FAILURE IN SERVER /b_getn index out of range

[prko]

While testing the latetst version which I rendered under OS X, I experienced the error

FAILURE IN SERVER /b_getn index out of range

after evaluating the following code:

{
var src= SinOsc.ar(440), amp= SinOsc.ar(440/20, 0, 2);
[src, amp, src * amp, (src * amp).clip(-1, 1)]
}.plot(20/440)

In 3.9.3, the code works well.
If the error occurs because 3.10 is not yet completed, Please close this issue.

[patrickdupuis]

The full message I'm getting on branch 3.10 is this. The last two lines appear a couple seconds after the failure message.

-> a Plotter
FAILURE IN SERVER /b_getn index out of range
WARNING: Buffer-streamToFloatArray failed!
Try increasing wait time

[miczac]

just tried this w/ 43e9244, can confirm.
Is this related / connected to what this PR #3848 addresses for supernova?

[prko]

I don't know.
However, I rendered SC with the supernova-enabled option, but supernova does not boot and I booted with sclang.

When I evaluate

Server.supernova

sc returns

-> Server

then evaluating

s.reboot

returns

server 'localhost' disconnected shared memory interface
'/quit' message sent to server 'localhost'.
-> localhost
Server 'localhost' exited with exit code 0.
Booting server 'localhost' on address 127.0.0.1:57110.
Error when parsing command line arguments:
the argument ('2097152.0') for option '--rt-memory' is invalid

Server 'localhost' exited with exit code 1.

[nhthn]

i found that this is sufficient:

{ DC.ar(0) ! 2 }.plot(0.0101)

strangely, this bug appears to be sensitive to the plot duration. this consistently works for me:

{ DC.ar(0) ! 2 }.plot(0.0100)

conjecture: the duration must translate to an integer number of samples.

if you check the Function:plot source code, you'll fin immediately translates to getToFloatArray, and if you replace plot with getToFloatArray the bug is still visible.

...wait a minute, i could have sworn it was actually loadToFloatArray at some point. when did that change? git blame tells me that it was changed to getToFloatArray recently, in #3088.

{ DC.ar(0) ! 2 }.getToFloatArray(0.0101) // fails
{ DC.ar(0) ! 2 }.getToFloatArray(0.0100) // OK

@telephon mind looking into this? thanks :)

[telephon]

...wait a minute, i could have sworn it was actually loadToFloatArray at some point. when did that change?

Sure it did, in the commit you refer to. loadToFloatArray also exists, but requires access to the same file system, while getToFloatArray will also allow plotting remote server signals.

I'll take a look.

[telephon]

Interesting.

// works fine
(
{
	var src= SinOsc.ar(440), amp= SinOsc.ar(440/20, 0, 2);
	[src, amp, src * amp, (src * amp).clip(-1, 1)]
}.plot(2 * 1633 + 67 / s.sampleRate)
)

/*
[ 'b_getn', 1, 0, 1633 ]
[ 'b_getn', 1, 1633, 1633 ]
[ 'b_getn', 1, 3266, 1633 ]
[ 'b_getn', 1, 4899, 1633 ]
[ 'b_getn', 1, 6532, 1633 ]
[ 'b_getn', 1, 8165, 1633 ]
[ 'b_getn', 1, 9798, 1633 ]
[ 'b_getn', 1, 11431, 1633 ]
[ 'b_getn', 1, 13064, 267 ]
*/
// fails
(
{
	var src= SinOsc.ar(440), amp= SinOsc.ar(440/20, 0, 2);
	[src, amp, src * amp, (src * amp).clip(-1, 1)]
}.plot(2 * 1633 + 69 / s.sampleRate)
)
/*
[ 'b_getn', 1, 0, 1633 ]
[ 'b_getn', 1, 1633, 1633 ]
[ 'b_getn', 1, 3266, 1633 ]
[ 'b_getn', 1, 4899, 1633 ]
[ 'b_getn', 1, 6532, 1633 ]
[ 'b_getn', 1, 8165, 1633 ]
[ 'b_getn', 1, 9798, 1633 ]
[ 'b_getn', 1, 11431, 1633 ]
[ 'b_getn', 1, 13064, 274 ]
FAILURE IN SERVER /b_getn in
*/

no time to check further. Maybe someone spots an off by one error somewhere here in getToFloatArray?

[telephon]

(
a = 2 * 1633 + 67 / s.sampleRate;
{
	var src= SinOsc.ar(440), amp= SinOsc.ar(440/20, 0, 2);
	[src, amp]
}.getToFloatArray(a, action: { |x| x.postln })
)


(
a = 2 * 1633 + 69 / s.sampleRate;
{
	var src= SinOsc.ar(440), amp= SinOsc.ar(440/20, 0, 2);
	[src, amp]
}.getToFloatArray(a, action: { |x| x.postln })
)

[mossheim]

Closed in #3875.