Typo3 exporter errors caused by array_merge and htmlspecialchars

[fishbone1]

Typo3 version: 11.5.9
Extension version: master 20297bb

I sometimes get a warning and and error when opening the page exporter. It happens after I added a rich text content with text and image. But it doesn't happen always. The warning part:

PHP Warning: array_merge(): Expected parameter 2 to be an array, string given in (...)/myproject/public/typo3/sysext/impexp/Classes/Export.php line 700

This is the typo3 code. I set a debugger breakpoint at line 700:

    protected function flatSoftRefs(array $relations): array
    {
        $list = [];
        foreach ($relations as $field => $relation) {
            if (is_array($relation['softrefs']['keys'] ?? null)) {
                foreach ($relation['softrefs']['keys'] as $spKey => $elements) {
                    foreach ($elements as $subKey => $el) {
                        $lKey = $field . ':' . $spKey . ':' . $subKey;
// Line 700:
                        $list[$lKey] = array_merge(['field' => $field, 'spKey' => $spKey], $el);
                        // Add file_ID key to header - slightly "risky" way of doing this because if the calculation
                        // changes for the same value in $this->records[...] this will not work anymore!
                        if ($el['subst']['relFileName'] ?? false) {
                            $list[$lKey]['file_ID'] = md5(Environment::getPublicPath() . '/' . $el['subst']['relFileName']);
                        }
                    }
                }
            }

array_merge tries to merge with $el which contains:

<p>Before</p>
<p><img alt="" data-alt-override="true" data-htmlarea-file-table="sys_file" data-htmlarea-file-uid="{softref:c64604e5faca1242d8358828cfb4170b}" data-title-override="true" src="/fileadmin/user_upload/myimage.png" style="" title="" width="300" height="73" /></p>

I can also see this structure in a successful export:

<softrefs type="array">
    <keys type="array">
        <softref_key index="rtehtmlarea_images" type="array">
            <softref_element index="content">&lt;p&gt;Content mit Bild:&lt;/p&gt;
                &lt;p&gt;&lt;img alt=&quot;&quot; data-alt-override=&quot;true&quot; data-htmlarea-file-table=&quot;sys_file&quot; data-htmlarea-file-uid=&quot;{softref:80768071c2028e4faefc5a619f078928}&quot; data-title-override=&quot;true&quot; src=&quot;/fileadmin/user_upload/myimage.png&quot; style=&quot;&quot; title=&quot;&quot; width=&quot;300&quot; height=&quot;73&quot; /&gt;&lt;/p&gt;</softref_element>
            <softref_element index="elements" type="array">
                <numIndex index="1" type="array">
                    <matchString>&lt;img alt=&quot;&quot; data-alt-override=&quot;true&quot; data-htmlarea-file-table=&quot;sys_file&quot; data-htmlarea-file-uid=&quot;91&quot; data-title-override=&quot;true&quot; src=&quot;/fileadmin/user_upload/myimage.png&quot; style=&quot;&quot; title=&quot;&quot; width=&quot;300&quot; height=&quot;73&quot; /&gt;</matchString>
                    <subst type="array">
                        <type>db</type>
                        <recordRef>sys_file:91</recordRef>
                        <tokenID>80768071c2028e4faefc5a619f078928</tokenID>
                        <tokenValue>91</tokenValue>
                    </subst>
                </numIndex>
            </softref_element>
        </softref_key>
    </keys>
</softrefs>

It seems the exporter expects an array for softref_element "content".

After this warning I get an error 500:

Core: Exception handler (WEB: BE): TypeError, code #0, file (...)/myproject/public/typo3/sysext/impexp/Classes/ImportExport.php, line 854: htmlspecialchars() expects parameter 1 to be string, null given

Related code from exporter:

    protected function addSoftRefs(array $softrefs, array &$lines, int $indent): void
    {
        foreach ($softrefs as $softref) {
            $line = [];
            $line['ref'] = 'SOFTREF';
            $line['type'] = 'softref';
            $line['msg'] = '';
            $line['preCode'] = sprintf(
                '%s<span title="%s">%s</span>',
                $this->renderIndent($indent + 2),
                htmlspecialchars($line['ref']),
                $this->iconFactory->getIcon('status-reference-soft', Icon::SIZE_SMALL)->render()
            );
            $line['title'] = sprintf(
                '<em>%s, "%s"</em> : <span title="%s">%s</span>',
                $softref['field'],
                $softref['spKey'],
// Line 854:
                htmlspecialchars($softref['matchString']),
                htmlspecialchars(GeneralUtility::fixed_lgd_cs($softref['matchString'], 60))
            );
            if ($softref['subst']['type'] ?? false) {
                if ($softref['subst']['title'] ?? false) {
                    $line['title'] .= sprintf(
                        '<br/>%s<strong>%s</strong> %s',
                        $this->renderIndent($indent + 4),
                        htmlspecialchars($this->lang->getLL('impexpcore_singlereco_title')),
                        htmlspecialchars(GeneralUtility::fixed_lgd_cs($softref['subst']['title'], 60))
                    );
                }

In this case the content entry is null where an array is expected by the exporter:

I think this code from t3x-rte_ckeditor_image is related:

    // Classes/Database/RteImagesSoftReferenceIndex.php

    public function findRef_rtehtmlarea_images($content)
    {
        $retVal = false;
        // Start HTML parser and split content by image tag
        $this->htmlParser = GeneralUtility::makeInstance(HtmlParser::class);
        $this->splittedContentTags = $this->htmlParser->splitTags('img', $content);

        $images = $this->findImagesWithDataUid();

        // Assemble result array
        if (!empty($images)) {
            $retVal = [
                'content' => implode('', $this->splittedContentTags),
                'elements' => $images
            ];
        }

        return $retVal;
    }

Maybe this is an exporter issue if string/null is allowed where it expects arrays. Unfortunately I don't know much about how this works...

[jpmschuler]

Confirm: with latest TYPO3 11LTS and PHP 8.1
Steps to reproduce: have a page with a tt_content element with an RTE image. Open export via right click context menu on page, get Exception

rtehtmlarea_images is the only non-core soft reference parser in the system.

Can't see something obvious in the code.