• lowprofile@lemmy.worldEnglish
    4·
    1 year ago

    Hi, multi-channel Emisar D2 user here, Model number 0135, emisar-2ch, using ToyKeeper’s latest Anduril 2 version, anduril.2023-08-07.emisar-2ch.hex: At my end, momentary turbo 3H/4H works just fine and as expected, so I do not have the same behavior.

    4H from ON for momentary turbo works with every channel mode, whether there is tint ramping or not, and 3H from ON for momentary turbo works when there is no tint ramping with the selected channel mode; all exactly as displayed in the diagram.

  • Reader9@programming.devEnglish
    1·
    1 year ago

    Good question, just wanted to point out your link

    [https://bazaar.launchpad.net/~toykeeper/flashlight-firmware/multi-channel/view/head:/ToyKeeper/spaghetti-monster/anduril/ramp-mode.c#L423]()

    markdown is broken but copy-pasting the url works.

      • Reader9@programming.devEnglish
        1·
        1 year ago

        I don’t own a multi-channel, but I did see something that looks interesting in the code (not familiar with the codebase and haven’t worked with C much):

        Could be a red herring but the function channel_has_args looks kind of like it might be doing the same work as if(!arg) for single-channel. Again this could be totally wrong but IF that were the case, then I’d expect that adding to the if statement could potentially help catch this? Like

        if (! arg) || (! channel_has_args(cfg.channel_mode) .

        Also in following the code style the actual statement would also need to include an IFDEF for multi-channel-config but maybe not necessary for you test locally?

        • Protiron@lemmy.worldOPEnglish
          1·
          1 year ago

          Also in following the code style the actual statement would also need to include an IFDEF for multi-channel-config but maybe not necessary for you test locally?

          So Mementary Turbo 3H work for you? If you don’t own multi channel, MT should be 3H from ON. I think l423 be identical for simple or multi channel.

          I can’t find where arg should be equals to 0… steady_state always call with arg <> 0

          • Reader9@programming.devEnglish
            1·
            1 year ago

            Sorry for the extraneous information about the ifdef. The important part was whether adding this extra check would fix the bug in the same way as deleting the original line did.

            I can’t find where arg should be equals to 0… steady_state always call with arg <> 0

            From what I could assume in the comments, this refers to the very first state in that steady-state (“first frame”) right when you have done 4H. Then it sets the current level to the appropriate turbo (set_level_and_therm_target).

            Your bug would be that it never set the level to turbo because it never executed set_level_and_therm_target because ! arg evaluates to false.

            Your bug fix was to delete the check, but that means that every instant it will keep trying to set the level to turbo instead of just setting it during the first frame and then maintaining the steady-state.

            My potential fix was to add something so that the statement if (! arg) || (! channel_has_args(cfg.channel_mode) would evaluate to true during the first frame so it can then call set_level_and_therm_target exactly once.

            @[email protected] (if you have a moment) does this seem close to the mark?

            • Protiron@lemmy.worldOPEnglish
              2·
              1 year ago

              | (! channel_has_args(cfg.channel_mode)

              I test that asap. I already try to comment it and it’s work fine :

              //if (! arg) {  // first frame only, to allow thermal regulation to work
    uint8_t tl = style_2c ? 150 : turbo_level;
    set_level_and_therm_target(tl);
//}

              But I am surprised that the functionality appears in the diagram and that no one remarks that it does not work (3H without Multi channel or 4H with) https://lemmy.world/post/1038159

              • containerfan@lemmy.worldEnglish
                1·
                1 year ago

                It appears in the diagram because it works. Just tested it again on my dual-channel D4V2, and it works like a charm.

            • Reader9@programming.devEnglish
              1·
              1 year ago

              On reflection i think this is incorrect. arg seems to represent the output level as of when the event was triggered.

  • Reader9@programming.devEnglish
    1·
    1 year ago

    Speaking of testing locally - how do folks test code changes? Actually flash them to driver, or is there some kind of way of simulating the inputs and outputs? I’m guessing there isn’t an Anduril emulator.

    • SammysHP@feddit.deEnglish
      3·
      1 year ago

      Yes, you could (theoretically) use an emulator for the microcontroller. I tried it a few times. And it isn’t fun. It takes a lot of effort to simulate all inputs, outputs are hard to interpret and all kind of effects of the real light aren’t reproduced.

      In reality we flash the build onto a flashlight and try it. Sometimes it’s only a prototype, disassembled. And rarely it’s an actual devboard which has all relevant parts of the flashlight nicely accessible (basically the microcontroller with required electronics, regulated power supply, low power LED to see the output, several LEDs for aux and button and the switch itself).

      • Reader9@programming.devEnglish
        1·
        1 year ago

        That makes sense that it’s easier to flash directly since there are so many different parts and variations to the actual destination hardware.

        I read that one of Toykeeper’s original design decisions was to separate out the generic UI code from the specific code that interfaces with the hardware (being vague due to lack of knowledge). Maybe it would be possible to incorporate some testing code that only evaluates the user inputs compared to expected outputs within the Anduril layer? Such as a single state transition like “when multi-channel is enabled and user selects 4H, selected output level is equal to turbo”. Such a test could be indicate unexpected behavior when UI code is being customized.

        Just adding to the imaginary backlog here!