Re: [filter-effects] feImage with missing image resource

On Nov 3, 2014, at 7:04 PM, Rik Cabanier <cabanier@gmail.com> wrote:

> 
> 
> On Mon, Nov 3, 2014 at 9:38 AM, Paul LeBeau <paul.lebeau@gmail.com> wrote:
> The Error Processing section of the SVG spec says: "The document shall be rendered up to, but not including, the first element which has an error."
> 
> Would not a bad input, such as a bad URL reference be counted as an error?  If so, processing should(?) stop at the filter primitive before the <feImage>.  Which to me sounds exactly like option 3.

In general that is true. I would assume that there are exceptions though. Especially with filter effects that do not actually paint but operate on the pixel values.

> 
> If we define that a missing image is allowed, it wouldn't be an error any more :-)
> Moving forward I believe we should make things more permissive.

I agree with the permissive fallback up to a certain point. Filter Effects influences the presentation of an element and we have to be careful with the things that happen if the filter does not just fail, but does something completely different. For missing feImages, I think all three cases make sense.

>  
> On 4 November 2014 06:25, Rik Cabanier <cabanier@gmail.com> wrote:
> 
> 
> On Mon, Nov 3, 2014 at 6:14 AM, Dirk Schulze <dschulze@adobe.com> wrote:
> Hi,
> 
> I checked the behavior of different SVG viewers on <feImage>[1] with a missing image resource. It appears that we have three behavior categories across 7 different SVG viewers (6 if you still count WebKit and Blink as one). Tested: Safari, Chrome, IE, Firefox, Opera (presto), Illustrator, InkScape (no result), Batik.
> 
> All viewers apply an SVG filter on content even if <feImage> has a missing resource. Other then that, viewers either:
> 
> 1) treat <feImage> as a transparent black image taking all primitive regions into account,
> 2) replace the missing image with a “missing image” icon and take all primitive regions into account or
> 3) treat <feImage> and all successor color manipulating primitives as null filter.
> 
> Example: https://www.w3.org/Bugs/Public/attachment.cgi?id=1531
> Description: https://www.w3.org/Bugs/Public/show_bug.cgi?id=27221#c0
> 
> Option 1) looks reasonable for me and is easy to implement/specify. Option 2) looks strange but at least gives authors feedback of what is going on. I am unsure how to specify 3) and how it actually is implemented. Maybe someone from Mozilla could clarify.
> 
> I agree that 1) is the most reasonable option. 
> Can you break out which browser does what option? 

The bug report has the following summary:

IE, Opera: Missing image is treated as transparent black.
Firefox, Illustrator: Missing image acts as a null filter even for filter primitives applied on top of feImage.
Safari, Chrome: General behavior as for IE, Opera but taking the "missing image" instead of transparent black.

Greetings,
Dirk

> 
>

Received on Monday, 3 November 2014 22:09:56 UTC