Firmware 4.31.19086

[DNSB]

Originally Posted by ownedbycats

Kobo uses SQLite.The documentation has a list of very specific ways corruption may occur.

I've looked at that page a couple of times and it seems to be a miracle that any database works for more than 3 minutes without corrupting.

[Quoth]

SQLite is better than Access*, but certainly not multiuser 0r multiprocess. Used as intended it's stable and free from corruption on a suitable filesystem, or correctly used filesystem.

(* Access used Jet Database Engine. Which is why MS released the MSDE, originally a cut-down SQL2000, in 1999/2000. Jet was / is terrible. Access and VB6 could use MSDE via ODBC. MS SQL was bought in from Sybase at about version 6)

[compurandom]

I think blaming sqlite for this issue is just distracting. The real issue is the filesystem sqlite is sitting on and its lack of proper shutdown at USB plugin.

Yesterday, I powered on and plugged in my '30 firmware device, and noticed there was a long delay before it would detect the USB, even with the force USB option in NickelMenu. I wonder if something has gone wrong with that detection and '31 is allowing it to connect too early.

I've said a couple of times I've had database corruption on firmware versions prior to '30, but I didn't have '30 installed long enough before it upgraded to '31 to see one there.

[ownedbycats]

Yeah, I was just posting it because... well, it takes something to corrupt a database that has rollback journals.

[WormGod]

Originally Posted by DNSB

At some points, I've had the available memory drop to below 9MB and often after switching fonts a couple of time, I've seen reboots.

How much available memory would you expect to see under normal circumstances?

And once it gets low, do you see it continue to drop when doing nothing for awhile except checking periodically?

[NiLuJe]

There's indeed only so much you can do to paper over the fact that FAT32 should just die a fiery death ;o).

And, ideally for this use-case, be replaced with a filesystem that makes sense (i.e., is actually journaled, not like you, exFAT ;p).

TL;DR: This is (unfortunately) why stuff moved away from USBMS: the whole thing is a pain to integrate into a system properly in the first place, and it usually sits on top of a truly terrible filesystem, putting it in the very situation it's the absolute worst at handling safely :/.

[ownedbycats]

Originally Posted by NiLuJe

There's indeed only so much you can do to paper over the fact that FAT32 should just die a fiery death ;o)

Ah yes, the filesystem that scrambles itself if you sneeze near it. But 'compatibility.'

[DNSB]

Originally Posted by WormGod

How much available memory would you expect to see under normal circumstances?

And once it gets low, do you see it continue to drop when doing nothing for awhile except checking periodically?

If I leave the Sage alone for a couple of hours, the available memory will increase marginally.

I am more concerned with the available memory dropping below 10% of the total memory. On my Sage, a couple of times before a spontaneous reboot, I saw numbers such as 340MB total memory, 18MB free, 9MB available. Right now, I see 340 total, 129 used, 85 free and 189 available. Switching fonts a couple of time including a switch to AR Ming drops the free and available memory numbers like a paralyzed falcon and also triggers the Your eReader is not authorized to open the book popup. Interesting since the current ebook was my font tester epub and I don't remember adding DRM to it.

[Rid]

Originally Posted by DNSB

and also triggers the Your eReader is not authorized to open the book popup. Interesting since the current ebook was my font tester epub and I don't remember adding DRM to it.

I have had this issue on my Aura One with kobo and sideloaded books (non DRM). Also had issues of books not showing on the reader after a sideload (had to turn on and off - or connect to computer and eject again).

Since I have been using this firmware I have done 2 resets and restored the DB the last time - it went to the setup screen. Thanks to JSWolf he mentioned that the Kobo Utilities plugin can do an automatic of the DB, if I had known this (and was not lazy doing backups of course) maybe the first 2 resets may not have been needed?

Since a couple of days ago I have started using 4.30.18838 and so far it seems OK...

[ownedbycats]

So I uploaded a new book to my device.

This time, instead of rebooting, I decided to then read one chapter of my ePub before reconnecting in hopes that it would give Kobo enough time to finish up whatever import processes might still be running.

Whether it was that or connecting with the ePub open, no errors!

[PeterT]

I experienced the DB corruption last night. Hadn't synced to calibre in a long time; I've been reading library books fulfilled from the Kobo store for some time, but wanted some other books.

Carefully I ensured I had no open books, connected to calibre, sent some books, disconnected and waited for the import to finish. Reconnected via NickleMenu and ... db corruption.

Luckily I had the backup from just before the 2nd connection, so quick restore and all was fine.

I wonder if post import the Kobo has to process the books to extract the cover, and that it's that step is preventing the dismount?

Sent from my Pixel 4a using Tapatalk

[ownedbycats]

Originally Posted by PeterT

I wonder if post import the Kobo has to process the books to extract the cover, and that it's that step is preventing the dismount?

I was thinking perhaps something like that, or for some reason the import process itself isn't properly shutting down -- rebooting prevents corruption, which seems to support it.

[jackie_w]

Originally Posted by PeterT

I wonder if post import the Kobo has to process the books to extract the cover, and that it's that step is preventing the dismount?

I always get the calibre driver to create the cover thumbnails/screensavers, during send-to-device, rather than letting the Kobo import process do extractions. I wonder whether that's why I haven't had any database corruptions?

[ownedbycats]

Originally Posted by jackie_w

I always get the calibre driver to create the cover thumbnails/screensavers, during send-to-device, rather than letting the Kobo import process do extractions. I wonder whether that's why I haven't had any database corruptions?

No, I have the driver send the covers and I've gotten corruptions.

[jackie_w]

Originally Posted by ownedbycats

No, I have the driver send the covers and I've gotten corruptions.

Just been lucky then, I suppose