KOReader: a document reader for PDF, DJVU, EPUB, FB2, HTML, ... (GPLv3)

[knc1]

Originally Posted by anton79ru

Guys, I've been using KOReader for three weeks now and I have encountered several instances of my PDF files becoming corrupt ...
- - - -

How do you know that the PDF file **on the Kindle** was not corrupt to start with, and that caused KOReader to crash?

Any chance that the FAT32 file system of USB mass storage is not working correctly or is itself corrupted?

How are you putting the PDF files onto USB mass storage of the Kindle?
Do you always "safely remove" and then "eject" the device?

[donMaranzano]

The thing what I'm missing is sharing quote in facebook. It this app supporting this function?

[knc1]

Originally Posted by donMaranzano

The thing what I'm missing is sharing quote in facebook. It this app supporting this function?

No.
You would be best to start a new thread for this different topic.

[donMaranzano]

No, I'm not hijacking, I thought that the app has that function. I thought it's function of the app.

[knc1]

Originally Posted by donMaranzano

No, I'm not hijacking, I thought that the app has that function. I thought it's function of the app.

try reading the post again please.

[anton79ru]

Originally Posted by hawhill

Hm. Do you run other software? ... it might leave some hint in the file "crash.log" as of why it crashed... at some point of document complexity, it might well crash, unfortunately. The newer versions *can* write to PDFs, but they will do so only for saving highlights.

Just KOReader, apart from the stock firmware. The crash.log file is empty. Most often, it happens when I close one file to open another. My PDFs vary in size from 500K to 2M or more. I do highlight a quite a few times in a single file.

At which point in time does KOReader save highlights? I assume that IF it happens after I close the file, AND it takes several seconds to accomplish, AND I try to open another file before the saving operation is complete, THEN it might be the case... I will try to reproduce that.

[anton79ru]

Originally Posted by knc1

How do you know that the PDF file on the Kindle was not corrupt to start with, and that caused KOReader to crash?

Any chance that the FAT32 file system of USB mass storage is not working correctly or is itself corrupted?

How are you putting the PDF files onto USB mass storage of the Kindle?
Do you always "safely remove" and then "eject" the device?

I download academic articles from libraries like Hein or Jstor - those are my files. Before they get corrupt, I'm able to read them on the Kindle and after they get corrupt, I do check the original downloaded file on my PC, so I know they're fine.

Never got any file system or USB errors. I either drag and drop or copy and paste files in Windows, from the local folder I download them to. And yes, I always safely remove the device.

Thanks for the checklist.

[knc1]

Originally Posted by anton79ru

Never got any file system or USB errors.

**I either drag and drop or copy and paste files in Windows**,

from the local folder I download them to.
And yes, I always safely remove the device.

Thanks for the checklist.

Hmm, that might be the problem - -
Windows (of all versions, since forever) has had problems with gratuitously changing <lf> to <cr><lf> in what should be a binary file (PDF files are **binary**, not **text**).

Do a binary compare to the original (presumed good) and the file on the Kindle to see if that is the "corruption" that has happened.

[NiLuJe]

Because I'm still a bit fuzzy, when you think one PDF gets 'corrupt', how does it render on a full-fledged PDF reader on your computer (ie. load the one from the Kindle over USB, not your original)?

(That might sound stupid, but I'm still confused, and wondering if you're not mixing 'corrupt' with 'sudden broken rendering in KOReader').

[hawhill]

However, I think we should test the writing into PDFs more closely. Maybe we should write to a new file and then do a (atomic, at least in theory) rename().

PS: As for sharing quotes - just recently, a HTTP plugin got into KoReader. Sharing quotes to Facebook is not there yet (I think), but it might well be in the not-so-distant future.

[anton79ru]

Originally Posted by knc1

gratuitously changing <lf> to <cr><lf>

Does it really matter for scanned files? Can screwing up this parameter result in deletion of entire pages? It does not happen right after I copy files onto the reader, but after certain time during which I read and highlight.

Originally Posted by NiLuJe

when you think one PDF gets 'corrupt', how does it render on a full-fledged PDF reader on your computer (ie. load the one from the Kindle over USB, not your original)?

This is a valid question. I'm really talking about irreversibly damaged files in this case. Even the size of the file is not the same (like ten times less) when checked on a PC with the Kindle connected via USB. Stuff is plainly deleted.

[anton79ru]

Originally Posted by hawhill

Maybe we should write to a new file and then do a (atomic, at least in theory) rename().

Another solution may be to create a backup copy at each instance of opening a file, then compare at the end of session and if the resulting file differs from the backup more than it should - replace the current file with the backup copy, otherwise - erase the backup and create a new one the next time that file is opened.

Does it make any sense to you or it's just another way to say what you have in mind?

[knc1]

Originally Posted by anton79ru

This is a valid question. I'm really talking about irreversibly damaged files in this case.
Even the size of the file is not the same (like ten times less) when checked on a PC with the Kindle connected via USB.
Stuff is plainly deleted.

1) ...
2) ...
3) Ah, no - that is just the file being truncated at the current file pointer position when the file is closed.
Probably a typo in the flags that the file was opened with or a thinko in the file closing.
(Directly adding a highlight mark would require merging the content prior, the highlight, and the content following - here you describe appending the highlight to the content prior and the content following being dropped into the bit bucket.
There is a reason that many systems keep things like added highlights in an external file.)

[hawhill]

I did not yet have the time to check the logic with koreader. It's most certainly not a general failure, since the functionality has been tested. It must be some kind of corner case.

anton79ru: that might be a way for testing when/if corruption happens at all, but that route seems to be overkill. If the file is not opened for writing at all, it just cannot be the fault of KoReader. But mitigating failure conditions is the second step, the first would be to find out why they happen. A job for my weekend, maybe.

[ignacy]

I have recently bought a Kindle Paperwhite 2 and installed KOReader, specifically koreader-nightly-20140425 (+ KUAL v2.3, MKK v20140214 & jailbreak), a few dictionaries and Tesseract 3.02. I am really impressed by the software, so kudos to the developers and all those who made this magnificent work!

There are a few minor things that don't seem to work, and since I haven't found any information about it, I thought I'd ask here before creating a bug ticket on GitHub.

• The KOReader User Guide says: "display config menu: single tap on the lower center of screen."
  When I single tap on the lower center, it just displays the following page. Single tapping on the upper center displays both the main and the config menus.
• Tapping and holding on a single word does display the definition of the word and the << and >> buttons are available if different dictionaries have entries for the word. However, the Highlight and Add Note buttons are grayed out (disabled).
• If I tap, hold and drag, so that more than one word is selected, a context menu offers five options, but only two are available. Highlight and Translate are enabled, but only Highlight works (Translate does not seem to do anything.) The options Add Note, Share and More are grayed out.
• The FileManager doesn't use the whole screen to list the directory contents (that is, there is a gap between the bottom of the list and the "page 1/N" text at the bottom).

Is that behavior normal for the version I have? (The latest version as I'm typing this.)

And some things that you might want to add in the future:

• In the Fontlight Level window: a cancel button and the current level being used.
• Ability to zoom to 200% (and pan by dragging, ideally with a mini map).
• Search for text.

Thank you.