-
Notifications
You must be signed in to change notification settings - Fork 7.6k
Extension Manager font is hard to read on Windows #4554
Comments
Tagging Sprint 28 since we should at least consider whether we need to do anything here before shipping. We might also want to scrub the UI for other uses of light+small Source Sans that would experience similar issues. |
I can confirm that small text in other places like the About dialog is also affected. (Also in the About dialog, you can see a change in image antialiasing -- all the gravatars are shifted about 1px within their frames). But I don't see hue changes anywhere other than the text, so this appear to not involve color profiles directly. It may be that something about compositing or antialiasing changed at the same time (or in support of the color correction) that is causing this. |
The CEF change definitely only affects text rendering in mainstream cases -- I've verified on a number of websites now. Text is noticeably thinner (especially small text) and colored text is noticeably desaturated, but images and background fills are completely unchanged. |
@JeffryBooher Please take a look if there was anything we could do in the short term (this sprint). @dangoor is out today, please chat with @peterflynn or @njx if you felt we should defer. |
@larz0 We'd like your thoughts on this. I'm proposing we do a workaround by bumping from font weight from light (200) up to normal on Windows. That's heavier than it looked before, but much more readable: With workaround: IMHO since this is primary content it's ok for the text to be less grayed out... |
I agree @peterflynn. Thanks for the heads up! |
Did some further research. Looks like this is caused by Chromium bug 146407, which has the same symptoms: fainter text and desaturated text colors only/especially on Windows. It broke in Chrome 22, which matches what we're seeing: in Sprint 27 we upgraded CEF from 1180 to 1453, which corresponds to upgrading from Chrome 21 to Chrome 27 -- and that's indeed when we first saw the problem. (Considering that was jumping forward 10 months in Chrome's history, I'm almost surprised that font rendering and a couple V8 crashes are the only unexpected changes we've seen!) |
Damn… hope that's a high priority bug for the Chromium Team. |
There's a comment in that bug by a Chrome dev saying the cause is that Chrome has started assuming all displays are calibrated to sRGB because "most displays today" are (which in my experience isn't true). It also seems that no other browser (including Safari) makes that assumption, judging by all the comparison screenshots there showing Chrome differing from the rest of the pack. But the developer comment almost suggests they consider it NAB. Would be interesting to get thoughts from some in-house font experts here (e.g. TypeKit) on what sorts of colorspace/gamma assumptions are safe... The bug comment thread near the end gets a little jumbled up with a different issue (137692) where certain poorly-hinted webfonts render with no AA at all on Windows. That's blocked awaiting DirectWrite support and was never working; whereas the "spindly" AA that's the issue here is effectively a "regression" from Chrome 22 onwards. There's plenty of momentum on the DirectWrite stuff but unfortunately little or none on the "spindly" issue we're hitting -- although it's possible using DirectWrite for all text would remove the sRGB assumption and go back to whatever the OS assumes natively (which seems more correct given that text looks better in IE). |
One interesting thing to note about this is that the effect depends on text brightness -- all text gets paler, so light colored text actually got fatter while dark colored text got more spindly. You can see an example of light text "fattening" if you compare the working set list between sprints. |
@larz0 I noticed one other place that might be affected by this: dialog box messages. The font size is a little bigger in dialogs, so it's not nearly as bad... but it's still noticeably thin. Do you think those are hard-to-read enough to warrant correcting too? |
…, which is a Chromium bug, by avoiding the lightest font-weight on Win. Lighten the text slightly so it's still a little muted, like the design looks on Mac and with older CEFs on Win.
Work around #4554 (Extension Manager font hard to read on Windows)
Closing. We'll file a separate bug for the other dialogs and punt that to next sprint. |
@peterflynn FYI #4567 |
We're looking at the same issue with Topcoat's website. Looks better on firefox. |
…eanups * origin/master: (30 commits) turn off optimization in acorn (but not acorn_loose right now) cleanup unit test prefs temporarily switch to my tern Revert "Revert "Workaround for the Tern crash."" Update README.md Work around #4554 (Extension Manager font is hard to read on Windows), which is a Chromium bug, by avoiding the lightest font-weight on Win. Lighten the text slightly so it's still a little muted, like the design looks on Mac and with older CEFs on Win. Revert "Workaround for the Tern crash." Updated by ALF automation. Re-add toolbar hover. Fix some button appearance issues. Updated by ALF automation. * Fix bug #4548 - remove Save As from folder tree context menu * Fix exception thrown when File > Save As invoked with nothing open * Update docs for working set events to reflect PR #4450 integration tests for registerInlineEditProvider Fixes after review JSDoc fixes. Fix for extensions compare show error message and add safety check For #4535, show error message but not 'remove' link for bad extension in dev folder Refactor provider callback for export Generalize registerInlineEditProvider and registerInlineDocProvider to take an optional priority parameter change upper limit to 16000 ...
… the worst offenders similar to #4554, making the font weight heavier and tweaking the text color to match Mac more closely.
The small, extra-lightweight font used for extension descriptions is very spindly on Windows, making the text hard to read and less attractive.
This seems to have regressed from before. Today:
Sprint 26:
I'm guessing this is the same Chromium/CEF font rendering change that caused #4391...
The text was updated successfully, but these errors were encountered: