With the current particle system, electronics(and all reactions) may behave different depending on when the individual particles were created. When a save loads, it creates them from the top left, going right and down, which is different than how a user would have created it manually. It has to do with how the particles are updated, first created to last, and some reactions only transfer instantly (SWCH) if all the updates follow the SWCH's direction, example, a line of SWCH you draw manually going right to left will turn on instantly from the right side, but use the line tool (or save and reload) and it will go slow. This can't really be fixed and it has always been this way, but i think a new saving method (not available yet) will fix the save/loading differences.

So to make something work consistently, make sure you save/load and test a lot.

@cracker64 (View Post)
True, but in this case there is some SPRK on normal conductors which is behaving inconsistently. And that *is* meant to go at a constant speed. In the video, there's a particle of PSCN in the top half that is sparked one frame before the corresponding particle in the bottom half. Up to that point, the progress of the spark is symmetrical.

I'm having a go at fixing it now. In sprk.c: "(parts[i ].life<3 || ((r>>8)<i && parts[i ].life<4))". I think those conditions may be wrong. It's particle r>>8 that will have a different life value depending on whether it is updated before or after the spark is transferred to it, not particle i.

@cip (View Post)
Thank you for giving such a detailed bug report! I wish all of them were like this one.

Yes, spark travel speed is supposed to be invariant.

Here's something to test (but note that saves/stamps from it won't open in the current version):
https://github.com/downloads/jacksonmj/The-Powder-Toy/powder-sprkspeed.exe
(source can be found at https://github.com/jacksonmj/The-Powder-Toy/tree/sparkspeed)

With that version, all three test cases work for me. Most other saves seem to still work, I would be interested to hear about ones that don't.


@cracker64 (View Post)
Actually, the new save format won't fix that. Storing coordinates for each particle didn't compress very well, so it does a similar thing to the current format except multiple particles can be in the same place.

The single frame delay in "INST fail test" save was because the wifi particles weren't firing at the same time.

Press D to enable debug mode and mouse over a particle. The particle ID will be displayed (top right corner, the number after the '#'). On each frame, there is a loop that updates particles (calculates interactions with other particles and moves particles around) in order of their particle ID.

On the frame when the input wifi particle receives a spark:
The top three output wifi particles are updated first. The wifi channel is off, so they do not create a spark. Next, the input wifi particle is updated, and because there is an adjacent spark it turns the wifi channel on. The remaining four output wifi particles are then updated, and because the wifi channel is on they produce a spark.

On the following frame, the top three output wifi particles produce a spark.

The fixed program delays all spark production by wifi until the following frame.


Text display was broken because I stopped SPRK from travelling instantly from left to right along SWCH - I've reverted that change and updated the download. Same link, https://github.com/downloads/jacksonmj/The-Powder-Toy/powder-sprkspeed.exe


It's based on a newer version of the source code than v70.1, so the sign tool no longer creates walls and the false walls are removed when loading saves.