JavaScript embedding doesn't add directory slash importing image paths, breaks images

[quimgil]

This looks like… a simple bug?

• Example source page: http://surtdelcercle.cat/index.php?view=blog&id=6

• Images are called with relative paths: src="media/torbrowser.jpg"

• Resulting URL: http://surtdelcercle.cat/media/torbrowser.jpg (correct)

• Imported page: https://la.confederac.io/t/surt-del-cercle/2865 (notice the broken images)

• The imported image URL: http://surtdelcercle.catmedia/torbrowser.jpg (broken)

Note the slash missing in “catmedia”.

I didn’t find a report for this bug, although this one might be related: Embedded topics don't handle "//"-images properly

Can anyone reproduce? Or is there a fix that we could implement on our side? (other than defining images with absolute URLs, which we could do moving forward but what about the articles already published)

[pfaffman (Jay Pfaffman)]

@quimgil

Can anyone reproduce?

Well, it’s been a while, but yes.

https://github.com/discourse/discourse/blob/0b3f9638047a37bf6626bf2d6aa74299be5cd55e/app/models/topic_embed.rb#L217-L224

The above code appears to normalize image URLs that start with a / (fixing up the hostname), but not those relative to the current URL.

Maybe that should instead be this?

    if src.present? && !src.start_with?('http')

If I’m right, I think that the code a few lines above has the same bug.