Troubleshooting Send to Kindle issues and findings

[tomsem]

I know there's an earlier thread but wanted to start fresh. First a few of the 'findings':

PDFs with more than 100 bookmarks will not convert to Print Replica
Update: This seems to be fixed now. Kindle Create still has issue with it.

ePubs with more than about 1300 ToC entries fail to convert
Fun fact: most editions of the Bible have 66 books and 1,189 chapters, so a ToC with entries for all of these would probably get through S2K's pearly gates. Maybe that was even the standard used for the arbitrary limit?
In this case Kindle Previewer 3 does not complain, but it gets rejected by S2K. The only recourse seems to be to thin out the ToC entries somehow.
It's also possible that it is triggering some sort of timeout and is not a ToC size limit as such. It takes about 40 minutes of processing before failure occurs. Calibre can take hours to convert these (hmm why?).

Conversion to Print Replica will preserve Bookmark hierarchy in its ToC
Bookmarks are converted to Print Replica ToC entries, and at least one level of hierarchy is preserved (have not tested more than that). Generated PDFs often lack any bookmarks and thus will have a minimal ToC by default. So it is nice to add some before sending it for conversion. I've seen at least on GitHub project that purports to automate this (scans PDF for things that look like headings), and am hoping eventually to script it myself (I plan to add page links for page references in Contents/Index/Footnotes too).
Kindle Create does not seem to let you create a multi level ToC, and if you make changes to one that's been imported, it gets flattened.

FWIW I sent Amazon Feedback about the Print Replica ToC limit (seems way too small). Probably should send one about multi-level issue too.

[tomsem]

(deleted)

[kyrilson]

If you convert to epub and then run through this (https://kindle-epub-fix.netlify.app), does it work? Just curious. I’ve been using that to fix any errors before I use the S2K function. Everything has worked perfectly so far (started using it back in Dec 2022 or so)


Sent from my iPhone using Tapatalk

[tomsem]

Originally Posted by kyrilson

If you convert to epub and then run through this (https://kindle-epub-fix.netlify.app), does it work? Just curious. I’ve been using that to fix any errors before I use the S2K function. Everything has worked perfectly so far (started using it back in Dec 2022 or so)


Sent from my iPhone using Tapatalk

It did change something for the files referencing the embedded fonts, but Kindle Previewer 3 says 'It may be incomplete or corrupt' without any further details. It no longer passes epubcheck ('Corrupted image file encountered').

So, no.

But it turns out that this was one of those First Reads purchases from Amazon. Yet it showed up in Docs in Content and Devices!

I deleted it from there, but it's still listed in Books, as expected. Did I send it with Send to Kindle for some reason? Don't know, but we can reclassify this issue:

Never Mind

[tomsem]

That said, I'm now getting the same error with another book. Originally ePub (one of Tor's free book of the month). No longer have the source files, and apparently didn't have backup in calibre (this goes back 6 years, before I got serious). I downloaded it to my Kindle, imported it to calibre, converted to ePub3. epubcheck passes, no errors or warnings. But Kindle Previewer 3 throws the 'Hyperlink not resolved in toc' error, and S2K rejects it (probably the same code).

I tried one of the online converters and the result didn't pass epubcheck. Also tried the ^^tool and it made changes, but it broke epubcheck.

So it again comes down to resolving:

E24010: Hyperlink not resolved in toc (One possible reason can be that the link points to a tag with style display:none):/private/var/folders/1c/s6n482rn5fs_ls6v8cdnqr5r0000gr/T/0bf4418c-75c0-48fe-a8bf-342aa3c493ac/cTemp/mTemp/mobi-76dTK3/text/part0001.html#UGI0-51bae2eb2418443c8e37a3f2d3286faf

It seems Kindle Create is tripping over toc references that include ID in the url after hash (#).

/text/part0001.html#UGI0-51bae2eb2418443c8e37a3f2d3286fa

I removed the hash and ID (there were two TOC entries in this case), and now Kindle Create and S2K accept it.

All's well that ends well, but it kinda looks like bug in the Kindle Create (and presumably Send to Kindle) validation code.

Can someone explain why a publisher would be using this URL form? Visually it doesn't seem different, or I'm not noticing anything.

Okay, I found the original .ePub and .mobi files from Tor. It seems the hash style is used in the .mobi, but not in the .epub. When I convert .mobi/.azw3 to .epub, it carries over this hash style that Kindle Create doesn't like. But if I convert the original .ePub to epub3 there are no hash style TOC URLs.

I wonder if calibre's conversion to ePub should dispense with carrying over the hash style tags or at least offer option to just leave the hash part off.

In any case, I'll start making use of the original epub (having found them all) rather than downloading to kindle and using that for source for adding to calibre.

[slm]

When I've had problems, I've used Calibre to convert to DOCX and then sent the DOCX to Kindle using Send to Kindle.

So far, that has worked like a charm, probably because the DOCX converter is less fussy.

[tomsem]

Originally Posted by slm

When I've had problems, I've used Calibre to convert to DOCX and then sent the DOCX to Kindle using Send to Kindle.

So far, that has worked like a charm, probably because the DOCX converter is less fussy.

Yes, everyone seems to have some trick, and thanks for yours.

Now that I understand this particular problem, however, ePub remains my go-to, unless the source is DOCX. Fewer conversions seems better.

[Quoth]

Delete

[tomsem]

It appears to be possible to create Print Replica locally by feeding the KPF output of Kindle Create to the KFX Output plugin CLI.

These can be side loaded to Scribe and the iPad app (where you can add highlights and notes). I have not checked yet to see if Share works (Scribe) or Export (iPad). Most likely not, but there's always screen capture.

Update: There is something different about KFX generated from KFX Output than that generated from Send To Kindle. The conversion from KPF to KFX is being handled by Kindle Previewer, so maybe that is missing something.

Side load KFX to the iPad app of the former does not let you select text etc, but side loading the Send To Kindle KFX (sent to Scribe) will (however there is no Pages preview as there is for Print Replica store content).

On the Scribe itself, there's very little difference that I have noticed so far. So it might be good enough to experiment with.

And Share/Export goes through the motions of things, but nothing gets exported (as one might expect).

From Kindle iOS app it does send email, but the .html file is empty. It looks like it is once again sending PDFs from Send To Kindle service, there was some sync of reading position with Scribe a couple of weeks ago, but that seems to be gone.

[tomsem]

Originally Posted by tomsem

From Kindle iOS app it does send email, but the .html file is empty.

Actually I looked at the html attachments again, it does export annotations for side loaded Print Replica (and I assume for any side loaded books). It's all generated on the device and sent out by the email client on the device.

[tomsem]

(deleted)

[tomsem]

Send To Kindle failed to ingest an ePub which I'd converted from AZW3.

Solution was to edit the AZW3 and fix font name references which were incorrect (name didn't match what was in font file), and also let it update some GUIDs which the checker said were not valid GUIDs, and generate a new ePub.

[JSWolf]

Originally Posted by tomsem

Send To Kindle failed to ingest an ePub which I'd converted from AZW3.

Solution was to edit the AZW3 and fix font name references which were incorrect (name didn't match what was in font file), and also let it update some GUIDs which the checker said were not valid GUIDs, and generate a new ePub.

When you convert from KF8 be it with Calibre's conversion or KindleUnpack, you will have errors. This is why it's good to use both Calibre's error checker and epubcheck to find them and fix them.

[tomsem]

Originally Posted by JSWolf

When you convert from KF8 be it with Calibre's conversion or KindleUnpack, you will have errors. This is why it's good to use both Calibre's error checker and epubcheck to find them and fix them.

Most of my books are in AZW3 and I don't yet have many of these converted to ePub, and do 'just in time' conversion when I want to read with an ePub app or device.

In this case, I had accidentally deleted the book from my library and wasn't sure how easy it would be to get Amazon to restore it, hence my first reaction was to get it it added back as personal document and that's when I ran into the issues.

As it turns out, Amazon restored it via Chat support in under a minute, my best experience with CS in a long time.

Long term, I plan to get everything in ePub, look for and fix any issues as I go along, but that will take awhile.