[Webkit-unassigned] [Bug 68682] REGRESSION (r93326): Thumbnails do not appear inside MobileMe gallery

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=68682





--- Comment #3 from James Simonsen <simonjam at chromium.org>  2011-09-29 21:50:40 PST ---
Short version:

The site is broken. We could do a site-specific hack to fix it.

Long version:

I don't have a me.com account, so I only debugged the reload case. I can debug the other case too, if desired.

Here's what I see...

The page seems to use some sort of preloading technique where it constructs an image element (I1) with the thumbnail's URL. The page then constructs 2 more image elements (I2 & I3) with the same URL. I3 is then used in HTMLCanvasElement::drawImage() and is what you normally see on the page.

During the first page load:
1. I1 is new and loads.
2. I2 and I3 reuse the CachedImage loaded by I1, because it hasn't expired.
3. I3 is drawn successfully.

During the reload: (all resources must revalidate because we're reloading)
1. I1 must revalidate and loads.
2. I2 must revalidate and loads.
3. I3 must revalidate.
4. drawImage(I3) is called, but I3 is still in the middle of revalidating. drawImage() fails.
5. I3 loads.

According to the HTML5 and HTTP specs, WebKit's new behavior is correct and the page is broken. The page should wait for I3's load event to fire before drawing to the screen.

Here's the steps I followed in the specs to come to that conclusion:
1. http://dev.w3.org/html5/spec/Overview.html#update-the-image-data step 6.
2. http://dev.w3.org/html5/spec/Overview.html#potentially-cors-enabled-fetch step 1.
3. http://dev.w3.org/html5/spec/Overview.html#fetch step 5.
(Note: We're not already downloading, because I2 already completed.)
4. http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html section 14.8, step 2.
(Note: We MAY use the cache without revalidation, but don't because the user pressed reload.)
5. http://www.whatwg.org/specs/web-apps/current-work/multipage/the-canvas-element.html#dom-context-2d-drawimage "If the image argument is an HTMLImageElement object that is not fully decodable ... then the implementation must return without drawing anything.
6. http://www.whatwg.org/specs/web-apps/current-work/multipage/embedded-content-1.html#img-good
7. http://www.whatwg.org/specs/web-apps/current-work/multipage/embedded-content-1.html#img-all

We fail at #5, because the test in #7 fails, because #3 hasn't completed yet.

Before bug 52153 was fixed, we never got to the revalidation step in CachedResourceLoader::determineRevalidationPolicy(). We used to always short-circuit out, because we'd already loaded the resource. The old behavior violates the HTTP spec regarding must-revalidate.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.