Kobo Bug thread

[Rev. Bob]

Originally Posted by JSWolf

In my CSS, I specified all sizes from 1px to 30px. I even tried em and pt. I also tried with and with an ending semicolon. I did not specify any font size in body as these are not relative sizes but absolute sizes (except for em).

They all came out at whatever size was specified by the font size slider.

Everything from 1 to 30? Try more like 12 to 120; use BIG jumps to make sure you're avoiding the "11px and 12px are the same effective size" and "that's too small to see" problems.

Further, if you're specifying px sizes, the font slider should have no effect on the display. That it does tells me that there's a problem elsewhere in the CSS, resulting in the reader never even seeing this set of rules.

I was proofreading a book recently that looked fine in iBooks and Calibre Edit, but completely shattered under Adobe Digital Editions. I was looking at the CSS for the affected rules, and everything looked good; I didn't get what the problem was. When I looked at the whole document, though, an early rule looked like this:

.c1 {
font-size: 1.25em;mx
}

That syntax error was causing ADE to ignore everything after that line, which was most of the rules, including the ones I was looking at. So even though the specific rules I was scrutinizing were fine, they weren't even being parsed.

[JSWolf]

Originally Posted by Rev. Bob

Everything from 1 to 30? Try more like 12 to 120; use BIG jumps to make sure you're avoiding the "11px and 12px are the same effective size" and "that's too small to see" problems.

Further, if you're specifying px sizes, the font slider should have no effect on the display. That it does tells me that there's a problem elsewhere in the CSS, resulting in the reader never even seeing this set of rules.

I was proofreading a book recently that looked fine in iBooks and Calibre Edit, but completely shattered under Adobe Digital Editions. I was looking at the CSS for the affected rules, and everything looked good; I didn't get what the problem was. When I looked at the whole document, though, an early rule looked like this:

.c1 {
font-size: 1.25em;mx
}

That syntax error was causing ADE to ignore everything after that line, which was most of the rules, including the ones I was looking at. So even though the specific rules I was scrutinizing were fine, they weren't even being parsed.

Please have a look at the following ePub and see if you can find anything not quite right with it.

https://www.mobileread.com/forums/att...2&d=1418401497

[Rev. Bob]

Originally Posted by JSWolf

Please have a look at the following ePub and see if you can find anything not quite right with it.

Well, let's see...

The NCX has loads of invalid entries. The font files aren't declared with the correct MIME type or even the right name. You're still going from 1-30 by single pixels instead of, as I recommended, a wider range by larger jumps. There are a lot of unnecessary rules in the CSS file, as well as a proprietary property (oeb-column-number) in the BODY rules that could possibly cause problems. (Yes, I know it's in the spec.)

Try this version. I multiplied all the pixel sizes by ten (so it's 10-300 instead of 1-30), removed the bad NCX entries, fixed the font references, and removed the O-C-N property. Pulls up fine in ADE, but try it on your device.

[Josieb1]

Originally Posted by Rev. Bob

The last time this happened to me, I did two things to fix it.

0. If you have a recent backup of the database, restore it; problem solved. If not, smack yourself gently on the face with a red snapper.

1. Use an SQLite utility to go to the database table (make a backup first!) that has all the shelves, mark everything "false" in the "_IsSynced" column, and mark everything "true" in the "_IsDeleted" and "_IsVisible" columns. (You can do that with one statement for each column, but the syntax eludes me at the moment.) Then go through and find the shelves you want to keep, marking them "false" in the "_IsDeleted" field. Make another backup, then disconnect the reader and see if the shelf list matches what you want. If it doesn't, try again until it does. Sync the reader to send the changes up to the cloud, and make sure nothing changed on the reader in the process. If you're good, delete the two backup databases and make a fresh backup of the "blessed by sync" version.

2. Make periodic backups, at least of the database. How periodic? Only you can answer that, but "after significantly changing the contents of the reader" is a good time. Note that with the Kobo, this means after letting it scan newly-added books, so the database is up to date and won't have to start by adding several entries. Yes, you'll have to reconnect the device to the computer. Small price to pay.

I don't have a backup (or a red snapper) and as much as i'd love to make one, once my kobo is back to normal, I don't quite understand your instructions. Sorry once you talk in techie code language my brain sort of freezes over

[Rev. Bob]

Originally Posted by Josieb1

I don't have a backup (or a red snapper) and as much as i'd love to make one, once my kobo is back to normal, I don't quite understand your instructions. Sorry once you talk in techie code language my brain sort of freezes over

Okay, lemme try more of a "do this, do that" approach...

I'm assuming here that you want to repair the shelves instead of just deleting them and starting over. I'm also assuming that when the dust clears, you want one shelf representing each set of duplicates, and that shelf should contain everything that was on any of its duplicates. That is, if you have five "Freebie" shelves, you want to end up with one that has all the books that are on any "Freebie" shelf, and so on.

TOOLS:

I use an add-on for Firefox called "SQLite Manager" - grab that and you're most of the way there. Once you have that installed, connect your reader to your PC, go to your reader's .kobo folder, and make a backup copy of KoboReader.sqlite. Now start the SQLite Manager add-on and open .kobo/KoboReader.sqlite with SM.

PROCESS:

In the left panel, you should see a few options, starting with "Master Table". Right under that, there's a "Tables" entry that, on my machine, shows 26 entries inside it. Click the little triangle beside "Tables (26)" to open it, and scroll down to the one that says "Shelf". Click on the word.

The main part of the window should now show all the shelves/collections on your device. The sixth column, "Name", should look very familiar, but the fourth column, labeled "InternalName", is more important. Click on that header, and it'll sort the shelves by name - very useful for finding duplicates, which is where your problems are coming from. As you scan over the list, you should see groups of duplicated shelves with different ID values but the same Name (and, perhaps, InternalName). If they don't have different IDs, we're in trouble and you should stop now (while you haven't changed anything) and let me know that. Otherwise, assuming all the ID values are different, we're in good shape; continue below.

First, we need to make sure all our changes are going to be sent to the server when you sync. Click on the "Execute SQL" tab and paste the following line into the white box:

UPDATE "main"."Shelf" SET "_IsDeleted" = "true", "_IsSynced" = "false"

Click the "Run SQL" button, and you should get a "no errors" message in the "Last Error" box. That's good. What you've just done is told your device to get rid of all your shelves, so you shouldn't stop there! You still have some work to do. Click back on "Browse & Search" and click "Name" to sort everything.

Now you have to fix that "same name" problem and tell the device NOT to delete the shelves you want to keep. It's tedious, and it means editing every row, but it has to be done, since it's the only way to separate the shelves. What I did was add a number to the end of each duplicate's Name and InternalName.

Double-click on a row and it'll bring up an "Edit Record" box. If it's a shelf you're renaming, edit fields 3 and 5 (InternalName and Name) so they match each other, but no other row. For instance, if you have four "Pending" shelves, leave one alone for now and change the rest to "Pending1" through "Pending3" so you can tell them apart. That way, you're keeping one "real" shelf from each set. On the one you don't rename, edit field 7 (_IsDeleted) to say "false". On each row, when you're done with your edits, click OK and confirm the change. The plugin should then tell you to click on Cancel to exit the box.

You may have to visit multiple pages, since the utility only shows 100 rows at a time. The bottom line under this big area handles that; it'll say something like "1 - 100 of 260" with arrows and double-arrows around it. The single-arrow advances to the next page; the double-arrow zips to the end.

Once you're done with those changes, go to Database at the top left, then click Close Database in that menu. (It saves automatically as you work.) Close SM, make a new backup of the changed file, disconnect the reader, and open up your collections view. Poke around a bit to make sure everything looks like you want. (All the "numbered" shelves should now contain zero books, if they're even visible, and the "not-numbered" ones should contain everything they're supposed to.) If that looks good, go ahead and sync. That should remove all the duplicated shelves from the cloud and update the shelves you're keeping, since we've told the device that all the shelves need to be updated.

Is that clearer, maybe?

[Josieb1]

Originally Posted by Rev. Bob

Is that clearer, maybe?

Thanks so much for posting all that, I'm out today so on Sunday I'll copy in all to a word doc and see if I can follow it.

[samhy]

Originally Posted by Josieb1

I've got the duplicated shelves problem back on my Kobo Aura H20 as a result 3.12. I mistakenly left the wifi on

I can't remember how to fix it, can anyone enlighten me, shelves are impossible now I've got 260 instead of three pages.

Maybe you can look here

[Josieb1]

Originally Posted by samhy

Maybe you can look here

Thanks Samhy, I'm just waiting for some announcement as to when the fix is in Calibre.

[PeterT]

Originally Posted by Josieb1

Thanks Samhy, I'm just waiting for some announcement as to when the fix is in Calibre.

It's already in the place it belongs; Kobo Utilities plugin.

[ethan3686]

Originally Posted by ethan3686

Whenever i go into the browser of my kobo glo and search MARS [planet] or any other topic on wikipedia website it tries to load it but the screen freezes in loading. I have to charge or reset the kobo glo to bring it back from the freeze state or it just crashes and automatically goes back to kobo home screen and work normal again. Why is this happening?

^ Can Someone help me with this?

[PeterT]

Unfortunately not much we can do. About the best any of us can do is to pass it o to to Kobo. (Yes.. I did try and yes.. I had a device freeze)

[frostschutz]

Not sure if this is the right place, but anyway...

The Kobo firmware accepts incomplete KoboRoot.tgz (cancelled download, lost connection, ...) and corrupts itself when extracting them. This is because it uses zcat to verify the KoboRoot.tgz, which fails to raise any error for incomplete files, and tar subsequently extracts half an update and even leaves outright corrupted files (like half a library). Worst case the device is bricked. Using tar instead of zcat to verify the tar file would catch most of such errors.

(in /etc/init.d/rcS)

zcat /mnt/onboard/.kobo/KoboRoot.tgz > /dev/null && tar zxf /mnt/onboard/.kobo/KoboRoot.tgz -C /

should be

tar tzf /mnt/onboard/.kobo/KoboRoot.tgz > /dev/null && tar zxf /mnt/onboard/.kobo/KoboRoot.tgz -C /
# or alternatively
tar Ozf /mnt/onboard/.kobo/KoboRoot.tgz > /dev/null && tar zxf /mnt/onboard/.kobo/KoboRoot.tgz -C /

(same for Kobo.tgz)

[JSWolf]

Originally Posted by frostschutz

Not sure if this is the right place, but anyway...

The Kobo firmware accepts incomplete KoboRoot.tgz (cancelled download, lost connection, ...) and corrupts itself when extracting them. This is because it uses zcat to verify the KoboRoot.tgz, which fails to raise any error for incomplete files, and tar subsequently extracts half an update and even leaves outright corrupted files (like half a library). Worst case the device is bricked. Using tar instead of zcat to verify the tar file would catch most of such errors.

Are you saying to change from using zcat to tar and all will be well?

[frostschutz]

I'm suggesting that the best tool to test tar files might be tar, and not zcat. As long as you don't use checksums, anyway.

[tshering]

Originally Posted by frostschutz

Not sure if this is the right place, but anyway...

The Kobo firmware accepts incomplete KoboRoot.tgz (cancelled download, lost connection, ...) and corrupts itself when extracting them. This is because it uses zcat to verify the KoboRoot.tgz, which fails to raise any error for incomplete files, and tar subsequently extracts half an update and even leaves outright corrupted files (like half a library). Worst case the device is bricked. Using tar instead of zcat to verify the tar file would catch most of such errors.

(in /etc/init.d/rcS)

zcat /mnt/onboard/.kobo/KoboRoot.tgz > /dev/null && tar zxf /mnt/onboard/.kobo/KoboRoot.tgz -C /

should be

tar tzf /mnt/onboard/.kobo/KoboRoot.tgz > /dev/null && tar zxf /mnt/onboard/.kobo/KoboRoot.tgz -C /
# or alternatively
tar Ozf /mnt/onboard/.kobo/KoboRoot.tgz > /dev/null && tar zxf /mnt/onboard/.kobo/KoboRoot.tgz -C /

(same for Kobo.tgz)

Could you please explain this a little more. zcat will throw an error as soon as a single byte of KoboRoot.tgz is missing. Won't it?