Root Access

Michael Frank

5:39 AM

@bob I accidentally click "Start new session"... Help?

Bob

6:16 AM

@MichaelFrank ??

Michael Frank

6:58 AM

@Bob In Firefox.

I did not click "Restore Session", now all my tabs are gone.

I know you are TabMaster, so please sensei, you're my only hope.

Bob

7:16 AM

@MichaelFrank Open your profile folder via about:profiles and go to sessionstore-backups

copy everything there to another folder

and then look for the most recent biggest file

close firefox, rename file to previous.jsonlz4, and open firefox again

(and pull from backup folder if it fails, cause it'll clobber)

what you can get back depends on whether you (a) killed firefox [via task manager] asap (recovery might not be clobbered if within 30s), (b) didn't close firefox yet (previous might not be clobbered; iirc it writes on save though it might clear when you click start new?), or (c) you can only recover from the upgrade backup from the last (minor) release upgrade

also if you want a faster resp poke me on matrix next time, I actually get phone notifications there :P

djsmiley2k

guess who'[s back

back again

wow that's le suck

both the sudo thing, and firefox handling of restoring tabs after accidently not restoring them

in chrome i go 'history -> recently closed and they're all there (if all in one window it'd show "14 tabs"

Biswapriyo

7:46 AM

What are those tools called which measure latency b/w a keypress and monitor? I saw some YouTube videos where people put the device on the display and measure the latency.

Bob

7:56 AM

@Biswapriyo I like to call them cameras

(seriously though, you can emulate them with a phone camera. bonus points if you have a slow mo mode)

Burgi

morning

Bob

8:43 AM

@djsmiley2k Actually, now that you mention it, I haven't been asked to restore/not-restore in quite a while.

djsmiley2k

o_O

Bob

I have mine set to always restore but if that's disabled the old session seems to still be accessible via the history menu *shrug*

djsmiley2k

as in it auto restores?

yeah

Does it only ask, if it's on auto, when it's crashed out?

because it maybe due to a tab causing the crash?

Bob

@djsmiley2k good point, that's the only place it still asks

Michael Frank

10:02 AM

@Bob Yea... It seemed to be clobbered.

Oh well!

Ave

10:17 AM

bleh, kernelcare.

that was 2 months ago

Ave

10:43 AM

Blog bot

11:15 AM

stackoverflow.blog/2019/10/15/…

bwDraco

tomshardware.com/news/… wat

NAND endurance is a problem, and here we have an industrial application managing to wear out flash.

djsmiley2k

11:47 AM

@Blogbot "we got the chance to sit down with Prashanth Chandrasekar, Stack Overflow’s new CEO, to hear about how he got into computer science and his vision for this community and company." lul

@bwDraco lol

Why on earth are they logging to the firmware storage location

bwDraco

12:18 PM

Yeah. Stuff that's expected to write lots of data over the life of the system should be on a separate, readily-replaceable board.

Also, use the correct type of storage! Choose high-endurance MLC flash if you need to write dozens of gigabytes to it every day for a decade or more.

This is a freaking automotive system. There is no reason to skimp on the storage.

If MLC isn't an option, at least properly size and overprovision the storage.

People just don't get that NAND endurance is not unlimited. I would expect folks working on automotive systems to know better.

And if it's nearly full, beware of write amplification (that's probably the thing they missed).

insideevs.com/news/376037/tesla-mcu-emmc-memory-issue

> 8 GB eMMC

What the fuck.

Why would anyone seriously believe that such a small flash memory chip could handle dozens of gigabytes of writes each day for a decade?

Also, that chip is almost full. Write amplification will kill it.

Totally idiotic.

djsmiley2k

12:41 PM

@bwDraco developers are stupid in everything other than development. I'm not even joking here.

I have developers who don't understand networking, sysadmin, ops, or anything else. They just know code.

bwDraco

1:05 PM

IMO the best way to cope with the limited endurance of flash memory in a complex industrial system is to treat it as a consumable. All storage will eventually fail, but flash memory has write endurance limits that must be accounted for.

Bob

@djsmiley2k :<

@bwDraco Meh.

Rotating logs are the poster child for flash-optimised filesystems that deal primarily with append writes.

Just configure the right buffer size and you have exactly zero write amp.

djsmiley2k

@Bob not all devs obv.

bwDraco

@Bob The infotainment software shares space with the logs. The system needs to do at least some wear-leveling so there will be a nonzero amount of write amplification.

That, too, is an idiotic decision.

You can't expect all logging to happen within a mere few hundred megabytes of space, repeatedly overwrite that limited area without any amount of wear leveling, and expect the flash to last long enough for an industrial application where it needs to operate at least a decade.

Bob

Ahem.

See: "flash-optimised filesystems".

bwDraco

True but I'm inclined to think they just used ext4 rather than something like F2FS. It's unlikely that the problem of NAND endurance even crossed the developers' minds.

Biswapriyo

5:03 PM

Does there something happen in SU? I got rep. down with "user was removed". I also see this in other >10K rep. users' activity.

djsmiley2k

5:43 PM

not just in SU.

djsmiley2k

7:27 PM

mmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm

bwDraco

nnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnn

djsmiley2k