[mu4e bug] URLs in gnus-view for HTML mails not recognized

[mbunkus]

Describe the bug

Functions that operate on URLs such as mu4e-view-go-to-url or mu4e-view-save-url fail to find URLs in HTML mails where the displayed text of the <a> element does not contain a URL (href != link text). They do find them if href == link text.

For example:

• Works: <a href="https://github.com/">https://github.com/</a>
• Does not work: <a href="https://github.com/">Github</a>

In case 1. the URL is numbered & offered when using mu4e-view-go-to-url. In case 2. it isn't.

To Reproduce

Using the gnus-based view, open any HTML mail that contains links where the link text isn't a URL. Invoke mu4e-view-go-to-url. Observe that the URL isn't numbered, and that you cannot open it.

Environment

Arch Linux
Emacs 27.2
mu+mu4e 1.6.3-1 (Arch Linux package's version number)

Checklist

• you are running vanilla emacs (i.e. without Doom, Evil, Spacemacs etc.) (otherwise, please try to reproduce without those
• you are running mu4e without any third-party extensions (otherwise, please try to reproduce without those)
• you are running either the latest 1.4.x release, or a 1.5.11+ development release (otherwise, please upgrade).

[djcb]

It's more of an accident that some URLs are "activated" in an HTML mail; we need to turn that off for such messages.

[mbunkus]

I've implemented two workarounds available here, in case anyone needs something like this.

The first function collects all URLs in the current view buffer from text properties & displays them along with the anchor's text in a read-only buffer. Enables easy clicking on them with the mouse, even if you're running Emacs over ssh, as I do (from Windows via putty). Pressing q will kill the buffer again, returning to the message view. I've bound this function to g.

The second function takes the URL stored in text properties at point and starts a wget process with that URL asynchronously. This natural key bind for this is k, obviously.

Both workarounds are just that; quick hacks that rely on certain conditions (e.g. it only works if the shr renderer is used by gnus for displaying HTML messages, meaning mm-text-html-renderer must be set to 'shr, which is the default nowadays), aren't configurable and definitely not meant to be included in mu4e proper. Just as a small help for others who might need something now.