Opened 8 weeks ago
Last modified 7 days ago
#64683 new enhancement
_print_scripts should use the wp_inline_script_attributes filter
| Reported by: |
|
Owned by: | |
|---|---|---|---|
| Milestone: | Awaiting Review | Priority: | normal |
| Severity: | normal | Version: | 6.9.1 |
| Component: | Script Loader | Keywords: | |
| Focuses: | Cc: |
Description (last modified by )
On my site, we want to use a content-security-policy. And in this policy, we would like to not include support for 'unsafe-inline' scripts.
We can include inline scripts, as long as they have a nonce in them. That is, instead of just a <script> tag, if they included a <script nonce="xxxxxxx">, where the nonce is generated on every page load, and if our Content-Security-Policy contains script-src 'nonce-xxxxxxxx'.
Some of the scripts generated by WordPress core—and, indeed, by plugins—print themselves out using the wp_get_inline_script_tag function. When a script does that, then our theme can add a filter on the wp_inline_script_attributes hook, which adds the nonce according to our own logic.
However, there are some inline scripts printed by WordPress core that do not use wp_get_inline_script_tag, and with these scripts, there is no way to for our theme to add a nonce to the script tag, and therefore no way to allow these scripts to run in the context of a Content-Security-Policy that does not allow 'unsafe-inline' scripts.
The scripts added by WordPress core are at least those that are added by wp_default_scripts. Ultimately, these are printed out using the function _print_scripts, in wp-includes/script-loader.php. It prints the script tag using
echo "\n<script{$type_attr}>\n";
where $type_attr is either the empty string or "type='text/javascript'".
I propose that _print_scripts be changed so that instead of echoing the script directly, it constructs the code it wants to output, and prints it onto the page using wp_get_inline_script_tag, so that themes or plugins can add filters on the wp_inline_script_attributes hook to add a nonce (or do anything else).
Is that a good approach? If so, I can submit a pull request.
If there's another approach that would be better, I could do that. Perhaps we want to have a different hook here for some reason.
Related: #58664