[00:43:38] <Krinkle>	 I'll try to make my way through a few today
[01:24:15] <Krinkle>	 TimStarling: I'm curious how the analysis of T343407 would care given fnEval, and e.g. comparing status quo `implement()`  to `impl( function () { return […] ] )`, and possibly a third scenario of `impl( function () { mw.loader.implement(…); ` which would avoid the Function.apply intermediary and possibly allow the call to be analyed better upfront. I can do some comparisons as well, I'll be back in 1h or so.
[01:24:16] <stashbot>	 T343407: ResourceLoader source map on localStorage cache hit - https://phabricator.wikimedia.org/T343407
[01:37:16] <Krinkle>	 ah, that'd complicate getting a hold of the function.
[01:47:04] <Krinkle>	 TimStarling: want me to land/release the Minify lib change?
[02:56:52] <TimStarling>	 you can +2 it, but I'm thinking about adding an index map helper, so maybe leave the release for a day or two
[22:56:35] <TimStarling>	 I've got source maps working now to a demoable quality including localStorage cache hits
[22:57:07] <TimStarling>	 there was some facepalm involved
[22:58:46] <TimStarling>	 turns out (new Function()) is implemented internally like eval("(function(){" + code + "})")
[22:59:08] <TimStarling>	 it adds a browser-dependent header to the eval'd buffer which offsets the source for mapping purposes
[23:00:31] <TimStarling>	 simplest workaround is to use indirect eval instead of the function constructor