Bug 1095590

Created attachment 894761 [details]
zip of maven project to build docbook and html from asciidoc sources

Seems to be treating the alt tag as 3 blocks of 5 chars each, and truncating the rest:

[Web Page Open on Simulated Device.] ==> Web P,age O,pen o

-----

If I create a sample .adoc file like this:

<<test.zip!/test/src/main/asciidoc/test.adoc>>

And I render it using this pom.xml and the maven asciidoc plugin

<<test.zip!/test/pom.xml>>

Then the resulting docbook looks like this:

<<test.zip!/test/target/generated-docs/test.xml>>

In that case, the docbook XML looks fine, as does the resulting HTML.

So, clearly there's a problem w/ the way asciidoctor is being used here, as the maven wrapper for asciidoctor works fine on your sample .adoc.

Created attachment 894763 [details]
zip of maven project to build docbook and html from asciidoc sources - with commas in the alt tag

This version of the project (test-commas.zip), with commas in the alt tag, results in broken docbook (but does NOT truncate the content):

.My great figure
image::images/4116.png[This description, containing commas, is an alt tag.]

becomes:

<figure>
<title>My great figure</title>
<mediaobject>
<imageobject>
<imagedata fileref="./images/4116.png" contentwidth="containing commas" contentdepth="is an alt tag."/>
</imageobject>
<textobject><phrase>This description</phrase></textobject>
</mediaobject>
</figure>

Created attachment 894765 [details]
zip of maven project to build docbook and html from asciidoc sources - with commas in the alt tag, FIXED by adding quotes in the tag

If we add quotes into the alt tag, the resulting docbook is once again clean:

.My great figure
image::images/4116.png["This description, containing commas, is an alt tag."]

becomes

<figure>
<title>My great figure</title>
<mediaobject>
<imageobject>
<imagedata fileref="./images/4116.png"/>
</imageobject>
<textobject><phrase>This description, containing commas, is an alt tag.</phrase></textobject>
</mediaobject>
</figure>

I have confirmed that the Ascidoctor application (i.e. the Ruby one) will convert the asciidoc from https://bugzilla.redhat.com/show_bug.cgi?id=1095590#c1 correctly, so this must be an issue with Asciidoctor.js.

Fixed in Build 1.6-SNAPSHOT 201405190849

This was caused by changing the $ identifier in Asciidoctor.js to opal$ so it would play well with angularjs. I missed two cases where this find and replace modified a regex where the $ was used as a string end marker.

opal$opal.cdecl(opal$scope, 'BoundaryRxs', opal$hash2(["\"", "'", ","], {"\"": /.*?[^\\](?=")/, "'": /.*?[^\\](?=')/, ",": /.*?(?=[ \t]*(,|opal$))/}));

opal$opal.cdecl(opal$scope, 'SkipRxs', opal$hash2(["blank", ","], {"blank": opal$scope.BlankRx, ",": /[ \t]*(,|opal$)/}));

Fixing these fixed the image importing.

The supplied test case now imports as 

<section>
	<title>Test Page</title>
	<figure>
		<title>My great figure</title>
		<mediaobject>
			<imageobject>
				<imagedata fileref="images/4116.png" />
			</imageobject>
			<textobject> 
				<phrase>Web Page Open on Simulated Device.</phrase>
			</textobject>
		</mediaobject>
	</figure>
</section>

Verified that alt text is correctly populated now assuming the correct syntax is used. In the example Nick gave in Comment #2 the content was still imported with the incorrect contentdepth and contentwidth values, however this happened with the maven build as well, so marking this as VERIFIED.