Width attribute on table col elements causes validation errors

[westonruter]

When post_content contains HTML such as:

<table class="auto-style1 sws_custom_table" style="width: 100%; display: table; background-color: #ffffff; border: 2px solid #eaeaea;" border="0" width="630" cellspacing="0"><colgroup> <col style="mso-width-source: userset; mso-width-alt: 8988; width: 190pt;" width="253" /> <col style="mso-width-source: userset; mso-width-alt: 6826; width: 144pt;" width="192" /> <col style="mso-width-source: userset; mso-width-alt: 8135; width: 172pt;" width="229" /></colgroup>
...</table>

This is getting converted to AMP as:

<table class="auto-style1 sws_custom_table amp-wp-inline-976a106f7543a9a2e8e3c8a97560be03" width="630"><colgroup><col width="253" class="amp-wp-inline-16106abec734b745f40158cc486aa0fe"/><col width="192" class="amp-wp-inline-91e2d2d529220418b62e99acb85f0928"/><col width="229" class="amp-wp-inline-43ca072c586bb2eaad4c6de773864f95"/></colgroup>
...</table>

This results in a validation error:

DISALLOWED_HTML: The attribute 'width' may not appear in tag 'col'.

First of all, maybe it is a bug in AMP that this attribute is not allowed. We should double check that first.

Otherwise, if it is really not allowed, then whitelist sanitizer may be incorrectly merging the width layout attribute among the attributes allowed on an element:

https://github.com/Automattic/amp-wp/blob/4e1cd227b10f2fcd6add44a24b078158aad41e53/includes/sanitizers/class-amp-allowed-tags-generated.php#L10573

See here:

https://github.com/Automattic/amp-wp/blob/4e1cd227b10f2fcd6add44a24b078158aad41e53/includes/sanitizers/class-amp-tag-and-attribute-sanitizer.php#L1468

layout_allowed_attributes is not globally-allowed it seems. Maybe that condition needs to be removed.

As reported on https://wordpress.org/support/topic/not-a-valid-amp-page-7/

[westonruter]

I can see that the problem here goes beyond just col element. The sanitizer is also not stripping width attribute from elements that don't even allow in regular HTML, for example:

<span width="123">not right</span>

The sanitizer erroneously lets the width persist.

[westonruter]

I can see that the problem here goes beyond just col element. The sanitizer is also not stripping width attribute from elements that don't even allow in regular HTML

Fix for this is in #1075

[westonruter]

Re-opening since #1064 is not merged yet.

[kienstra]

Moving To "Ready for Merging"

If it's alright, I'm moving this from "Ready for QA" to "Ready for Merging." This probably doesn't need functional testing.