Wrong output html if table's caption contains headers

[azhidkov]

Hello, on this example HTMLPurifier 4.9.2 returns bad output html:

$text = '<table border="1" style="width:300px">
    <caption>
        <h2><strong>caption</strong></h2>
    </caption>
    <tbody>
        <tr>
            <td style="width:64px;text-align:right"><strong>Life:</strong></td>
            <td style="width:80px"><strong>063-560-10-10</strong></td>
        </tr>
        <tr>
            <td style="width:64px;text-align:right"><strong>Skype:</strong></td>
            <td style="width:80px"><strong>saleusa</strong></td>
        </tr>
        <tr>
            <td style="width:64px;text-align:right"><strong>Viber:</strong></td>
            <td style="width:80px"><strong>0973151111</strong></td>
        </tr>
        <tr>
            <td style="width:64px;text-align:right"><strong>e-mail:</strong></td>
            <td style="width:80px"><strong>info@saleusa.com</strong></td>
        </tr>
    </tbody>
</table>
';

$config = HTMLPurifier_Config::createDefault();
$p = new HTMLPurifier($config);
$html = $p->purify($text);

Output html is:

<h2><strong>caption</strong></h2>
    
    <strong>Life:</strong>
            <strong>063-560-10-10</strong>
        <strong>Skype:</strong>
            <strong>saleusa</strong>
        <strong>Viber:</strong>
            <strong>0973151111</strong>
        <strong>e-mail:</strong>
            <strong>info@saleusa.com</strong>

But if I remove h2 tag (change caption to <strong>caption</strong>) everything OK.
Is this a bug or may be I just not turn on some settings?

[ezyang]

In HTML 4.01, the contents of caption must be inline elements; h2 is a block element so it can't be allowed.

As for why we lose all of the table structure, I haven't investigated closely, but I suspect what may be happening is that h2 is triggering us to insert closing tags for the table. The subsequent open/close table tags are then dropped as not valid and then you lose the formatting.

There might be an easy hack to work around this particular case but I am not sure what a general fix might be. Perhaps try setting Core.LexerImpl to PH5P?