1800941 - Product image at www.triumphmotorcycles.co.uk does not load in Firefox, but does in Chrome and Safari

Status: NEW

(Core :: Graphics: ImageLib, defect)

Description

[Thomas Wisniewski [:twisniewski]]

The 360-degree view feature on this product page is broken because the big image containing all of its frames does not load in Firefox, with the message Image corrupt or truncated appearing in the console log. It also does not load in a new tab; it becomes blank after loading most of the image. This image seems to load fine in Chrome and Safari, leading to this being a webcompat issue.

Comment 1

[Timothy Nikkel (:tnikkel)]

The image is 69k pixels in width, and we have a 64k limit.

See https://bugzilla.mozilla.org/show_bug.cgi?id=1770513#c4 and following comments for some context on that. So perhaps we should increase that limit.

Comment 2

[Thomas Wisniewski [:twisniewski]]

Of course for webcompat, it would make sense to match whatever limit Chrome and Safari have (or to get them to change their limits). Otherwise we'll probably just run into more sites unsure of what's going on and calling it a Firefox bug since it works in the other browser engines. Do you happen to know what limits (if any) Blink and WebKit use?

Comment 3

[Timothy Nikkel (:tnikkel)]

I don't know their limits off hand but after the patch in bug 1770513 we will only have two limits: the total size of the decoded surface fits in a signed int (we are not likely to want to relax this requirement I don't think), and the decoded surface can fit into the surface cache. The surface cache holds all decoded images in the process, once it hits it's limit we cannot decode any more images. I don't think I've ever seen any bug reports about hitting it's limit. The size of the surface cache will usually be 2GB on desktop. The image here is 276MB. Sites using images over 2GB are probably going to run into problems on mobile where phones can sometimes still only have 2 or 3 GB total memory. So I think we probably won't run into more webcompat issues after bug 1770513 lands.

Comment 4

[Thomas Wisniewski [:twisniewski]]

Ah, I see. Alright, that sounds fine to me. Let's hope bug 1770513 indeed is all we need.