How can you achieve the same thing with node is script? That is compiling all dependencies into a single is file when you want to release the script?

I recommend post-processing the file with https://github.com/vercel/ncc

shadow-cljs can also do that on its own but its less reliable and requires more configuration

@thheller you responded when this was brought up recently in another channel: andare (self host compatible core.async: https://github.com/mfikes/andare), doesn't work with shadow due to this function https://github.com/thheller/shadow/blob/master/src/main/shadow/util.clj#L29, specifically the :require on [cljs.core.async.impl.ioc-macros :as ioc] which isn't in andare. Was wondering now if you'd consider removing it or gutting it of the internal async stuff. I've removed it locally and have had no problems running my regular shadow workflows, for what that's worth. Andare's implementation is probably not ideal but seems like the best way to accomplish getting core.async on self-hosted cljs for now. You might have a better sense if anyone is using that go! macro though.

@thheller just FYI: this is one minimal snippet I was able to produce on which GCC (`v20200504`) would fail:

function aa() {
    var a = {}, b = null;
    a.a = 1,
    b.a = 2,
    Object.defineProperties(a, b);
}

@andrewboltachev odd code. did you report it?

@thheller not yet actually. But this is my code. Actual one looks more like this: http://dpaste.com/1053C50#line-87

which comes from that "turf jsts" module (it's min version, which I've beautified)

the selected line confuses the GCC

if replaced with sth = Object.defineProperties(Bi, Vi); it would pass

what it's doing is about that: this Bi object just got an update with Object.defineProperties by adding some Vi properties to it. But the object isn't used anymore so it wants to optimize that extra work (calling Object.defineProperties) away

hmm yeah no clue. the code looks valid to me so must be a bug in closure

defineProperties has no return value I think. it just modifies the first argument in place

so that should work fine

yes sure

but having that Bi (or a in my case) won't be used anymore

won't it want to optimize it away?

just include your shortened example and the original code in the report

you can reproduce the error with the web interface which helps

https://closure-compiler.appspot.com/

aha