We might want to merge https://chromium.googlesource.com/chromium/blink/+/cbf0eb3458947ff81a836599a692a9c1b32724c3 Now that detecting meta@charset doesn't block processing the resource, we need to be more careful about how we prioritize encodings from meta@charset relative to other mechanisms. Specifically, we meta@charset is supposed to be less important than autodetection from the "<?xml" preamble but more important than encodings detected from sniffing the content itself.
We might take the regression test, but the code change is rushed, confusing, and it's just fixing a recent Blink regression anyway.
Also, what's "NTP"?
Landed the test in <http://trac.webkit.org/r153916>.
(In reply to comment #2) > Also, what's "NTP"? NTP = New Tab Page. It's the page you see in Chrome when you create a new tab. That's just where I noticed the problem, but the probably also occurred on other web sites. You shouldn't need the code change unless you want to try to detect meta@charset without buffering. If you do want to do that, there's another (even uglier) CL that you'd need to merge from later.
Thanks Adam!