Firmware 4.35.20400

[Aleron Ives]

You can't update because 4.34 is the latest version. Kobo pulled 4.35 for unknown reasons.

[jospalau]

Originally Posted by Aleron Ives

You can't update because 4.34 is the latest version. Kobo pulled 4.35 for unknown reasons.

Wtf. No explanation? Like BOOX made recently with their last firmware? Though I think they made a big mess 😂

Sent from my CPH2021 using Tapatalk

[Uncle Robin]

I had a book in my wishlist that was in both English and devanagari. When I checked the preview, it looked as in the light image. So today, I bought it to test the devanagari font in 4.35. It looks as shown in the photo my Sage. What is perhaps evern weirder is that when I open the book in the Web reader using Read Now, it looks exactly the same as it does on my Kobo, and not as in the preview.

I checked my device and the version numer is still 4.35, but the release nots make no mention at all of fonts, as they originally did when the firmware was first released. Something is VERY screwy!

[Uncle Robin]

In yet another ad for de-DRMing, this is what the de-DRMd copy looks like in the Calibre viewer

[jackie_w]

@Uncle Robin,

As I understand it the way it works is, if you have 4.35.20400 installed and wi-fi is on and you open a kepub which the device detects as containing Bengali or Hindi text then it triggers a download of 2 font files into the /fonts/kobo/ subdirectory (NotoSansBengali-Regular.ttf and NotoSansBengali-Bold.ttf) and then auto-checks 'Noto Sans Bengali' as one of the fallback fonts for this kepub, in the [Aa] menu.

Did those 2 font files get auto-downloaded to your Kobo when you opened your kepub?

I don't know the rules for how the device detects the presence of Bengali/Hindi. It may be HTML markup such as xml:lang="hi" or xml:lang="bn" or it may just be the presence of certain unicode characters.

[Uncle Robin]

Originally Posted by jackie_w

@Uncle Robin,

As I understand it the way it works is, if you have 4.35.20400 installed and wi-fi is on and you open a kepub which the device detects as containing Bengali or Hindi text then it triggers a download of 2 font files into the /fonts/kobo/ subdirectory (NotoSansBengali-Regular.ttf and NotoSansBengali-Bold.ttf) and then auto-checks 'Noto Sans Bengali' as one of the fallback fonts for this kepub, in the [Aa] menu.

Did those 2 font files get auto-downloaded to your Kobo when you opened your kepub?

I don't know the rules for how the device detects the presence of Bengali/Hindi. It may be HTML markup such as xml:lang="hi" or xml:lang="bn" or it may just be the presence of certain unicode characters.

No fonts were downloaded, as far as I can tell. Bengali as a fallback font for devanagari would be as much use as Chinese. And it's interesting that the Web Reader online looks exactly the same as the display on my Sage, despite the preview (presumably hosted on the same server) looking fine. And while everything I know about ebook HTML could fit in a gnat's navel, the fact that the de-DRM'd version displays correctly suggests the issue is not with the coding of the book itself

[jackie_w]

Originally Posted by Uncle Robin

Bengali as a fallback font for devanagari would be as much use as Chinese.

The downloaded Kobo version of 'Noto Sans Bengali' lists both Devanagari and Bengali as supported unicode ranges. But as I can't read either of them I can't say how good/bad it is.

[Uncle Robin]

Originally Posted by jackie_w

The downloaded Kobo version of 'Noto Sans Bengali' lists both Devanagari and Bengali as supported unicode ranges. But as I can't read either of them I can't say how good/bad it is.

I alreadsy had Noto Sans Devanagari on my Sage, sideloaded some time ago. No Noto Sans Bengali on my Sage.

[ghira]

NiLuJe: Thanks. I have no patches installed. The only thing I've done other than use standard settings is to enable "optimizeLegibility"

[lensmann]

Originally Posted by Uncle Robin

I had a book in my wishlist that was in both English and devanagari. When I checked the preview, it looked as in the light image. So today, I bought it to test the devanagari font in 4.35. It looks as shown in the photo my Sage. What is perhaps evern weirder is that when I open the book in the Web reader using Read Now, it looks exactly the same as it does on my Kobo, and not as in the preview.

I tested this out on my Kobo Aura One, which is also on 4.35. I tried it on Pocket rather than by buying a Hindi book from Kobo - buying a book I can't read would be going farther than I could justify even in the interests of tinkering, and if it works in Pocket it should also work on the main reader.

Anyway, everything seems to work as it should. When I synchronized Pocket, it said it would need to download a font for missing glyphs, and it downloaded Noto Serif Devanagari. The article I'd saved then displayed text in Hindi, which seemed to my inexpert eyes be the same as in the original. See screenshot below, and the original page (it was the top news story on BBC Hindi at the time).

[EDIT]
OK, a bit of searching revealed a whole bunch of free books in Hindi on the Kobo store. I got the one with the most intriguing cover, which turned out to be a book of short poems / couplets (I think). It displays fine on the Kobo Web reader (see attached image).

On my Aura One, however, things are a bit weird. When I use Noto Serif Devanagari as a supplemental font and 'Publisher Default' as the main font, then a few glyphs aren't displayed. The attachment "devanagari-test-kao-1.jpg" is quite representative of what a typical page looks like. However, when I set Noto Serif Devanagari to be the main font, then everything seems to display correctly (see "devanagari-test-kao-2.jpg").

From what I can tell (but do remember that this isn't actually a language I can read), the issue seems to relate to combining marks. Comparing the two versions, the first missing glyph seems to be U+093C DEVANAGARI SIGN NUKTA, while the second seems to be U+0941 DEVANAGARI VOWEL SIGN U. I don't know if every instance of these signs fails to render properly, or if there's something about this particular combination that is causing the problem. But, in any event, everything seems to display fine if I set Noto Serif Devanagari as the main font, and I'm definitely not seeing the issues you saw with your book where Devanagari text displays as English letters.

[jackie_w]

Originally Posted by lensmann

I tested this out on my Kobo Aura One, which is also on 4.35. I tried it on Pocket rather than by buying a Hindi book from Kobo - buying a book I can't read would be going farther than I could justify even in the interests of tinkering, and if it works in Pocket it should also work on the main reader.

Anyway, everything seems to work as it should. When I synchronized Pocket, it said it would need to download a font for missing glyphs, and it downloaded Noto Serif Devanagari. The article I'd saved then displayed text in Hindi, which seemed to my inexpert eyes be the same as in the original. See screenshot below, and the original page (it was the top news story on BBC Hindi at the time).

Thanks for posting this example. When I copied a small sample of this Hindi text into my test kepub it also triggered the downloaded Kobo versions of NotoSerifDevanagari-Regular.ttf and NotoSerifDevanagari-Bold.ttf. The fact that my previous sample text of what I *thought* was Hindi did not, confirms my suspicions that I really ought to leave trying to answer Kobo language support questions to others better qualified.

[NiLuJe]

@ghira: No patches either over here (as I barely ever use Nickel anymore) (and 4.35, of course ;p). I can't tell you how it fared on 4.34, as it's been years since I used collections anywhere outside of calibre, but 4.35 *did* affect collision handling, as it seems to have fixed another collision issue introduced a few months ago.

[JSWolf]

Originally Posted by NiLuJe

@ghira: No patches either over here (as I barely ever use Nickel anymore) (and 4.35, of course ;p). I can't tell you how it fared on 4.34, as it's been years since I used collections anywhere outside of calibre, but 4.35 *did* affect collision handling, as it seems to have fixed another collision issue introduced a few months ago.

So would you say that on a Libra 2 running KOReader that KOReader would be better on 4.35?

[NiLuJe]

I have no idea how a Libra 2 would even enter the equation? I'm talking about the sunxi display driver, so that effectively means the Elipsa & Sage only. And, specifically, the Sage, as the Elipsa never got the updated collision handling code found on the Sage (it's still terrible/broken, but differently ;p).

But, as far as KOReader is concerned, I'm making far more liberal usage of the flag that bypasses most of said brokenness than Nickel is.

[Uncle Robin]

@jackie_w and @lendsman Thank you both for your experiments. If you copied the text from the preview, that would explain a lot, especially about why the preview renderes just fine but Kobo's own Web Reader does not. If the preview is using a Noto devanagri font, that's Unicode (and one I have on my reader too). The embedded fonts are not Unicode - in my quest to them I actually found an online tool to convert from the fonts used to a Unicode font.

The issue seems to be with the embedded font not being called by the kepub for whatever reason, and perhaps that's why once I had removed the kepub from my device I couldn't redownload it (it was only removed from the device, not put in the archive).

THe point of the experiment (apart from wanting to read the book ofc) was to test the advertised improved handling of Hindi in 4.35 Kobo's Hindi has always been problematic in my limited experience, one of the very first books I ever bough from Kobo was a Hindi translation of the Jungle Book and it too was a mess - Hindi characters but messed up and missing glyphs and still effectively unreadable. So I was hoping for better things from 4.35. If my experience is in any way indicative, it might explain part of the reason for pulling the firmware