Actual result:

“<span></span>” → “”
“<span id="…"></span>” → “<span id="…"></span>”
“<span class="…"></span>” → “”
“<span style="…"></span>” → “”
“<span title="…"></span>” → “”

Expected result:

“<span></span>” → “<span></span>” or “”
“<span id="…"></span>” → “<span id="…"></span>”
“<span class="…"></span>” → “<span class="…"></span>”
“<span style="…"></span>” → “<span style="…"></span>”
“<span title="…"></span>” → “<span title="…"></span>”

MediaWiki strips empty SPAN attributes, but not empty DIV. See:

http://www.mediawiki.org/w/index.php?oldid=588727&action=edit&preview=yes

a.d.bergi wrote:

(In reply to comment #2)

MediaWiki strips empty SPAN attributes, but not empty DIV.

Sure, but this bug is about empty spans.
The mentioned http://en.wikipedia.org/wiki/COinS metadata format even /specifies/ the use of <span> tags...

Btw, <div style="display:inline;"> is a horrible workaround for <span>.

<div class="…" style="display:inline;" />

is not a workaround for

<span class="…" />

because

A<div class="…" style="display:inline;" />B

generates

A
B

Okay okay okay. Before anyone else comes to put any other example I just want to clarify that my comment was to:

A) Notice the inconsistency between stripping empty SPAN vs DIV. I would expect none or both to be stripped.

B) DIVs can be used WHERE APPROPIATE as a replacement (for example, to insert an empty element with an ID or CLASS for presentational changes or script manipulation).

And about the last example,

<div>A<div class="…" style="display:inline;" />B</div>

generates

A B

Notice it's all surrounded by an additional DIV.</rant>

(In reply to comment #5)

<div>A<div class="…" style="display:inline;" />B</div>

generates

A B

Yes, but expected result is

AB

without a whitespace.

Copy from T134423#2266304:

Re the stripping - Tidy actualy does not strip every self-closing tag and OTOH it strips some empty pair tags.

<span/> - stripped
<span class="foo" /> - stripped
<span></span> - stripped
<span class="bar"></span> - stripped
<div/> - stripped
<div class="foo" /> - NOT stripped
<div></div> - stripped
<div class="bar"></div> - NOT stripped

So, Remex is no longer stripping in the way Tidy stripped before (notably, empty elements div and span #3, are not removed). That's Good.

On an aside, I was under the impression that Remex would actually strip the self-closed tags which cannot be self-closed under HTML (that's both div and span), which is why we have the Linter category. The output I'm getting instead is:

<span></span>
<span class="foo"></span>
<span></span>
<span class="bar"></span>
<div></div>
<div class="foo"></div>
<div></div>
<div class="bar"></div>

Spans and divs 1 and 2 (versus 3 and 4 for each) are clearly not being stripped and I'm pretty sure they are supposed to be (either that or passed through as self-closing elements). @ssastry do you know why? 😃

In T29786#4507268, @Izno wrote:

So, Remex is no longer stripping in the way Tidy stripped before (notably, empty elements div and span #3, are not removed). That's Good.

On an aside, I was under the impression that Remex would actually strip the self-closed tags which cannot be self-closed under HTML (that's both div and span), which is why we have the Linter category. The output I'm getting instead is:

<span></span>
<span class="foo"></span>
<span></span>
<span class="bar"></span>
<div></div>
<div class="foo"></div>
<div></div>
<div class="bar"></div>

Spans and divs 1 and 2 (versus 3 and 4 for each) are clearly not being stripped and I'm pretty sure they are supposed to be (either that or passed through as self-closing elements). @ssastry do you know why? 😃

We decided to provide a backward compatibility fix for self-closing tags to reduce some of the workload of editors. But, the plan is indeed to remove that fix once the use of self-closed tags has come down as editors fix the linter issues for that category. The b/c fix is to convert a <tag /> to <tag></tag>.