mescalinum

I'm writing some externals with js, and I hit a bug.

This is the simplest test-case for reproducing it, so please ignore the fact that the two externals do nothing useful :).

inlets = 2;
outlets = 1;

function anything() {
    a = arrayfromargs(messagename, arguments);
    post('inlet ' + inlet + ': ' + a + '\n');
    if(inlet == 1) {
        outlet(0, 'foo');
    }
}

or in words, if it receives something on the second inlet, it will emit a 'foo' message.

Then we attach a bang to its second inlet, and a message "$1" on its first inlet, and we connect the external outlet to the message inlet.

What should we expect from this piece of code?

2) js object receives something on second inlet, then it outputs 'foo'

3) message "$1" receives 'foo', and outputs 'foo'

4) js receives 'foo' on first inlet, so it should do nothing

What happens instead, in step 4, is that js thinks that it is receiving 'foo' on the second inlet, so it goes in an infinite loop.

This is also confirmed by what we observe in the Max console:

inlet 1: bang
inlet 1: foo
inlet 1: foo
inlet 1: foo
...

I have the feeling that the global variables in js (like inlet, messagename, ...) makes it non-reentrant, so they fail in this scenario.

Should I go on and report this bug to Cycling 74?

inlets = 2;
outlets = 1;

function anything() {
	a = arrayfromargs(messagename, arguments);
	post('inlet ' + inlet + ': ' + a + '\n');
	if(inlet == 1) {
		outlet(0, 'foo');
	}
}

What should we expect from this piece of code?
Here is my guess:
1) click on bang
2) js object receives something on second inlet, then it outputs 'foo'
3) message "$1" receives 'foo', and outputs 'foo'
4) js receives 'foo' on first inlet, so it should do nothing

bug-js-object-not-reentrant

nevermind the screenshot... "js a" should really be a $1 message

I simplified the testcase while writing the post

nevermind the screenshot... "js a" should really be a $1 message
I simplified the testcase while writing the post


yes the behavior is not as one would expect. It seems to be related to the feedback-loop. I assume it happens because the internal "batch-stack" is never finished and therefore "inlet" does not get updated. (JS always runs in the low-priority thread and is single threaded).

A [deferlow] object between the [js] outlet and the [$1] message will lead to the expected result.

yes the behavior is not as one would expect. It seems to be related to the feedback-loop. I assume it happens because the internal "batch-stack" is never finished and therefore "inlet" does not get updated. (JS always runs in the low-priority thread and is single threaded).
A [deferlow] object between the [js] outlet and the [$1] message will lead to the expected result.


[bug?] js object not reentrant?