The Delay1_Ctor method calls Delay1_next(unit, 1);, which should normally then do the right thing. But:

float x1 = unit->m_x1;
<...>
LOOP(inNumSamples & 3,
        x0 = ZXP(in);
        ZXP(out) = x1;
        x1 = x0;
    );

This will output unit->m_x1, which has been set by Delay1_Ctor to the input. So yes, this is wrong.

Somehow I have a hard time testing this though.

Trying your example (and a variant) I do get the expected result.

(
a = {
    var trig = Impulse.ar(0);
    Poll.ar(Delay1.ar(trig), trig, "hello");
    FreeSelf.kr(T2K.kr(trig) <= 0);
    Silent.ar(1)
}.play;
)

(
a = {
    var trig = Impulse.kr(0);
    Poll.kr(Delay1.kr(trig), trig, "hello");
    FreeSelf.kr(trig <= 0);
    Silent.ar(1)
}.play;
)

Returns: hello: 0

Probably close this one. I must have been seeing this only in my faulty (now-retired) local branch. Using a new local branch, forked off from very recent master, I don't see the problem anymore.

b = Buffer.alloc(s, 64, 1);

a = {
    var trig = Impulse.ar(0);
    RecordBuf.ar(Delay1.ar(trig), b, loop: 0, doneAction: 2);
    Silent.ar(1)
}.play;

b.getn(0, 5, _.postln);
--> [ 0, 1, 0, 0, 0 ]

Same for kr.

I need to reopen this issue. I just got bitten by it in a class yesterday.

a = {
	var sig = Impulse.kr(0);
	var delay1 = Delay1.kr(sig);
	[sig, delay1].poll(sig);
	FreeSelf.kr(sig <= 0);
	Silent.ar(1)
}.play;

UGen Array [1]: 1
UGen Array [0]: 1

The context in class was to demonstrate the lowpass-ish behavior of a two point averaging filter: (sig * 0.5) + (Delay1.ar(sig) * 0.5). The impulse response of this filter should be 0.5, 0.5, 0, 0, 0, 0 ... but instead, I get:

b = Buffer.alloc(s, 2048, 1);

(
{
	var sig = Impulse.ar(0),
	delay1 = Delay1.ar(sig);
	RecordBuf.ar((sig + delay1) * 0.5, b, loop: 0, doneAction: 2)
}.play;
)

b.getToFloatArray(wait: -1, action: { |data| d = data.postln });

FloatArray[ 1.0, 0.5, 0.0, 0.0 ... ]  // no

EDIT: What I don't understand is why we are getting different behavior in different systems, at different times. According to https://github.com/supercollider/supercollider/blame/28aea6c19309ee496dc73fb32a85fd8e1dc25016/server/plugins/FilterUGens.cpp#L1951, the line unit->m_x1 = ZIN0(0); has never changed since the initial revision 😵

This will output unit->m_x1, which has been set by Delay1_Ctor to the input. So yes, this is wrong.

Yes, this is wrong, and unit->m_x1 also needs to be reset to 0.0 at the tail of the ctor.

I'll bump this up to the top of the list for the next batch of init sample fixes.

EDIT: What I don't understand is why we are getting different behavior in different systems, at different times.

Impulse was fixed in the intervening years, so perhaps --> [ 0, 1, 0, 0, 0 ] was a case of two wrongs made a right, Impulse was fixed, and we're left with a right (Impulse) and a wrong (Delay1), which now gives [ 1.0, 1.0, 0.0, 0.0, 0.0 ]? 😵‍💫

PR opened: #6110

I also tested all examples in this thread to confirm they're fixed.