Settings->Audio prompts-> Power alert "Average power 65536W" during ride when power meter not connected

This is being fixed in the 23.15 beta

• Fix power meter audio prompts playing when sensor is disconnected

I submitted this to product support on 07/03/2025 so this is a quick turnaround - thanks Garmin!

PubBike said:

Obviously the erroneous high number needs to be fixed, but going beyond that it would be good if the power alert was only generated when a power sensor is connected.

Not that it really matters to the end user, but the explanation for both of these issues is that 65536 is used internally in certain contexts to mean "no power value".

This means that code which handles power has to take that special case into account. And in this case, there must be new/changed alert code which did not do that. So it's likely a single bug which caused what looks to the end user as 2 separate issues.

Ofc the real questions are:

- how could this suddenly start happening when Garmin has been dealing with this special case in its own code for probably 20 years now?

- how did this get past testing (either automated or manual)?

- why do the same kind of bugs pop up over and over again (we've also recently seen issues like incorrect distance / time alerts with big numbers like 4 billion)?

PubBike said:

And combine that with some issues with their software architecture, and general development error handling philosophy.  It would seem obvious to me that you only do something with a piece of data - e.g. produce an alert - if the relevant data is valid and/or the sensor producing it is present, else it could cause the watch to crash by trying to process an invalid value.  That doesn't seem obvious to Garmin.

Yeah, I think I tried to address that above by explaining that 65536 is probably used internally as an invalid value for power. Obviously somebody didn't realize that and/or forgot, when making a change to the alert code in question. (well the actual invalid value for power in certain cases would be 65535, but the general point is that 65536 probably isn't meant to be valid data.)

It's not that Garmin literally forgot to handle errors, it's that there's a certain scenario which is indistinguishable from valid data if you don't know what you're looking for. There's a subtle difference between not handling a certain error and not even knowing that something is an error in the first place.

That's the point I was trying to make. It's very similar to the issue with the spurious time/distance alerts that have gigantic values (e.g. 4 million/billion-ish) that was reported recently.

The explanation is not meant to absolve Garmin - quite the opposite, as it shows that Garmin is not able to reliably work with their own system that they should have full knowledge and control of.

The only point is that what seems to you to be two separate issues -  "erroneous high number" and "power alert [should only] be generated when power sensor is connected" - are actually probably 2 symptoms of the same issue: 65536 is internally used to signify that power data is invalid and the alert code should realize that fact.

By the same logic, seeing a spurious distance alert of 4 million or billion km is probably not 2 separate issues, but just 2 symptoms of the same issue.

PubBike said:

So I don't begrudge anyone for not reporting it.

Neither do I, my point was that this kind of thing happens with Garmin all the time, and I don't think it should be the job of end users to find this stuff as if we are unpaid beta testers (oh wait).

I do agree that Garmin has systemic problems with software. Users have been complaining about this forever.

I would not be surprised if the fix for this issue in 23.15 is just to add an if-statement e.g.

if lap average power<65536 then queue average power audio prompt

...which of course we know would be just more of the same poor programming practice.