EPUB2 to EPUB 3 errors

[Psymon]

I recently reinstalled Sigil, and discovered in the process that there's now that EPUB2 to EPUB3 converter plugin -- woohoo! I had so much trouble before trying to convert one of my ebooks to EPUB3 (simply so that I could add in one little audio file) that I finally gave up on it. :/

That plugin is great, though! I still ended up with a ton of errors, but I just finished going through them all one-by-one and correcting them all. Now I've finally got my ebook validating with NO errors -- although this is before I add in that little audio file, mind you -- but I still have a 289 "warnings," virtually all of which seem to be this one:

SVG hyperlinks should have a human-readable title (using the 'title' child element or the 'xlink:title' attribute)

I have absolutely NO idea what this means -- even after searching all over the 'net for an answer. It would seem that all the errors arise wherever I have plain old HTML links (i.e. <a href="whatever">), but I'm clueless how to fix this.

Or should I just simply ignore these "warnings"?

So that's my first question. My second question arises from when I add in the audio file, which is basically just this line of code...

<audio controls="controls" src="../Audio/tombowling.mp3" style="text-align:center; width:100%;"><i>[Sorry, your device doesn’t support audio]</i></audio>

...but just that then gives me these six errors when I validate...

ERROR (RSC-005) at "Walden_and_Beyond_epub3.epub/OEBPS/content.opf" (line 15, col 32):
Error while parsing file 'attribute "opf:role" not allowed here; expected attribute "dir", "id" or "xml:lang"'.

ERROR (RSC-005) at "Walden_and_Beyond_epub3.epub/OEBPS/content.opf" (line 27, col 34):
Error while parsing file 'attribute "opf:scheme" not allowed here; expected attribute "id"'.

ERROR (RSC-005) at "Walden_and_Beyond_epub3.epub/OEBPS/content.opf" (line 28, col 34):
Error while parsing file 'attribute "opf:scheme" not allowed here; expected attribute "id"'.

ERROR (RSC-005) at "Walden_and_Beyond_epub3.epub/OEBPS/content.opf" (line 10, col 47):
Error while parsing file 'assertion failed: @refines missing target id: 't1''.

ERROR (RSC-005) at "Walden_and_Beyond_epub3.epub/OEBPS/content.opf" (line 11, col 69):
Error while parsing file 'assertion failed: @refines missing target id: 'create1''.

ERROR (RSC-005) at "Walden_and_Beyond_epub3.epub/OEBPS/content.opf" (line 12, col 49):
Error while parsing file 'assertion failed: @refines missing target id: 'create1''.

I haven't the slightest idea what any of that means -- at least, I know what the content.opf file is, but what is it that I'm supposed to do to fix this?

Thanks so much in advance (and as always -- I'm so grateful for these forums!). It would be so cool if I can finally get this converted!

[Toxaris]

Did the ePUB2 validates correctly with Flightcrew and ePUBCheck before converting to ePUB3? If so, please report these errors also in the ePUB3 plugin thread. I am wondering this, since adding an audio file should not give those errors in the opf at all.

An image must have the property for an alternative title in it. So, for example, <img alt="description of image" src="<image>" />. To be honest, most people use alt=""...

I do think we need to see the OPF to give a better answer.

[DiapDealer]

It sounds to me as if the opf metadata section has somehow "reverted" to epub2. Did you edit the epub3 with any other programs after using the Epub3itizer plugin?

[Psymon]

My original EPUB2 version validated perfectly -- it's already up on the iBooks store, in fact. I just opened that, ran the EPUB3 converter plugin, ran it through Pagina where I got a ZILLION errors (give or take a few million) fixed all the errors -- but still had those hundreds of "warnings" -- and then added in that one line of code for the audio file.

Oh, and I've been doing all this EPUB3 editing in Sigil, too -- would that be the issue here? I realize that that is (or at least was) meant for EPUB2 editing.

[Psymon]

Oops, sorry, I neglected to respond better to this...

Originally Posted by Toxaris

An image must have the property for an alternative title in it. So, for example, <img alt="description of image" src="<image>" />. To be honest, most people use alt=""...

I do think we need to see the OPF to give a better answer.

I'm not getting any errors/warnings re my image tags (which all have alt tags). In fact, before adding in that line of code for the audio file (and the audio file itself) I had no errors at all, just a couple hundred "warnings" all related to my links.

And re my OPF file (I presume you mean the content.opf file), here it is, attached...

[DiapDealer]

Originally Posted by Psymon

Oh, and I've been doing all this EPUB3 editing in Sigil, too -- would that be the issue here? I realize that that is (or at least was) meant for EPUB2 editing.

That all depends on what version of Sigil you're using. The latest version should be no problem. It has the inherent ability to edit EPUB3s. Older versions didn't (or could only do so on a very limited basis).

So yes, some of the work done by the Epub3-itizer plugin could conceivably be undone (and those exact errors you're getting introduced as a result) if you're not using the latest version of Sigil when editing the subsequent Epub3-itizer plugin's EPUB3 output.

Easy way to tell if that's what's going on is to run the Epub3-itizer created EPUB3 through EpubCheck immediately after it's produced (before ever opening the new EPUB3 in Sigil). It should pass validation (or very, very close to it depending on the plugin version). That way you'll know if editing the EPUB3 with Sigil is introducing the errors or not.

Either way, I recommend making sure you're using the very latest version of Sigil (v.0.9.3) and the Epub3-itizer plugin.

[KevinH]

Your content.opf file is mixing epub2 style dc metadata with epub3 style dc metadata with refines. So unless there is a bug in epub3-itizer that no one else has hit (always a possibility) something has converted the epub3 dc creator, dc title and dc identifier schemes back to epub2. Sigil version 0.9.3 will not do this since it literally prevents the metadata editor from being invoked under epub3. But earlier versions of Sigil will if you invoke the Metadata Editor gui. What else touched the file after epub3-itizer output it? Was it a Sigil version earlier than Sigil-0.9.3?

If need be I can walk you through fixing the metadata for epub3 by hand, removing all of the calibre introduced pieces and properly creating epub3 metadata.

Just let me know.

KevinH

[Psymon]

Hey, thanks guys. I'm using Sigl v. 0.9.3, and epub3-itizer v. 0.3.5 -- and yes, I got all those errors right after simply running the filter on my valid EPUB2 file and then running that through epubcheck. A lot of the errors (as opposed to warnings) had to do with width tags in images and tables, and valign tags in tables, which I switched out to the CSS equivalents -- I'm still stuck with all those warnings, though (289 of them), which I still don't understand the meaning of, and then I got more errors once again when I added in the audio file/code (as I mentioned earlier).

So I guess there is a bug, if it's supposed to come out "perfect." Either that or there's something wrong with my book that's crashing the script when it runs -- but it validates just fine (in EPUB2) so I have no idea what/where there that might be.

So yeah, I guess any help in getting things right would be much appreciated!

[KevinH]

Hi,
I will edit your content.opf and post it for you to use as a model. Will you try rerunning epub3-tizer on your valid epub2 version and saving it under a new name. Then rename the new epub output from epub3-itizer to end with .zip and unzip it and post just the unzipped content.opf file here. That way we can see if the problem is with Epub3-itizer or Sigil.

Thanks,
KevinH

Update: see below for a content_fixed.opf you can use to hand correct your version.

As for your warnings, I see under the manifest properties that svg is being used heavily in this epub. If you use svg and there are links in it you should add the xlink:title property as mentioned in the warning.

Also, how are you checking the epub2 version before running epub3-itizer? Are you using the FlightCrew plugin or the epubcheck plugin? Given the preponderance of svg in this file you should probably use the epubcheck plugin to do your epub 2 checking as well. Also remember that epubcheck warnings are NOT the same as errors in any way.

KevinH

[Psymon]

Originally Posted by KevinH

Hi,
I will edit your content.opf and post it for you to use as a model.

Oh, that's very kind of you, thank you.

Will you try rerunning epub3-tizer on your valid epub2 version and saving it under a new name. Then rename the new epub output from epub3-itizer to end with .zip and unzip it and post just the unzipped content.opf file here. That way we can see if the problem is with Epub3-itizer or Sigil.

I would -- I don't mind doing so (I'm giving the book away for free on the iBooks store, too) -- but it's too big to post here. Even before adding in the MP3 file it's about 24 megs in size. You can download a slightly-earlier version of it off my website, though -- I'm sure it'll come out with the same "issues," as most of the changes between that version and the most recent one are aesthetic/typographic...

http://www.psymon.com/walden/ebooks/

The book in question (of the four there) is the main anthology, Walden and Beyond -- and if only for reference, it's the very last chapter (the "Postscript," an article which I wrote myself) where I wanted to add in the audio file to play Henry's favourite song.

[Hope it's okay for me to post the link to my website like that -- I recall running into a problem here before that "advertising isn't allowed" in these threads, but that's not my intention here, I'm just sharing the link so you can get the book!]

[EDIT: Oops, I mis-read what you wanted -- sure, I can send along just the content.opf file, but having just sent you the link for the book itself, is that better, so you can run it on that EPUB2 version and actually see yourself all the errors I ran into?]

[KevinH]

Hi Psymon,
You misunderstand. I don't want the epub, just the content.opf file that is zipped up inside the output of the epub3 epub3-itizer generates. Epubs are simply special types of zip archives after running epub3-itizer.
Thanks,

KevinH

[Psymon]

Originally Posted by KevinH

Hi Psymon,
You misunderstand. I don't want the epub, just the content.opf file that is zipped up inside the output of the epub3 epub3-itizer generates. Epubs are simply special types of zip archives after running epub3-itizer.

Oh, as I said in my "edit" to my last post, I thought that the actual book itself would be better for you -- but here, attached, is the content.opf file, immediately after I ran my EPUB2 version through it (and before I fixed all those "width" and "align," etc. issues).

[KevinH]

Hi Psymon,

Okay it is correct as it is written, which means that ePub3-itizer is not at fault here. So something in Sigil-0.9.3 is messing up your epub3 when reloaded. I will try to figure out what that is. I just grabbed your book to see what the svg warnings were about.

SVG hyperlinks should have a human-readable title (using the 'title' child element or the 'xlink:title' attribute)

Your book has many sections like the following in its code:

  <div class="svg_outer">
    <div class="svg_inner">
      <svg height="94%" version="1.1" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1200 1800" preserveAspectRatio="xMidYMid meet" width="94%" xmlns:xlink="http://www.w3.org/1999/xlink"><image height="1800" id="title-original" width="1200" xlink:href="../Images/walden-titlepage.jpg"/></svg>
    </div>
  </div>

It is just a warning that says your need to add an xlink:title="Walden-TitlePage-Image" attribute (or something along those lines).

This has nothing to do with epub3-itizer operation. And they are just warnings and can be ignored.

Hope this helps,

KevinH

[KevinH]

The bug is in Sigil-0.9.3 when you add any existing file (such as an audio file) somethign causes the metadata to be rewritten. I will track down what and get it fixed for the next release.

The best workaround is to add the audio file to the epub2 version BEFORE running ePub3-Itizer.

Thanks,

KevinH

[Psymon]

Originally Posted by KevinH

The bug is in Sigil-0.9.3 when you add any existing file (such as an audio file) somethign causes the metadata to be rewritten. I will tack down what and get it fixed for the next release.

Interesting. Would it help if, say, I went back to my last valid EPUB2 version, then add in the audio file (and the code for that, of course), and even though that "invalidates" the EPUB2, then convert it from there to EPUB3, and see what happens?

Actually, if only out of curiosity, I'll try that -- certainly it would be nice if that does then "instantly" convert my book flawlessly.

[EDIT: What the heck is wrong with me today??? That IS what you said to do -- d'oh. That's the second time today. ]

Oh, and before doing that, I'll first fix all those other things that were giving me errors (re width, align, etc.), plus all those SVG warnings I got, of course. That's interesting, too, re your explanation of the latter -- I couldn't understand the reference to the "SVG link," since what they were pointing to were just plain HTML links, but I guess all those warning messages relate to links that point to an SVG image. That's where epubcheck should tell me the error is -- not in any/all links pointing to that.

Did I inadvertently find a bug in epubcheck, too?