Anyone ever run into local scope corruption within a cfc in cf21 when using ArrayMap in parallel mode, followed by an ArraySort? By local scope corruption, I mean that the local scope is now based off the ArraySort closure, outside of the closure, so that function level variables are no longer seen

Can you give us a repro case, demonstrating the issue?

/**

* @output false * @accessors false **/ component{ /** * @output false **/ public Foo function init(){ return this; } /** * @output false **/ public array function doThing(){ return doThing2(); } /** * @output false **/ private array function doThing2(){ cfmail( from="debug@localhost", to="debug@localhost", subject="Local Scope 1", type="html" ){ WriteDump( var=Local, showUDFs=false ); }; Local.aValuesIn = [ "a", "b", "c", "d", "e" ]; Local.aValuesOut = Local.aValuesIn.map( function( item ){ switch( item ){ case "a" : { return new Bar( 1 ); } break; case "b" : { return new Bar( 2 ); } break; case "c" : { return new Bar( 3 ); } break; case "d" : { return new Bar( 4 ); } break; case "e" : { return new Bar( 5 ); } break; } }, //enable parallel to break code true, 5 ); cfmail( from="debug@localhost", to="debug@localhost", subject="Local Scope 2", type="html" ){ WriteDump( var=Local, showUDFs=false ); }; Local.aValuesOut.sort( function( e1, e2 ){ return compare( e1.getId(), e2.getId() ); } ); cfmail( from="debug@localhost", to="debug@localhost", subject="Local Scope 3", type="html" ){ WriteDump( var=Local, showUDFs=false ); }; return Local.aValuesOut; } }

(normally a big wodge of code like that would be better in a trycf gist or just a normal gist rather than inline, but cool

Local Scope 2:

Also that's not very stand-alone. Is being in a CFC a requirement?

yes

If yer repro code poss requires zipping, it's not been pared back enough to just be a repro case, I think. I mean that code has a cfmail call in it?! How is that necessary to reproduce the situation?

My point is your repro case should only include code necessary to demonstrate the issue. Nothing more. Nothing less.

the mail is just to capture the state of local to verify it's busted

This is a repro of your problem: https://trycf.com/gist/a14e90f6775e943be73d643a1f44ef07/acf2021?theme=monokai

what bug fix?

I had a bug in my sort callback

mmm

anyway, that reproduces your problem. And anyone can just click "run" to see it

well, at least i'm not crazy (in that there is a problem)

Seems to be peculiar to

sort

. Similar code using

reduce

is fine

I've run into this issue a few times, recently, finally decided to track it down

yeh I noticed that. I don't really see how that should even possibly contribute to it. But... erm... it doesn't surprise me of how ColdFusion has implemented higher order functions and function expressions etc.

It doesn't even have to be set parallel within the function. any use of parallel map during the request will trigger it

Maaaaaaaaaaaaaaark!!!!! @Mark Takata (Adobe) 😉

Appreciate the help with this, been bugging me for 3 days

Well that one is... wow. Quite a bug. I think you're correct btw, there's some kind of call behind the scenes that is borking up (technical term) the scopes. Lemme send this in with the two other things I found yesterday 😛

Did this ever get put into the bug tracker? I'm eagerly excited for the next patch to resolve this so I can clean up a bunch of code

Jim, this update was security-only. We are trying to separate the patches from now on whenever possible. The bug fix patch should be coming soon, I'll check to see if it got in that.