[1.1.81] Belts under inserters pulse weird item counts

[splee]

Having an inserter grab from the same tile of belt that's pulsing a contents value to the circuit network is outputting the whole belt contents when only one new item flows in.

The "good" setup is:

• have items flowing down a belt and put into a chest by an inserter (capacity 1)
• wire the 2nd to last tile of belt to both input & output of an arithmetic combinator
• belt doesn't enable/disable but does send contents, in pulse mode
• inserter isn't taking items from the tile of belt pulsing the signal, but a later piece of belt
• arithmetic combinator configured as Everything + 0 -> Everything so it tallies the historic total of item flow

The combinator output differs from box contents by a fixed constant, because items have passed through the circuit-ed belt but not yet been picked up by the inserter. Makes sense. When the inserter picks one item off the belt, one new item flows onto the belt, the belt pulses a value of 1 and the combinator "tracker" increments by 1. All makes sense.

I can imagine using this functionality for stuff. It's the "good" setup, the baseline. The leftmost in the included blueprint, which is probably easier.

The "weird" setup(s):

• the belt and combinator still wired & set up the same way
• have the inserter grab from the same belt tile that is pulsing the circuit value

OK so this one is weird in several ways. When one lane is full and the inserter takes an item, the belt outputs a signal pulse of up to 4 even though only one new item entered the circuited belt. So it's pulsing the total number of items on the belt?

But if both lanes are full, the inserter picks 1 item off, the belt still pulses 4 for a full belt even though there are 8 items over both lanes. So in some cases the signal pulse isn't even the count of belt contents, but lane contents.

I can't really see how it's useful to pulsing full belt contents when only one item flows in, but maybe someone else can. The thing that really makes this a bug (to me) is the discrepancy between the output signals when 1 vs 2 lanes are filled with items, and the discrepancy between counts when the inserter is taking from the circuit pulsing belt rather than from subsequent belt after the circuit pulse.

Expectation:

I'd hope that all three circuit networks of the attached blueprint count the green chips at the same rate, and agree with each other & the box contents (to within a constant because belt buffer). However, they don't increment at the same rate.

Identification credit goes to SeiferKatt, clipped from their stream, they also talk through some of it:

https://www.twitch.tv/seiferkatt/clip/T ... -5Rw7CfOyQ

I just reproduced this - both 1.1.80 and 1.1.81 - and built the demonstrating blueprint thing.

[Loewchen]

viewtopic.php?f=48&t=48324

[splee]

Fair enough, the player workaround is still just "use the good setup and not the weird ones" and that works out fine.