter ".+\/dist\/prod\/index.js$" \ --no-percent-bars --only-dir --depth 1 258K └─┬ . 73K ├── syntax 70K ├── runtime 49K ├── compiler 22K ├── opcode-compiler 7.1K ├── manager 6.8K ├── util 5.7K ├── program 4.1K ├── validator 3.7K ├── debug 3.7K ├── reference 2.8K ├── vm 2.1K ├── node 1.7K ├── destroyable 1.6K ├── wire-format 563B ├── global-context 511B ├── encoder 469B ├── vm-babel-plugins 155B └── owner
main	cd packages/\@glimmer du --ignore_hidden \ --reverse --apparent-size \ --filter ".+\/dist\/dev\/index.js$" \ --no-percent-bars --only-dir --depth 1 650K └─┬ . 181K ├── runtime 162K ├── syntax 104K ├── compiler 63K ├── opcode-compiler 29K ├── debug 23K ├── manager 19K ├── validator 17K ├── util 13K ├── program 9.5K ├── reference 6.2K ├── destroyable 5.5K ├── global-context 5.0K ├── vm 4.2K ├── node 2.4K ├── wire-format 967B ├── encoder 870B ├── vm-babel-plugins 497B └── owner	cd packages/\@glimmer du --ignore_hidden \ --reverse --apparent-size \ --filter ".+\/dist\/prod\/index.js$" \ --no-percent-bars --only-dir --depth 1 301K └─┬ . 80K ├── runtime 74K ├── syntax 50K ├── compiler 25K ├── opcode-compiler 21K ├── debug 11K ├── manager 8.7K ├── util 7.6K ├── validator 5.9K ├── program 4.1K ├── reference 3.0K ├── destroyable 2.8K ├── vm 2.1K ├── node 1.6K ├── wire-format 1.4K ├── global-context 604B ├── encoder 469B ├── vm-babel-plugins 155B └── owner

---

I applied this branch to ember-source 5.10, (which can be tested here) and came out with these production build sizes:

# 5.10 + this branch
❯ du --no-percent-bars --apparent-size --filter ".js$" dist/assets/

4.4K   ┌── chunk.2772c50f98341aff4d18.js
 98K   ├── chunk.04801e0d37536e3eb752.js
394K   ├── vendor.89b515acd0ef98846b07babf4e0f28a9.js
496K ┌─┴ assets

# 5.10 

❯ du --no-percent-bars --apparent-size --filter ".js$" dist/assets/ 

4.4K   ┌── chunk.39bc453487af3c4b70e2.js
 97K   ├── chunk.0bb4a1b44f780f8fed74.js
411K   ├── vendor.a80a7a73fcea3dc743e6a8ccc8536d9e.js
513K ┌─┴ assets

It is 17K smaller overall.

That's still far from the 52KB production vendor growth reported by emberjs/ember.js#20628

---

In a future PR, I'd like to switch the build to SWC, because SWC's minifier solves the problem we have:
https://play.swc.rs/?version=1.7.24&code=H4sIAAAAAAAAA12OTQqDMBCF957iIS4iFALJ0ssYdYqCNSEmNKl496at1trd%2B%2FnmMZzjbgdH0BZGzUm0uiP0ZCm7%2Bql1g54QWCixwJLzNrkK69FFFi6IFzz%2BiCo7GNW0TP32yZ1GEiBORNyIjILR1uFLUlA3MxJLbAZwDu2d8Q5zr%2F3YoaFPus3UAgISsq5SvGfFEpgoVxRLZLnI3%2Br1odyVYLlMcTpanwuGSssfAQAA&config=H4sIAAAAAAAAA31VPY%2FbMAzd71cEmjsUGTp0vivQIehQ4FZBsShHV1kyRCqXIPB%2FLy07H5fQ2Sw%2BPlLix%2FPpZbVSH9ion6sTf%2FKhNxkhX85swWMkc2CLomMP2GTfk%2Fp2RglHyJmAUE3DhCgyuQUaWYDr7%2Bv1zFAhJYQzY7Z1Pnp3vM3ZpK7PgHhjYyuHLB1Ewq%2F8GcvpcwQol1v7NqUAJj5BtEHtI0ELWQrcpBBMj6D3JgtRxpua7DFJKUawEFjd59SLeLSefIqc8xG1YKxukgUB8hka8nuQaJyLaRH5ecJ7KmxhW9q29vmODXsTiiEhJxxqS%2Fi2QtRd8kjalSiVcAIXajCBc3Hvmd7pDFRyfOR9JB8XevIPgCsQDGI0HUhxq4fjeVpiu6dMHx2PLB0FnOdbemWElouqvXdCZcfKQCYvdTODLQ2MlW2k68zwQvnQW9DgHM%2BKEBo%2FPTU7Kem46MkJAPfXOGmqJkBftnABHxfiCfyLX0nygM0enaHdMorHbpvCkwQd0C7ZJw7cCkrLcGaVOPTLeIkWeDTAii4FK%2FAoArwAlHSoevkwG7weHFG3IW2vMjE7DBcd7kxs7%2Ff9osWZZc6l3H2VWE67SbaEOgEndU1wUlUf6tfvv%2Fr17V1v%2Fry%2BjWJeo6thGKYEL3MS1dVA19%2FIOEGT%2BP9QV6ezzl%2FuqDxuzsz6rOE%2FQnOQmpIGAAA%3D

Terser fails at this:

And terser failing at this is one of the reasons we've had so much dev code in prod

[github-actions]

duration phase no difference [-175ms to 29ms]
renderEnd phase no difference [-1ms to 2ms]
render1000Items1End phase no difference [-18ms to 4ms]
clearItems1End phase no difference [-4ms to 2ms]
render1000Items2End phase no difference [-18ms to 0ms]
clearItems2End phase no difference [0ms to 2ms]
render5000Items1End phase no difference [-58ms to 8ms]
clearManyItems1End phase no difference [-1ms to 2ms]
render5000Items2End phase no difference [-65ms to 1ms]
clearManyItems2End phase no difference [-4ms to 2ms]
render1000Items3End phase no difference [-23ms to 7ms]
append1000Items1End phase no difference [-24ms to 29ms]
append1000Items2End phase no difference [-15ms to 9ms]
updateEvery10thItem1End phase no difference [-4ms to 1ms]
updateEvery10thItem2End phase no difference [-1ms to 0ms]
selectFirstRow1End phase no difference [0ms to 0ms]
selectSecondRow1End phase no difference [0ms to 0ms]
removeFirstRow1End phase no difference [-1ms to 1ms]
removeSecondRow1End phase no difference [0ms to 1ms]
swapRows1End phase no difference [0ms to 1ms]
swapRows2End phase no difference [0ms to 0ms]
clearItems4End phase no difference [-2ms to 1ms]
paint phase no difference [0ms to 0ms]

[16:59:30] Generating Benchmark Reports [started]
[16:59:39] Generating Benchmark Reports [completed]

Benchmark Reports    

JSON: /home/runner/work/glimmer-vm/glimmer-vm/tracerbench-results/compare.json

PDF: /home/runner/work/glimmer-vm/glimmer-vm/tracerbench-results/artifact-1.pdf

HTML: /home/runner/work/glimmer-vm/glimmer-vm/tracerbench-results/artifact-1.html

[lifeart]

Amazing work!

[NullVoxPopuli]