Remove limits of ES6 features in standard library code #20700
Conversation
sbc100
force-pushed
the
es5_transpile
branch
2 times, most recently
from
cbb258d to
ad1c276
Compare
When transpiling to ES5 use closure with `SIMPLE_OPTIMIZATIONS` rather then `WHITESPACE_ONLY`. This means that polyfills cant be included as needed and removes the limits on what ES6 features we can use. The downside of this is that is slows down builds for ES5 users but this seems like a reasonable tradeoff. Fixes: emscripten-core#11984
sbc100
force-pushed
the
es5_transpile
branch
from
ad1c276 to
423a29a
Compare
|
ping.. |
| - JavaScript library code can now use the full range ES6 features and we rely | ||
| on closure compiler to transpile for ES5 when targetting older browsers. | ||
| For those that want to would rather perform transpilation seperately outside | ||
| of emscripten you can use the `-sPOLYFILL=0` setting. (#20700) |
There was a problem hiding this comment.
What is the connection between transpiling and polyfills? That POLYFILLS disables transpiling seems slightly surprising to me and I think might be for other readers.
There was a problem hiding this comment.
Yes, I kind of lumped them together.
I do think its somewhat logical though since folks who use -sPOLYFILL=0 are basically saying "I promise to install my own polyfills".. otherwise the code simply won't run, and in almost all cases that I know of polyfills are normally installed by transpilers, so what they are really saying is "I promise to my own transpilation" to ES5 (or whatever).
I've not actually heard of anyone using -sPOLYFILL=0 in the wild, but if we hear back from suchusers that they want to use this setting and still have emscripten take care of transpiling we can consider a separate setting.
3 participants
When transpiling to ES5 use closure with
SIMPLE_OPTIMIZATIONSrather thenWHITESPACE_ONLY. This means that polyfills cant be included as needed and removes the limits on what ES6 features we can use.The downside of this is that is slows down builds for ES5 users but this seems like a reasonable tradeoff. It also makes debugging harder for such users since closure will minify the names of locals in this configuration even though we pass
--formatting=PRETTY_PRINTwhich seems to prevent minification of global names.Fixes: #11984