Weird highlight boxes

[giuli635]

• KOReader version: 2024.03.1
• Device: Android 11 and Arch Linux

Issue

Weird highlight boxes created. This happens throughout the document, but more frequently when is near a math expression.

Steps to reproduce

I just don't know if it's legal uploading the pdf jajajaja.

[Frenzie]

Presumably that's just the math saying it's that tall, unless you're saying it doesn't happen in other viewers.

[giuli635]

It doesn't happen in Evince for example. I've just checked if its a rendering issue but it also saves the highlight that way. But as I said, in Evince I dont have this problem.

(Edit)
Perhaps it's worth mentioning that reflow shows highligths better.

[Frenzie]

Based on some random PDF (https://arxiv.org/pdf/2305.17493.pdf) I'd say it's a bit of a toss up what happens to come out better where. It does generally seem to come out a bit nicer in Firefox, probably mostly as a happy coincidence due to how they lay out text. (It's regular HTML with fancy positioning, which is fundamentally quite different from how native PDF implementations work.)

The (hopefully) good news is that there are likely to be some improvements here in the future regardless, due to various updated dependencies.

Okular

Evince, KOReader on top

[Frenzie]

The (hopefully) good news is that there are likely to be some improvements here in the future regardless, due to various updated dependencies.

You can probably get a preview of that in a newer MuPDF app (whether on desktop or Android), unless it doesn't expose text selection.

[giuli635]

Okay, I'm doing a last recap with all the info I realize could be useful, if you tell me I just have to wait, I'll wait :)

Evince selection before any modification in KOReader

Evince opening the file modified by KOReader

KOReader selection before any modification

Doing this, when I want to use reflow and then select in desktop it crashes, but I don't have time to see the logs, I'll see what's going on later.

The page of the pdf

I leave here the page of the pdf test.pdf.

[Frenzie]

Curious, the geometry doesn't seem out of the ordinary in the PDF but as shown the selection returned is rather tall.

(Some relevant output from mutool trace.)

    <span font="DLYZCT+LatinModernMath-Regular" wmode="0" bidi="0" trm="9.96264 0 0 9.96264">
        <g unicode="𝜆" glyph="4470" x="236.69" y="353.349" adv=".583"/>
        <g unicode="𝑥" glyph="1319" x="242.49822" y="353.349" adv=".572"/>
        <g unicode="." glyph="15" x="248.19684" y="353.349" adv=".278"/>
        <g unicode="𝑥" glyph="1319" x="250.96645" y="353.349" adv=".572"/>
        <g unicode="𝑧" glyph="1321" x="256.66508" y="353.349" adv=".465"/>
    </span>
    <span font="IWOTBY+LibreBaskerville-Regular" wmode="0" bidi="0" trm="10.161893 0 0 9.96264">
        <g unicode="a" glyph="66" x="264.867" y="353.349" adv=".554"/>
        <g unicode="n" glyph="79" x="270.4967" y="353.349" adv=".689"/>
        <g unicode="d" glyph="69" x="277.49827" y="353.349" adv=".675"/>
    </span>

04/09/24-21:40:58 DEBUG dict lookup word: 𝜆𝑥𝑦.𝑦𝑥 {
  "124x270+497+545"
} --[[table: 0x7c9d79e20440]]

But yeah, at the moment there's a good chance it'll be fixed automatically once MuPDF is finally updated.

[benoit-pierre]

Nope, I get the same tall selection boxes with a build using MuPDF 1.24.1.

[giuli635]

So is a problem of MuPDF? Not of KOReader?

[Frenzie]

Yes, here's a screenshot from MuPDF 1.24 for Android.

[Frenzie]

As a random thought, I think the Latin Modern Math font, used here for math, had a tall lineheight issue about a decade ago. Although that wouldn't explain why only MuPDF has this issue.

[giuli635]

So, maybe I should create an issue in MuPDF?

[Frenzie]

I suspect this issue already covers it: https://bugs.ghostscript.com/show_bug.cgi?id=700466

Note how the screenshot of http://beta.hebrewbooks.org/pagefeed/hebrewbooks_org_9717_1.pdf shows a very tall selection in the header, when the box actually hugs the Hebrew characters.

[giuli635]

That's five years old, do you think this will be addressed?

[Frenzie]

Your test case could be useful because it shows a clear contrast between desired and undesired behavior. I think it'd be worth adding there, unless @benoit-pierre has different thoughts.

[benoit-pierre]

I would report the issue upstream, yes.

[giuli635]

Thank you guys, you've been very helpful! Do I leave the issue open?

[Frenzie]

Let's leave it open for now.

[Chlorpheniramine]

• KOReader version: 2024.03.1
• Device: Android 11 and Arch Linux

Issue

Weird highlight boxes created. This happens throughout the document, but more frequently when is near a math expression.

Steps to reproduce

I just don't know if it's legal uploading the pdf jajajaja.

#11716
This look quite similar to the condition I encountered. I propose to merge it to this thread and close it if necessary.