Bug with partial maps and outputs

Bug Reports
1

Hi!

Apologies for the many bugs in one post, let me try to mention all of them since there are quite a lot.

Bug 1:
The input of an output device and the output of the output device are not the same.

I work with quite a lot of output devices and regularly find this error. Since I work in a multitude of TMD’s with input devices, an error like this completely destroys the setup.

The error is displayed here;
The rightmost expander is showcased, you can see the output.

image
image671×404 70.1 KB

here the connected output device shows something really weird. Something has been done to it, It has taken my rock mask and overwrote it for like 40%.

image
image950×481 114 KB

Resulting in the following output:

image
image658×655 96.3 KB

This happens randomly consistent and seems to be with combiners and outputs from what I have come across. I can not replicate it unfortunately, but it does happen regularly.

3 Likes
2

Hi there,

Thanks for the bug reports! The first one is quite concerning. A few questions:

  1. What build version is this happening in?
  2. In Program Settings->Storage, set Allocation Caching to None and see if this changes anything. I would not expect it to fix the issue, but if your partial results change, it implies certain things about what’s happening that are useful to know.
  3. Are there any macros in the chain of devices where this occurs?
2 Likes
3

Hi Stephen!

Apologies for my sloppy bug report, let me try to clarify some stuff!

  1. This is the newest, 4047 version. I just recently upgraded.
  2. Let me do that and work today, then by the evening I will update this message with any results.
  3. No nothing, mainly file inputs, just one create water device, some composite devices (unpack and pack water core 3026), and filters.
1 Like
4

Admin note:

I have moved this bug seperate from the others for less confusing followup.

1 Like
5

Hi Stephen,

I have done all the steps you mentioned, but unfornately it still happens.

Some further things to add to the puzzle;
The wrong outputs that are being displayed in the output device are NOT even connected, ANYWHERE to that output device.

It also happens if a build is fully exported, then somewhere very niche a device gets changed, which does NOT influence the output anymore, then the output can still be overwritten by some junk mask. Even when the memory usuage is put on ‘Keep all results’.

I hope this helps a tiny bit.

Zhyrr

1 Like
6

I am still extremely interested in this bug, but I cannot replicate it at all.

What I suspect is happening is that some device is running into an error during processing, but is incorrectly NOT flagging an error. This propogates its wrong, half-built data through the network. The fact that completely “wrong” output is being displayed is likely due to the allocation cache. The allocation cache lets a device use previously-allocated but unused memory which is faster than allocating new memory. This means that if a device incorrectly does not write to its entire area, you’ll see the prior contents.

The root cause is almost certainly a single device class but without more info or having it happen to me, I can’t track it down much further.

Anyways, just an update.

2 Likes
7

I have encountered a similar situation before, which occurred in a very previous version where the prerequisite was severe memory shortage while using disk memory. I have not attempted to reproduce this issue, but I have indeed received similar outputs: outputs that were completely different from expectations, some layers were ignored, and the outputs looks strange.

1 Like