Source maps?

[samreid]

See phetsims/phet-core#28

We are wondering if chipper 2.0 should provide source maps.

[jonathanolson]

It would be nice to have, but may have some extra complexity (multiple source maps would be needed, they would need to be placed in a specific path so they could be read, etc).

[samreid]

Can you describe why multiple source maps would be needed? Would there be a single unmangled JS file before the mangle step that could be used as the sole sourcemap?

[jonathanolson]

Can you describe why multiple source maps would be needed? Would there be a single unmangled JS file before the mangle step that could be used as the sole sourcemap?

Correct, unless we can figure out how to combine source maps (might be possible), we'd have one for each group of code we run uglify on (right now at least twice, once for preloads, once for require.js stuff).

[samreid]

Understood, thanks. We could potentially combine preload + requirejs before uglify or skip sourcemaps on preloads if we think it's less risky.

[jonathanolson]

We could potentially combine preload + requirejs before uglify or skip sourcemaps on preloads if we think it's less risky.

Yes. Right now we have each preload split up (I believe) so we can identify which each one is, but I think having one source-map per HTML file would be significantly improved.

[zepumph]

@ariel-phet says we seem to not get enough bugs, and hard to replicate ones at that, for this to be worth the time.

Deferring until someone asks for one, or if we find a strong need for one. Removing chipper:2.0 label to go for it later.

[samreid]

From #701

Sam Reid [2:51 PM]
Babel transpile has implications for profiling, I took some notes here: #691 (comment)

Jonathan Olson [2:52 PM]
hmm, interesting. I wonder if there are flags we can provide that would give us better debugging information?

Sam Reid [2:52 PM]
Maybe something to name those anonymous functions?
key: “decayVertical”,
value: function(i0, sign, width) {

would become

key: “decayVertical”,
value: function decayVertical(i0, sign, width) {

Jonathan Olson [2:53 PM]
I know it was decided a while ago that we never wanted to use source maps, but it would help in this case

Sam Reid [2:53 PM]
Is it easy to generate source maps in the babel step?
Or will that obfuscate the profiling anyways?

Jonathan Olson [2:55 PM]
may slightly obfuscate profiling results, but Chrome might be able to show both with/without source maps. I’m not sure, I haven’t used them much
and I believe all of the tooling we use can generate source maps