[boskid][2.0.28] Same decider condition's status changing based on insertion order

[arl85]

Hi,
reading wiki seems that any condition involving EACH will never output 1 for a signal which has a zero value, even if the condition is one that would be passed for a zero value.

For example:

Code: Select all

     + EACH >0
OR   |
     + EVERYTHING = 0
     

Output: DOT signal

no input -> no output

immagine.png (78.75 KiB) Viewed 428 times

That's ok, but... if I first insert the EVERYTHING = 0 condition, that condition remains ON.

please note the insert verb: if I simply move the EVERYTHING condition on top, that condition remains OFF, but if I remove the EACH one, EVERYTHING one turns on and remains on even if no output is returned when the EACH condition is added.

immagine.png (79.34 KiB) Viewed 427 times

I'll try to better explain with a video.

[robot256]

I can't reproduce this on Vanilla 2.0.28. "Everything=0" evaluates true, but as soon as I click the "add condition" button, the output signal disappears, even before selecting the "Each" signal. In your video, the output signal remains in the GUI even though there is a condition entry with an unconfigured signal selection. Can you post a save with the combinator in that state?

Also, is the game paused or in editor mode while you are doing this?

[Muche]

It's not about an output signal. It's about the visual green status of Everything=0 sub-condition.

If Each is added first, the moment it's added, the combinator is intenally switched to vector mode and subsequently added Everything=0 takes that into account and evaluates to false (due to no signals - no evaluation).

If it's added first, it evaluates to true (as everything=0). When Each is added, the combinator is switched to vector mode, but Everything=0 subcondition is not reevaluated, so it stays green, seemingly keeping thinking it's still in normal scalar mode.

This looks to be only visual inconsistency, after saving & loading the visuals of subcondition in both cases is consistently false.
It also seems to happen only when setting up the combinator, after copy+paste/blueprinting it's built correct.

[boskid]

Thanks for the report. It looks like this will be fixed in 2.0.42 as part of 120624. I see exactly how it failed: when decider was evaluating conditions while in "each" mode, i placed an early exit when there were no signals on the input side and this caused conditions to not be evaluated and one of the side effects of evaluating those conditions was checking if conditions were fulfilled for the gui purposes. "Everything" condition when added as second simply never had its condition evaluated to have gui follow up the changes while when it was added first, it was evaluated and when each signal was later added, this evaluation now stopped updating.