Empty tags get converted into self-closing tags

[hannaeckert]

Expected Behavior:

In Antisamy up to and including version 1.7.6, I got this behavior:
Input: <div></div>
Output: <div></div>

Current Behavior

Since Antisamy version 1.7.7, I now get this behavior:
Input: <div></div>
Output: <div/>

---

This applies to any empty tag within the input, even if nested deep within other tags.
I would like for Antisamy to not change the structure of empty tags when processing the input.

---

Until this is fixed, my workaround is to stick to version 1.7.6.

Please let me know if you need further information or details on this.

[davewichers]

@spassarop - Can you research this? @rbri - Do you think this change was caused by a change in neko-htmlunit? If so, was it deliberate?

[rbri]

Will have a look at the end of the week, sorry

[rbri]

@hannaeckert you a right, this is a change introduced in version 1.8.8.

Spend some time to check and document the behaviour of empty tag handling across the version 1.7.7, 1.7.8 and the head. And also implement some minor changes to emit more correct html.
Looks like this improvement to take care of html empty tags (like br) and emiting them in there html5 form already solves some backward compatibility problem.
Finaly i only had to introduce some special handling for div and textarea.

@davewichers @spassarop please have a look

[hannaeckert]

@rbri Thanks for getting back to me so quickly 😊

I'll run a check if this fixes my issues and get back to you as soon as possible.

[hannaeckert]

@rbri

I just built a snapshot version with:

$ git clone https://github.com/nahsra/antisamy
$ cd antisamy
$ mvn -Dmaven.test.skip package

(I had to skip the tests because one [as far as I could tell unrelated] test failed during the build)

When I feed in this input:

<div></div>
<span></span>
<i class="some-test-class"></i>
<a class="fun-test-class"></a>
<table id="having_fun_with_empty_tags"></table>
<table>
    <thead></thead>
    <tbody></tbody>
</table>
<table>
    <tbody>
        <tr>
            <td></td>
        </tr>
    </tbody>
</table>
<textarea></textarea>

I got this output:

<div/>
<span/>
<i class="some-test-class" />
<a class="fun-test-class" rel="nofollow"></a>
<table id="having_fun_with_empty_tags" />
<table>
    <thead/>
    <tbody/>
</table>
<table>
    <tbody>
        <tr>
            <td></td>
        </tr>
    </tbody>
</table>
<textarea/>

---

When I do the same thing with Version 1.7.6, I get this:

Input:

Input:
<div></div>
<span></span>
<i class="some-test-class"></i>
<a class="fun-test-class"></a>
<table id="having_fun_with_empty_tags"></table>
<table>
    <thead></thead>
    <tbody></tbody>
</table>
<table>
    <tbody>
        <tr>
            <td></td>
        </tr>
    </tbody>
</table>
<textarea></textarea>

Output:

<div></div>
<span></span>
<i class="some-test-class"></i>
<a class="fun-test-class" rel="nofollow"></a>
<table id="having_fun_with_empty_tags"></table>
<table>
    <thead></thead>
    <tbody></tbody>
</table>
<table>
    <tbody>
        <tr>
            <td></td>
        </tr>
    </tbody>
</table>
<textarea></textarea>

---

The main Problem with this new behavior is, that <div>, <span>, <i>, <table>... etc. are no valid void- / self-closing-tags in html5. So when a browser sees something like <i class="fa fa-trash" /> it gets converted into <i class="fa fa-trash">, which leads to changes in the dom-structure. If you have this <i class="fa fa-trash"></i> this issues does not exist.

---

I hope this info is helpful 😊

[rbri]

Hi @hannaeckert,

i think i got your point, and hopefully this is fixed with the pr i made (will check this later)

$ git clone https://github.com/nahsra/antisamy

if you like to have a look at my fixes, you have to locally merge my pr before you build a new version with maven.
Maybe you can have a look at the new test, to get an idea about the changed output after the fix.

https://github.com/nahsra/antisamy/pull/680/changes#diff-6770fca9b2d2a28c1ccecfa79e0401d1c4a9dc2ee5763025e3b6c75135b0860fR2994

[hannaeckert]

Hi @rbri,

I just tested your latest version by building from your branch (I checked it out with the github client: $ gh pr checkout 680):

When I rerun my test, this is what I'm getting:
Input:

<p></p>
<div></div>
<span></span>
<i class="some-test-class"></i>
<a class="fun-test-class" rel="nofollow"></a>
<table id="having_fun_with_empty_tags"></table>
<table>
    <thead></thead>
    <tbody></tbody>
</table>
<table>
    <tbody>
        <tr>
            <td></td>
        </tr>
    </tbody>
</table>
<textarea></textarea>

Output:

<p/>
<div></div>
<span/>
<i class="some-test-class" />
<a class="fun-test-class" rel="nofollow"></a>
<table id="having_fun_with_empty_tags" />
<table>
    <thead/>
    <tbody/>
</table>
<table>
    <tbody>
        <tr>
            <td></td>
        </tr>
    </tbody>
</table>
<textarea></textarea>

It looks way nicer, but <p>, <span>, <i> and <table> is still converting to invalid self-closing tags.

[rbri]

@hannaeckert many thanks for the fast feedback, will check again

[rbri]

Hi @hannaeckert,

sorry for the back and forth here, but antisamy is not my project, i only like to help (and make sure neko-html (one of my projects)) is not the root :-)

Spend some time on this again and i think i have a better understanding of this now.

My current understanding:

• AntiSamy spits out (hopefully) valid XHTML
• this implies that self closing HTML tags (e.g. br, hr) should be written as self closing XHTML tags like <br/>. For backward compatibility with older antisamy version we should write this without a blank before the slash (even if XHTML suggests the blank)
• all other tags should be written with a closing tag

I hope these assumptions are now correct.
The new PR #683 is based on this assumptions and implements and tests this approach for all HTML tags

[hannaeckert]

Hi @rbri,

don't worry, I'm happy to help.
Yes, as far as I can tell you understood the problem perfectly 😊👍

---

I tested your most recent version (pr 683) and now it works nicely:

Input:

<p></p>
<div></div>
<span></span>
<i class="some-test-class"></i>
<a class="fun-test-class"></a>
<table id="having_fun_with_empty_tags"></table>
<table>
    <thead></thead>
    <tbody></tbody>
</table>
<table>
    <tbody>
        <tr>
            <td></td>
        </tr>
    </tbody>
</table>
<textarea></textarea>

Output:

<p></p>
<div></div>
<span></span>
<i class="some-test-class"></i>
<a class="fun-test-class" rel="nofollow"></a>
<table id="having_fun_with_empty_tags"></table>
<table>
    <thead></thead>
    <tbody></tbody>
</table>
<table>
    <tbody>
        <tr>
            <td></td>
        </tr>
    </tbody>
</table>
<textarea></textarea>

---

Thank you for your work 😊

[rbri]

@hannaeckert thanks for the fast feedback, lets see what the others are thinking

[rbri]

@davewichers @spassarop pr is ready for review - any chance to have a look at it; and btw there is a second pr that fixes another problem.

[spassarop]

Thanks. I’m aware of the PRs but I’ve been busy and sick so I delayed the review. I should be able to review this weekend.

…

On Fri, 22 May 2026 at 14:45 RBRi ***@***.***> wrote: *rbri* left a comment (nahsra/antisamy#679) <#679 (comment)> @davewichers <https://github.com/davewichers> @spassarop <https://github.com/spassarop> pr is ready for review - any chance to have a look at it; and btw there is a second pr that fixes another problem. — Reply to this email directly, view it on GitHub <#679 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHL3BMJR5DT2K5MGSWS22RD44CG35AVCNFSM6AAAAACYI3ZIBCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DKMRRGMZTQMBZGY> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>