2019.28.1

[chronopc]

Seems to be primarily from 28.2 to 28.3.1.

Sitting here with 24.4 still and no update.

Same here. I just want to be able to link my profile to my phone.

 

[inthecity]

Both the app and the car should show the same thing, and neither will truncate the version number. What confused me in the beginning were these version numbers, which I would actually consider build numbers and this other version that people throw around like version 6, 8, 9, etc.

We're currently on Version 9.0 of the firmware, and we get build updates represented by this 2019.28.3.1 numbers. The 28 represents the week of the year that the build was made.

Thank you for taking the time to explain that.

 

[ibGeek]

just got 28.3.1 last night. nothing new in the release notes.

That is to be expected. Lets explain:

So the build number of a Tesla release looks like this: 2019.28.3.1
The 2019 is the year of the build.
The 28 is the week that the build was created.
The 3 is the major revision of the week 28 release. (moderate changes made after the week of release)
The 1 is the minor revision of the week 28 release. (minor changes made after the week of release)

OK with that in mind there will always be changes to the release notes if the year and week change. There will rarely be release note changes for the major revision. There would be just short of never any changes to release notes for a minor revision.

In the case of this weeks build a moderate change occurred that added the fix for Dog Mode, and the release notes were changed.
I believe that this as added to 2019.28.2
Then there were issues with UK owners and Model S owners that resulted in 2019.28.3 and .3.1
Because these were minor fixes they were covered by the last line of every release notes page.

Hope this helps.

 

[ibGeek]

Both the app and the car should show the same thing, and neither will truncate the version number. What confused me in the beginning were these version numbers, which I would actually consider build numbers and this other version that people throw around like version 6, 8, 9, etc.

We're currently on Version 9.0 of the firmware, and we get build updates represented by this 2019.28.3.1 numbers. The 28 represents the week of the year that the build was made.

Sorry, didn't notice that you had already explained this.

 

[chronopc]

Finally got the update. Went from 24.4 to 28.3.1.

 

[jkdman123]

One user for a particular model??? All that tells you is that person had no issues (yet) for what they actually tried during the test period. It doesn't tell you *anything* about the tens of thousands of other users of that same model who have different settings, different hardware attached to their car (e.g. USB sticks), etc.

You’re reading too much into what I said. I meant that once an update version is *validated* then it’s *validated*. It should be released to everyone it applies to simultaneously. This is not hard to understand as it happens that way with pretty much all other software.

 

[MarcG]

You’re reading too much into what I said. I meant that once an update version is *validated* then it’s *validated*. It should be released to everyone it applies to simultaneously. This is not hard to understand as it happens that way with pretty much all other software.

And my point (and others who have chimed in on this) is that it is very difficult to validate every possible driving scenario before releasing.
Think of all the variables involved in Autopilot for example. Hundreds of thousands of combinations and permutations.

 

[willow_hiller]

And my point (and others who have chimed in on this) is that it is very difficult to validate every possible driving scenario before releasing.
Think of all the variables involved in Autopilot for example. Hundreds of thousands of combinations and permutations.

I think this is where the Tesla data advantage and their new training computer "dojo" come into play.

Sure Tesla cannot physically test all AP scenarios prior to a new launch, but they have a vast trove of real-life training data. Their archive contains radar, sonar, and camera data, and can be played through the NN. The NN wouldn't know the difference between a full set of sensor data and real testing. And if Dojo is fast enough, Tesla can run through millions of real life scenarios simultaneously, recording where the net behaved better and where it behaved worse.

 

[SomeJoe7777]

You’re reading too much into what I said. I meant that once an update version is *validated* then it’s *validated*. It should be released to everyone it applies to simultaneously. This is not hard to understand as it happens that way with pretty much all other software.

The car's software is not monolithic. It is not just one big update. The process works similar to a Linux system, where the car queries the update server to find out what updates for each module are available that are applicable to the particular car. Once all the updates that are available for all modules are determined, then the server packages all of them together into an update package that is specific for that car.

Each car has dozens of modules -- drive inverter, BMS modules, body controllers, ABS controllers, AP module, Bluetooth/WiFi module, etc. Tesla has multiple revisions of every module, and several modules that have had a manufacturing change or chipset change. E.g. some cars have a Bosch radar controller, others have a different manufacturer. Some cars have a revision D drive unit, others have a revision P drive unit. (Model S vehicles can have one of up to twenty drive unit revisions.) Some cars originally shipped with a revision B headlight controller got into an accident and when they rebuilt the car, it now has a revision E headlight controller.

Every one of those different modules, manufacturers, and revisions has different firmware. Thus, there are tens of thousands of different possible combinations of specific car configurations that are in service out on the road. It is not possible for Tesla to test every combination. In some cases, a particular car may actually be unique, in that it may be the only single car in existence with a particular combination of individual modules.

With this type of update mechanism, it is highly risky to apply update packages to every car at once. If the new firmawre for drive unit rev R has a conflict with body controller rev C, you need to find out about that as early as possible and stop the rollout before you brick hundreds of cars. That's just common sense.

In addition to that, these are cars. We're not updating Tweet docks or colored mouse cursor accessories where you can just willy-nilly put out buggy updates and not actually do anything that matters in the real world. Here, if you brick a car's ABS controller, you're going to kill someone. Common sense dictates that you have to take a far more conservative approach.

Your insistence that an update should be available to everyone simultaneously reflects lack of knowledge and shallow thinking.

 

[cdubois]

You’re reading too much into what I said. I meant that once an update version is *validated* then it’s *validated*. It should be released to everyone it applies to simultaneously. This is not hard to understand as it happens that way with pretty much all other software.

Staging software rollouts is quite accepted across the software industry. Multiple companies I have been a part of do this for various reasons including safety, reliability, and even just controlling server bandwidth needed to deploy (hundreds of) thousands of updates simultaneously.

Microsoft stages rollouts with large Windows updates, and for the very reason that they cannot test everything even with an army of internal testers plus public beta testers. They recently even needed to stop & recall a sizable Windows update from the field after they had big problems on a small number of systems. Here is an article discussing Microsoft's slow rollout process and this issue in their own words.

Updated version of Windows 10 October 2018 Update released to Windows Insiders

Here is a noteworthy excerpt from Microsoft's own post:

We intentionally start each feature update rollout slowly, closely monitoring feedback before offering the update more broadly. In this case the update was only available to those who manually clicked on “check for updates” in Windows settings. At just two days into the rollout when we paused, the number of customers taking the October 2018 Update was limited. While the reports of actual data loss are few (one one-hundredth of one percent of version 1809 installs), any data loss is serious.​

Tesla is well within accepted software development rollout practices for such an innovative and complex system. The only difference for Tesla is that they do not allow ordinary users to click "check for updates" as a further control in their rollout cycle.

 

[Mayoo]

Got 28.3 last night

 

[Kappelan]

Got my 28.3.1 least night as well. Looks like now you can attach the key card to a specific profile like you could with the phone before.

 

[Daniel in SD]

You’re reading too much into what I said. I meant that once an update version is *validated* then it’s *validated*. It should be released to everyone it applies to simultaneously. This is not hard to understand as it happens that way with pretty much all other software.

It would be crazy to release safety critical software to hundreds of thousands of cars all at once.

 

[Sklith]

Ever since I installed 2019.28.3.1 on my Model 3, the early collision warning feature has been going nuts while driving in normal stop-and-go traffic. I'm thinking about changing the warning time to late instead of normal.

 

[Firedog7881]

When I ordered my car last year, had this text exchange with a friend of mine. Still feel the same way today.

View attachment 444103

That is some great forethought. I appreciate your hesitation and I don’t blame anyone for not getting FSD, especially if you did your homework and came to the conclusion out of educating yourself. However if you spent $5K on a feature that you weren’t fully aware of the limitations then that is on you. I remember hearing about aggressive timelines that have come and gone but I am ok with progress of any kind regardless if they met an ambitious arbitrary goal that Elon picked out of the air because someone asked for a date. He is overly optimistic and I truly believe he is just ambitious to a fault and not out of malice intent to defraud. I hope I’m right, I can only base my stuff in the same things we all read on here.

 

[GregRF]

I'm on 28.3.1 and set up driver profiles to phones and key cards.

I have now noticed that it seems to link the seat heater to the profiles in some manner. So now if I turn on the seat heater while in Easy Entry profile and switch into gear the seat heater will turn off. I don't think it used to do this.

 

[surfrasch]

hey folks.....loaded 28.3.1 last Thursday, and have run into an issue that seems to coincide with the download. My passenger side window has been open an inch when I come back to the car....after locking and unlocking. No one exiting the passenger side, ....I just walk away and the car is locked (didn't check the window)....then I come back an hour later and the passenger side window is open an inch or so. It's the opening distance used when you open the door. No clue if it's happening directly when I'm existing the driver's side....but now I'm watching out for it.

It's happened like 4x since last week....it's not consistent, seems random, never seen that one before, it's opening on it's own and staying open? Has kind of freaked me out, because it's happened in parking lots.

Anyone else noticed this?

 

[neptunesfinest]

hey folks.....loaded 28.3.1 last Thursday, and have run into an issue that seems to coincide with the download. My passenger side window has been open an inch when I come back to the car....after locking and unlocking. No one exiting the passenger side, ....I just walk away and the car is locked (didn't check the window)....then I come back an hour later and the passenger side window is open an inch or so. It's the opening distance used when you open the door. No clue if it's happening directly when I'm existing the driver's side....but now I'm watching out for it.

It's happened like 4x since last week....it's not consistent, seems random, never seen that one before, it's opening on it's own and staying open? Has kind of freaked me out, because it's happened in parking lots.

Anyone else noticed this?

It's been an issue for a while. It could be one of two things:
1. You're hitting the button by mistake when exiting the car (the buttons are extremely sensitive so you may not even notice)
2. It's actually rolling itself down (this happened to me back in December...twice..both times in the rain!)

Make it a point to check the window when you exit. If you can confirm that it's closed when you walk away and actually rolling itself down sometime after, make note of the time frame and bring the car in for service. They'll check the logs and they'll be able to tell if there was a spurious signal. In my case they had to replace a module in the passenger door.

 

[surfrasch]

Make it a point to check the window when you exit. If you can confirm that it's closed when you walk away and actually rolling itself down sometime after, make note of the time frame and bring the car in for service. They'll check the logs and they'll be able to tell if there was a spurious signal. In my case they had to replace a module in the passenger door.

thanks....that's what I'm doing now....the first time I thought it was my own hand doing it....but then when it repeated randomly it surely has something to do with the window control.