H. James Harkins

> I fully agree that the behavior of a window should only be concern of the Window class and IMO this is what I suggested as the proposed change allows...

> So is there a chance that this could get merged if it gets written as a WindowDef? It was proposed by the developers of Jitlib but I know people...

@patrickdupuis Perhaps 1e-04 is demanding too much precision (see `classvar precision = 1e-04;`). On my Linux machine, I ran it several times and didn't see any failures. It's entirely likely...

> I'll try different precissions and see if the failures stop. They should: 0.754712 - 0.754595 = 1.17e-4 so it only just barely missed passing. Probably precision should be 0.01....

``` ERROR: Message 'prNumPeersChanged' not understood. RECEIVER: Instance of Frame { (0x55b04b4bd4d8, gc=54, fmt=01, flg=00, set=03) indexed slots [7] 0 : instance of Method CollStream:pos_ (0x55b045979200) 1 : Integer 0...

I can't find the reference right now, but there is a known bug about `try` being called within other control structures. My guess is that JMc implemented `try` by exploiting...

Here is a minimal-minimal reproducer (of the previously-reported bug, which I can't find now): ``` [1, 2, 3].collect { |x| try { 1 + x.flurghle } { x } }...

So it may be that the throw unwinds one and only one item from the stack. If additional items had been pushed, they remain, and mess up subsequent operations.

Another reproducer (using cruciallib, but not a bug in cruciallib): ``` Instr("testenv", { |freq = 440, nonstandardenv, amp = 0.1| SinOsc.ar(freq, 0, amp) * EnvGen.kr(nonstandardenv, doneAction: 2) }, [\freq, nil,...

``` x = [nil]; "hello".postf(try { max(x, 1) }); ``` Based on Julian's simpler case, I'll update the issue title: there is no loop *outside* `try` here, but there is...