I'm over here with a plunger and snake, clearing a clog out of the 
listserver.

So this is what happened.

Phil posted a very long message to the list, with a Tesla process list, and 
the listserver held it because of its size.

Phil CCed Lawrence, and the replies went back and forth between them, ALSO 
going to the server but not being distributed.

And the reason they weren't distributed is that Phil and Lawrence top-posted 
all the way through, so a 256k byte message stayed tacked on to every 
message.   By the last post, the message had grown to 352kB.

Folks, please don't do that. To terminally mix my metaphors, it gives the 
listserver indigestion.

To set things straight, here's the blocked thread, condensed to a manageable 
size:

==========

Lawrence, Thu, 18 Apr 2024 12:30:33 +0000 (UTC) 

> all modern software on complex systems with a large OS will have some
> amount of memory leaks, so it's always a good idea to do an occasional
> reboot to clear things up.  I've seen Tesla uptimes without reboots of
> over 2 years, but they are usually running like molasses by then. 

That's such a defeatist attitude.   Memory leaks aren't inevitable and there 
are lots of tools to find them.    It's the result of shoddy, lazy 
programming, not some fact of nature.      Howevercomplicated you think the 
Tesla software is, the Linux OS is much bigger project and more complicated 
and luckily Linus has always had a low tolerance for stupid excuses.  

-----

Phil, Thu, 18 Apr 2024 08:14:20 -0700 (15:14:20 UTC)

You can't compare an embedded system designed for industrial use to a large-
surface GUI on a large OS with a lot of apps.  

You probably interact with a handful of systems every day that run linux. 
They all run many layers, and It probably usually isn't the Kernel.   Do you 
have the source for the top layers?  Do you even have a shell, let alone 
root? Even if you are a developer of your caliber, and somehow had source 
and access, (almost never true) it's simply not feasible to attempt to debug 
someone else's system for some memory leaks that may cause reduced 
performance after weeks of uptime.  

Here's the process list on a Tesla Model 3 MCU just to give you an idea of 
what you'd be up against:  

(See original message for list)

*****

-----

Lawrence, Thu, 18 Apr 2024 15:28:38 +0000 (UTC)

I'm not saying it's "your" problem to solve.   It's the guy who wrote it. 
And getting the process list is a good place to start.   I'm actually not 
overwhelmed by it.     Most of the processes have a pretty miniscule memory. 
  Sort by memory usage and compare over time.   As I said I'm not a Tesla 
guy, but I used to have a pretty good understanding of Linux.     I see some 
of them use "QT".   Personally I think "QT" has become very bloated from 
what used to be.    I wouldn't use a "QT"application anymore.   That's my 
opinion anyway.   I also see "minijail".    Is it working? This isn't rocket 
science.   It's just work.  

-----

Phil, Thu, 18 Apr 2024 08:44:19 -0700 (15:44:19 UTC)

On a system this large there is no one "guy".  It's made up of many years of 
legacy code with a team of hundreds of different developers over time. They 
aren't going to just throw out QT and re-code it from scratch. There's too 
much inertia, and no matter what you think, nobody cares about fixing the 
leaks.   I'd argue the Linux kernel itself is highly bloated compared to 
where it was when I started using it.  

-----

Lawrence, Thu, 18 Apr 2024 15:53:21 +0000 (UTC)

Of course it's not one "guy".    But it most likely is "one guy" who made 
the mistake. Anyway, if nobody cares about memory leaks, then they will 
never get fixed...and lowering standards probably isn't the path to 
excellence.  

I personally am stubborn enough to care. Anyway, since you have the data, 
why not at least narrow it down to what process is growing over time?    
That ought to give you a path toward fixing it.....and you were the one to 
ask me what I'd do...and I'm telling you.  

-----

Phil, Thu, 18 Apr 2024 09:36:57 -0700 (16:36:57 UTC)

I have way better things to do, and I don't have source code, I'm sure as 
hell not going to bother decomp.  

I don't mind an occasional reboot.

-----

Lawrence, Thu, 18 Apr 2024 17:02:58 +0000 (UTC)

I completely understand.   As I tried to say it's not your problem, it's the 
people who wrote it.     

As I said, I don't have a Tesla, but looking through your process list, I 
see a lot of what looks like QT apps.    As an early user of QT, it used to 
be pretty good when it was 10 people and made really pretty X apps when most 
of the stuff looked pretty stodgy.    But honestly the good things about it 
was that drew standard widgets and menus....but they've tried to make it way 
bigger than was necessary and tried to lock the user into using their thread 
stuff, their SQL stuff, their OPENGL, etc.    IMHO I'm a smart enough to 
make my own decisions about that.   I just want it to draw some menus and 
get some callbacks for the mouse/keyboard.    I had to debug through their 
SQL stuff one time because it didn't work, and that was enough for me.  I 
think it's gotten ridiculous and I don't want to be forced into using their 
paradigms, their IDE's.   I'll pick my own thank you. If I were Musk, or who 
ever, I'd seriously look at gnome.    Most of the linux world seems to be 
using it and not QT anymore.   I'm guessing that others have come to the 
same conclusions as me.   For peoplewho actually understand these things, 
most GUI stuff is more or less the same when you get down to it.   There is 
no reason it can't be changed when your horse is not going in the direction 
you want it and won't change, you should probably change horses.  


David Roden, EVDL moderator & general lackey

To reach me, don't reply to this message; I won't get it.  Use my 
offlist address here : http://evdl.org/help/index.html#supt

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = 

     So if you ever need info about anyone at Harvard, just ask.
     I have over 4,000 emails, pictures, addresses, SNs.
     People just submitted it.  I don't know why.
     They trust me.  Dumb f***s. 

                 -- Facebook's Mark Zuckerberg, 2003/11/19
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = 

_______________________________________________
Address messages to ev@lists.evdl.org
No other addresses in TO and CC fields
HELP: http://www.evdl.org/help/

Reply via email to