Firmware 4.31.19086

[ownedbycats]

Yes, I've been able to replicate the dirty bit issue by leaving a kepub open during connect.

[DNSB]

Originally Posted by rashkae

Trampling your Database might be a relatively unusual occurrence. But the Kobo leaving the dirty bit set on it's filesystem has been 100% reproducible for me (on two different kobos).. there should be no excuse for Kobo being unable to find this,, it's right in the face if you look for it.

Interesting. Since Kobo updated the fat filesystem check (dosfsck aka fsck.fat) to v4.1 2017-01-24 several firmware revisions back, I haven't seen the dirty bit set popping up the Windows file system check dialog. When Kobo updates to v4.2, that will clear the dirty bit on run regardless of whether it was set or not.

[ownedbycats]

Originally Posted by ownedbycats

Yes, I've been able to replicate the dirty bit issue by leaving a kepub open during connect.

I should also note that this also happens to me on 4.30. (It would explain the handful of scan&fix errors I was getting prior to my initial 4.31 upgrade.)

[DNSB]

Originally Posted by ownedbycats

Yes, I've been able to replicate the dirty bit issue by leaving a kepub open during connect.

Originally Posted by ownedbycats

I should also note that this also happens to me on 4.30. (It would explain the handful of scan&fix errors I was getting prior to my initial 4.31 upgrade.)

What I managed to get with a kepub open during connect was a read-only mount. Again, I couldn't reliably duplicate it.

[rashkae]

Okay,, When I have another bit of free time, I'll try testing on a couple different PC's.. I tried multiple Kobos, but have only ever used my own computer. It would be quite the kick in the teeth if it turns out this only happens with certain usb chipsets, or something tricky like that.

[davidfor]

Originally Posted by DNSB

Interesting. Since Kobo updated the fat filesystem check (dosfsck aka fsck.fat) to v4.1 2017-01-24 several firmware revisions back, I haven't seen the dirty bit set popping up the Windows file system check dialog. When Kobo updates to v4.2, that will clear the dirty bit on run regardless of whether it was set or not.

Originally Posted by DNSB

What I managed to get with a kepub open during connect was a read-only mount. Again, I couldn't reliably duplicate it.

The issue seems to be that sometimes the kepub renderer is not closing a file if a kepub is open when the device is connected to the PC. That means that sometimes the file system is not unmounted cleanly and the dirty bit is set. Then sometimes when you disconnect the device, the filesystem is mounted read only on the device.

The problem for working it out is the "sometimes". For me, it didn't happen all the time. And each of the "sometimes" above didn't happen every time. I think I was seeing about 50% for the dirty bit but less than 10% for the read-only. I assume there was something in the books, or the user configuration, or something like how long between turning a page and plugging in, or when the last sleep was (that commits the reading session) or some other absurd sounding thing going on. And my devices all behaved a little differently. I could never figure it out. Which I think you saw elsewhere.

[DNSB]

Originally Posted by davidfor

And my devices all behaved a little differently. I could never figure it out. Which I think you saw elsewhere.

I do remember seeing that. Having had to deal with a few erratic issues (my memories of race conditions are not among my fond ones), I feel quite a bit of sympathy for anyone who is having to fix one. Give me a nice consistent problem that I can hit with that big hammer...

[svenlind]

Originally Posted by ownedbycats

If you use Calibre, get Kobo Utilities and set it up to backup your device on connection; it'll automatically handle it for you. Restoring is a few more steps: you need to unzip the last good backup and plop the files into the .kobo folder.

Thank you for that info

The possibilities of Calibre will never cease to amaze me.

[compurandom]

Originally Posted by davidfor

The problem for working it out is the "sometimes".

I think part of the problem is that there are at least 4 separate and distinct failure modes here, all with different causes and symptoms, and each with a multi-step, timing sensitive trigger.

[John F]

Originally Posted by davidfor

The issue seems to be that sometimes the kepub renderer is not closing a file if a kepub is open when the device is connected to the PC. That means that sometimes the file system is not unmounted cleanly and the dirty bit is set. Then sometimes when you disconnect the device, the filesystem is mounted read only on the device.

The problem for working it out is the "sometimes". For me, it didn't happen all the time. And each of the "sometimes" above didn't happen every time. I think I was seeing about 50% for the dirty bit but less than 10% for the read-only. I assume there was something in the books, or the user configuration, or something like how long between turning a page and plugging in, or when the last sleep was (that commits the reading session) or some other absurd sounding thing going on. And my devices all behaved a little differently. I could never figure it out. Which I think you saw elsewhere.

While I can appreciate that it may be hard to reproduce consistently, 50% of the time sounds rather good for reproducing a bug.

And if there are user's who can reproduce it 100% of the time, Kobo could reach out to the users for help in reproducing it.

If they thought it was worth it.

Although being fed bug reports with a silver spoon is always nice.

[JSWolf]

I am surprised that 4.31 has not been pulled.

[MGlitch]

Originally Posted by John F

While I can appreciate that it may be hard to reproduce consistently, 50% of the time sounds rather good for reproducing a bug.

And if there are user's who can reproduce it 100% of the time, Kobo could reach out to the users for help in reproducing it.

If they thought it was worth it.

Although being fed bug reports with a silver spoon is always nice.

It’d be up to the users to contact Kobo as the users are the only ones who know if they’re reproducing the issue 100% of the time. I’ve reported issues before, the last one I recall was with the Forma and some issue with the Lock Screen in the days and weeks after the Forma released. Reporting was easy just use the feedback button on Kobos site and list the steps to produce the issue. Anyone here who’s been able to reproduce it 100% has already spent longer posting about it here than it would take to report it to Kobo.

And yes Kobo employees read this forum but are likely hindered by the inner workings of how software changes can be made. Should this change? Sure. But it’d be faster to deal with the system we have than to hope a new one replaces it and then allows for these bugs to be fixed.

[davidfor]

Originally Posted by John F

While I can appreciate that it may be hard to reproduce consistently, 50% of the time sounds rather good for reproducing a bug.

And if there are user's who can reproduce it 100% of the time, Kobo could reach out to the users for help in reproducing it.

If they thought it was worth it.

Although being fed bug reports with a silver spoon is always nice.

That's 50% of my times. Not 50% of everyone's connections. And overall, it was less than 10% of the connections when there was an real problem. The dirty-bit wasn't really an issue. It was when either the database became corrupt, or the file-system was remounted read-only. So, a lot harder to track down. I was never able to reproduce it consistently. And I know Kobo were not either.

[Calenorn]

Based on what has been observed so far, could we make a list of best practices that will minimize the risk of database corruption when using the 4.31 firmware?

[JSWolf]

Originally Posted by Calenorn

Based on what has been observed so far, could we make a list of best practices that will minimize the risk of database corruption when using the 4.31 firmware?

One of the things to do is to go back to the home screen before you load any eBooks be it side loading or syncing.