Display problems in book viewer

[haertig]

I'm running the latest version of Calibre on Linux: version 3.10

If I jump forward in a book, the viewer loses its brains for wrapping text. Each paragraph becomes one long line that disappears off to the right of the viewer, never to be wrapped. This is a repeatable problem.

Example:

Book: "Relic", by Douglas Preston & Lincoln Child
Open in viewer: Progress indicator at upper left of viewer displays "1.0/728"
Overwrite "1.0/728" with "300", hit enter, to jump to page 300 of the book
Note that text does not wrap
Note that purple "Next Page" button does not advance page
Randomly jump around in book using vertical scrollbar on right of viewer
After a bit of random jumping, wrapping starts to work again

This happens in other books too, I just picked "Relic" as an example since it's a commonly found book that others may well have to test for themselves.

The two attached screenshots below illustrate the "no wrap" error case, and the "correct wrap" case (after jumping around randomly and returning to page 300).

[kovidgoyal]

You need to attach a book where this happens, I cannot reproduce on my linux system with a couple of random books.

[haertig]

Thanks for the quick response.

Attaching books would require me stripping the DRM and then putting them out there in the public arena. That is not something I will do. I can't distribute those, even for a noble cause such as this (for you/others to troubleshoot). Such is the stupid DRM world we live in.

I will be glad to help in other ways if possible - sending a log file or something else that Calibre (or Linux) may make available for troubleshooting.

I tried about a dozen of my books. They ALL behaved as I illustrated above, not one single one behaved normally. This test was done for books that I bought from Amazon, books that I got free from Amazon (Prime), and books that I bought from B&N years ago. I tried AZW3, EPUB and MOBI formats - these all failed in the same manner.

Obviously, this must be something specific to my particular computer/OS/installation since you cannot duplicate it, yet it is 100% repeatable for me on every book I tried.

This is not a big deal for me, since I read my books on an eReader, not in the Calibre viewer. I just happened to notice this behavior randomly, I have no idea why I was originally jumping forward in books, or even why I was using the Calibre viewer in the first place. I was just playing around I guess.

[BetterRed]

Originally Posted by haertig

Thanks for the quick response.

Attaching books would require me stripping the DRM and then putting them out there in the public arena. That is not something I will do. I can't distribute those, even for a noble cause such as this (for you/others to troubleshoot). Such is the stupid DRM world we live in.

Either create a bug report here ==>> Bugs : calibre, attach the book and then mark the report 'private'.

Or obfuscate the EPUB with the ScrambleEbook plugin and post it here.

BR

[haertig]

Well, alrighty then! A pretty cool plugin. Calibre developers and Mobileread users think of everything!

I picked a shorter book, so the resulting scrambled file is smaller. I hope it is small enough to not run afoul of any forum rules on upload size (you'd think the website user interface for uploads would automatically enforce size limits).

I still think it's probably something peculiar to my setup, and not the ebook itself (since they ALL exhibit the behavior for me, and none do for kovidgoyal). However, attached is one such book that I manually verified as exhibiting the problem.

In the attached EPUB, jump to page 200 by overwriting the progress field in the upper left of the viewer. You should see a line that starts with "USU HBKP". And you should see that the text does not wrap.

Next, drag the scrollbar on the right up and down a bit randomly, then click outside of the scrollbars drag handle a few times, then drag it some more, then click some more until you see the text wrapping start to work again. Then drag the scrollbar back to where you see "USU HBKP" again. This is on page 197 at this point, not the page 200 it was originally found on. Now, you should see that the text is wrapping appropriately. The attached screenshot shows the book at the end of the test described above, when the text is wrapped appropriately. I forgot to take a screenshot of the no-wrap error condition (and have since deleted the scrambled book from Calibre).

[BetterRed]

Originally Posted by haertig

I still think it's probably something peculiar to my setup, and not the ebook itself (since they ALL exhibit the behavior for me, and none do for kovidgoyal). However, attached is one such book that I manually verified as exhibiting the problem.

Must be me too then, because I see the behaviour you describe, when the viewer is in 'Paged Mode' on Windows 10. But only if I enter the 'page number' 200 and hit enter, if I slide the scroll button down to the "USU HBKP" line (or hit PgDn till I see it) I'm on page 197 with line wrapping.

BR

[kovidgoyal]

Doesn't happen for me on either windows or linux, jumping to position 197 or clicking the chapter in the table of contents panel.

Try running the viewer as

calibre-debug -w file.epub

see if any errors are printed in the terminal when the problem occurs.

[haertig]

Here's what happened in the terminal (I used [ code ] tags below, so you'll have top scroll side-to-side to see all my comments):

david@Davids-Linux-Desktop ~/Desktop $ calibre-debug -w ./scrambled.epub 2>&1

Initial opening of the ebook (I didn't hit any keys)
Turning on automatic hidpi scaling
devicePixelRatio: 1.0
logicalDpi: 96.0 x 96.0
physicalDpi: 81.5518394649 x 81.6428571429
Using calibre Qt style: True
Viewport cfi: /2/4/2/2[title^-image]@0:0

Then I hit "200 <enter>" and the following immediately appeared, and text wrap was screwed up
ReferenceError: Can't find variable: book_indexing

Then I hit "300 <enter>" and the following immediately appeared, and text wrap was still screwed up
ReferenceError: Can't find variable: book_indexing

Then I hit "200 <enter>" and there was NO ReferenceError this time, and text wrapping was CORRECT, but page 200 was not the same page as it was originally

Then I hit "197 <enter>" and I was back at my original place (that used to be page 200), and text wrapping remained CORRECT

Viewport cfi: /2/4/28/1:1
Viewport cfi: /2/4/4/1:0
Viewport cfi: /2/4/4/1:0
david@Davids-Linux-Desktop ~/Desktop $

The "Viewport cfi" lines appeared asynchronously (they didn't appear to be directly related to any of my keystrokes, they just popped up whenever they felt like it).

[haertig]

I thought about this for a while - why would the "book_indexing" variable not be found initially, but evidently it is found later? At least Calibre stops complaining about it later. So I though maybe this was a timing issue - when you first open the eBook you have to wait for indexing to happen, or something like that.

So I opened the eBook and walked away from 3 minutes, without hitting any keys. This should have given Calibre enough time to do whatever automatic indexing it wanted to do (since my previous tests all lasted less than one minute, I figured 3 minutes would be long enough to wait).

Anyway, after waiting those 3 minutes, the problem still occurred and the ReferenceError still showed up in the console, just as before. So I guess this proves my "timing issue" theory was a wrong assumption.

[BetterRed]

FWIW - when I type 200.0 in the 'page' number and press enter I end up at 200.2 with no line wrap, when I page up and down a bit I end up at 197.0 with line wrap.

I tried it out on two other books, same behaviour, one went to 199.9 with no line wrap, after a couple of page ups and downs it landed on 145.0 with line wrap.

Seems like transient behaviour when book is first opened, after that line wrapping seems to be OK.

Normally I never look at that page number box, much less type into it.

BR

[kovidgoyal]

Something is going wrong during the viewer initialization after the HTML is loaded. When you jump to another location far enough away a new HTML file will be loaded and the viewr does initialization again. Apparently after a couple of such initializations it settles down. The problem is there is no good way to figure out what is causing the initial issue, especially since I cannot replicate it.

[haertig]

This is really a non-issue for me. I just happened to notice the behavior when I was playing around, so I thought I'd report it. Chances are nobody else has, or ever will, notice it again!

Calibre is excellent software. I really appreciate all of your brain power and dedication that went into creating it. A gift to the world that you should really be proud of. Thank you!