Bug 16893 – text-shadow offset-values wrong since Safari 3.0.x

• Component: CSS
• Status: RESOLVED FIXED

[Harald Szekely]

This bug can be seen in all versions of safari 3 on Windows an Mac.

We use text-shadow heavily in our web-applications, sometimes as shadow, sometimes as bevel effect. I don't know exactly when it happened, but the bug was definately not present in Safari 3.0.0 and 3.0.1. It appeared in one of the versions afterwards. 

Until that bug you would use a CSS-Declaration like that: "text-shadow: 0 1px 0 #000;"  for getting a 1px black not antialiased "duplicate" text below the original text.

Now you have to enter: "text-shadow: 0 2px 0 #000;" for getting the same result! Why? 

If you enter: "text-shadow: 0 1px 0 #000;" you'll see no shadow at all! What for is an offset of 1px i it doesn't show, but shows in other browsers?

This change breaks all of our page-designs and was not announced anywhere, as far as i know! Now that Opera 9.5 supports text-shadow as well, the problem gets even worse! If you build pages for Safari, they now look awfull in Opera und vice versa! 

Will that change remain? If yes, then we'll have to stop using text-shadow, for the results of using it will be totally unpredictable! I think an offset of 1px has to be 1px, not 0px, as it is now! 

Is there any documentation or stuff where you people have talked about the change? I can find no article about that anywhere onthe web, so i think, noone has recognised it yet. Although that seems strange to me ...

Bye!

[Adam Roben (:aroben)]

Harald, can you provide us with a test-case that shows the difference between current Safari, old Safari, and Opera? We did make a change in this area recently, but our intent was to bring Safari better in line with other browsers. We'd definitely want to know if we've made the situation worse!

[Harald Szekely]

Created attachment 18474 [details]
Opera 9.5b

 This is how Opetra 9.5 Beta renders: (color: #fff; text-shadow: 0 2px 0 #000;) on Windows XP with Clear Type enabled. It shows the correct 2px offset.

[Adam Roben (:aroben)]

(In reply to comment #1)
> We did make a change in this area
> recently, but our intent was to bring Safari better in line with other
> browsers.

Specifically, we fixed bug 14736 and bug 12943.

[Harald Szekely]

Created attachment 18475 [details]
Safari 3.0.4 (523.15)

This is how Safari 3.0.4 (523.15) renders: "color: #fff; text-shadow: 0 2px 0 #000;" on Windows XP with Clear Type enabled. It renders the offset as 1px only.

The effect is the same on Macintosh Tiger and Leopard.

[Adam Roben (:aroben)]

(In reply to comment #4)
> Created an attachment (id=18475) [edit]
> Safari 3.0.4 (523.15)
> 
> This is how Safari 3.0.4 (523.15) renders: "color: #fff; text-shadow: 0 2px 0
> #000;" on Windows XP with Clear Type enabled. It renders the offset as 1px
> only.
> 
> The effect is the same on Macintosh Tiger and Leopard.

As far as I can tell, the rendering you've posted is how any currently-released version of Safari would render your example. We've recently brought WebKit in line with Opera's rendering by fixing bug 14736, but this fix has not yet made it into a release. Would you be willing to try a WebKit nightly build from <http://nightly.webkit.org/> on Tiger to see if it makes Safari's rendering closer to Opera's? (The fix will be available on Windows and Leopard, but the fix is not in WebKit itself so a nightly build won't show a difference.)

[Harald Szekely]

> Would you be willing to try a WebKit nightly build from
> <http://nightly.webkit.org/> on Tiger to see if it makes Safari's rendering
> closer to Opera's? (The fix will be available on Windows and Leopard, but the
> fix is not in WebKit itself so a nightly build won't show a difference.)

I can do that, but it will take some time. Here at work there's no Mac where i can play around with nightly builds. I'll post the results when i've done the test. 

(In reply to comment #5)
> Would you be willing to try a WebKit nightly build from
> <http://nightly.webkit.org/> on Tiger to see if it makes Safari's rendering
> closer to Opera's? (The fix will be available on Windows and Leopard, but the
> fix is not in WebKit itself so a nightly build won't show a difference.)

The fix in the nightly addresses Leopard as well, so you can test with a nightly on Leopard or on Tiger.

(In reply to comment #6)
> I can do that, but it will take some time. Here at work there's no Mac where i
> can play around with nightly builds.

I don't know the exact situation where you work, but I thought I'd mention that nightly builds do not install anything or modify the system in any way (in particular, they do not replace the system WebKit framework or affect the behavior of Safari when not launched using the WebKit nightly).

[Harald Szekely]

> I don't know the exact situation where you work, but I thought I'd mention that
> nightly builds do not install anything or modify the system in any way (in
> particular, they do not replace the system WebKit framework or affect the
> behavior of Safari when not launched using the WebKit nightly).

Thanks, but i know that. Although i lost all my bookmark icons and never got them back when i first tried to test a nightly some time ago. Had to delete all icons manually. 

But my boss won't let me do that anyway, sorry. I can test that in about 2 or 3 hours at home. 

[Dave Hyatt]

This is a bug in Leopard.  Shadows are off by 1 pixel throughout the system.  It's not a bug in Safari but is in fact a bug in the underlying shadow code used by the OS.  It has been fixed and will be patched in a software update soon.

I am closing this out, since it is not a bug in WebKit.  Thanks for the report.

[Dave Hyatt]

If you wish to file this shadow bug on Leopard itself you can do so at http://bugreport.apple.com/

However they are aware of the bug and have fixed it, so it probably isn't necessary to do so.

[Harald Szekely]

Sorry, but how can this be a bug in Leopard when i see it on Windows too? Just clear that for me please! Do i have Leopard on my Windows? ;-)

[Adam Roben (:aroben)]

(In reply to comment #12)
> Sorry, but how can this be a bug in Leopard when i see it on Windows too? Just
> clear that for me please! Do i have Leopard on my Windows? ;-)

More specifically it is a bug in CoreGraphics on Tiger, Leopard, and Windows.