Battery graph removes/ignores valid data points

[Pat McGowan]

There seems to be some heuristic in use to try to eliminate faulty data. I see correct graphs where the phone has run out of battery on a steep curve. then after a recharge and some further use, the graph is smoothed out and the low battery level data points are no longer used. This is hiding some valid events in the history.

[Sebastien Bacher (seb128)]

Thanks, could you describe a bit more the values of the data points that are ignored?

One thing that we do is

" /* TODO: find better way to filter out suspend/resume buggy values,
         * we get empty charge report when that happens, in practice batteries don't run flat often,
         * if charge was over 3% before it's likely a bug so we ignore the value */
        if (up_history_item_get_state(item) == UP_DEVICE_STATE_EMPTY && currentValue > 3)
            continue;"

The reason we do that is that on suspend/resume upower sometimes give us a point stating that the battery is empty, even if it's charged to 90%, if we don't filter those out we have "pits" in the graph around suspend/resume times sometimes

[Pat McGowan (pat-mcgowan)]

I can reproduce it and grab the files from /var/lib/upower, but essentially let the phone completely run out or battery and shut down then recharge, the entire curve is flattened out and you no longer see that the battery ran to less than 5%.

[Sebastien Bacher (seb128)]

can you share a screenshot and the data file next time you have the issue?

[Sebastien Bacher (seb128)]

closing since that's old and without update