4.12 Crashing on OS X

[gmint]

I was excited to try the new QLC+ but ran into a snag where as soon as I try to trigger a collection (and I suspect send DMX) I'm getting an immediate crash every time. I *suspect* that it's related to this fixture definition which I partially built in 4.11 and then edited in the 4.12 editor. I would have uploaded the def earlier, but it's a large definition and I wasn't quite done with it yet. I've attached the fixture def, but if you need the actual show file, I can send that too.

Blizzard-N-Trance-NEW.qxf

(41.71 KiB) Downloaded 122 times

[mcallegari]

Given that 4.12.0 is supposed to be backward compatible with "old" definitions, to understand why the crash happened I need to know:
- the macOS version
- which output adapter you're using
- on an empty project, add one fixture (the blizzard one) and move some channel on Simple Desk. Does it crash ?
- does it happen only with the blizzard fixture or also with others ?

thanks

[mlohrey]

I tried adding the fixture you uploaded to a workspace and ran a few scenes without any problems.

Maybe the workspace is the issue? Can you upload it?

Cheers

Mark

Mac OSX 10.14.1
QLC+ 4.12.0

[gmint]

Given that 4.12.0 is supposed to be backward compatible with "old" definitions, to understand why the crash happened I need to know:
- the macOS version
- which output adapter you're using
- on an empty project, add one fixture (the blizzard one) and move some channel on Simple Desk. Does it crash ?
- does it happen only with the blizzard fixture or also with others ?

thanks

I'm on OS X 10.14.1 and using the fixture definition in a blank workspace doesn't seem to crash it, so that's not the issue. I've attached the workspace below. I'm using Art-Net and E1.31 so I have no directly attached interfaces (best thing I've ever done, BTW).

Lime.qxw

(353.83 KiB) Downloaded 169 times

[mcallegari]

I have reproduced the crash on Linux (clicked on Fruit stripes) and fixed it.
However, I'd like to understand why it was happening.
It seems there is a mismatch between a RGB Matrix fixture group size and the size of the RGB Script used.
Your project is quite complex so I didn't go much in details, but since you know how you designed it, can you please check if there's is something about the RGB scripts you use ?

[EDIT] I think I found something. You have a fixture group named "Bar" with size 3x3. However, if you extend it (eg. 4x5) you will see there are more heads that are not used! Is that wanted or a mistake ?
I think if you remove those unused heads, QLC+ shouldn't crash.

[gmint]

Massimo,

Yes, I'm always removing and adding new lights to this particular layout. I try to keep it relatively clean, but I've still got various things in there that I probably have replaced years ago. In any case, I'm sure what happened with that RGB script is that I changed the size of that Matrix at one point. I'll go in and check that particular RGB script so I can work on 4.12. Thank you for looking into this so quickly!

Update: Those fixtures in the Bar group had gotten added inadvertently. I removed them and it seems to be working correctly now.

[mcallegari]

For everyone reading this. There's a technical explanation for this crash, caused by a change on 4.12.0 aiming to optimize RGB matrices.
Let's consider a 50x50 pixel matrix, but only 10 heads (pixels) are assigned, in a sparse order.
On every step change, the 4.11.2 code was cycling through all the 50x50 cells (so 2500 times), mostly finding no heads assigned.
4.12.0 cycles on the heads assigned, so only 10 times and go straight to the point.
The issue here is that I didn't check the out of bounds heads, like in George's case, so it crashed trying to access a variable out of its allocation.

[gmint]

For everyone reading this. There's a technical explanation for this crash, caused by a change on 4.12.0 aiming to optimize RGB matrices.
Let's consider a 50x50 pixel matrix, but only 10 heads (pixels) are assigned, in a sparse order.
On every step change, the 4.11.2 code was cycling through all the 50x50 cells (so 2500 times), mostly finding no heads assigned.
4.12.0 cycles on the heads assigned, so only 10 times and go straight to the point.
The issue here is that I didn't check the out of bounds heads, like in George's case, so it crashed trying to access a variable out of its allocation.

Massimo,

Thank you for the explanation. It does seem like the fade behavior of the matrices may have been affected somewhat. I was trying to get a video quick of the difference between what I was seeing in 4.11 and 4.12, but I need to pull a backup of the file because I must have overwritten it, and I can't do that at the moment. But to describe in terms which are hopefully descriptive enough, it appears that when I run the "one by one" matrix, the tail (so to speak) never fades out until the entire matrix is done running unless I set the fade out speed to zero. Before, I could adjust the length of the tail by adjusting the fade out speed.

[mcallegari]

George, please open a new thread for fade out issue.
I just tested on a simple project and it seems to work as expected. Or at least, it works like 4.11.2.
One thing I noticed though is that the preview doesn't show fade transitions, but the DMX monitor shows channel values are fading.
So, please, share a simple project where I can eventually see the issue you're seeing. Thanks