<trackbot> Date: 25 September 2014

<scribe> scribenick: birtles

<scribe> scribe: birtles

Filters (implementation status, testsuite, next steps)

ChrisL: filters is obviously a widely-deployed technology in all the browsers and implementations
... so I expected the filters spec for CSS would be at a similar level of maturity
... I was looking at the testsuite and it only has 30 something tests
... presumably because they have to be reftests and have to be a certain style
... so it's not many tests
... and when I ran through them in different browsers there were so many failures
... it seemed like all failed in IE
... I was surprised to see that
... I wondered what's happenning, are the tests wrong? what's up?

krit: many reference tests for filter effects test filters in HTML
... and IE doesn't support filters in HTML
... also similar for WebKit
... and we added some new filter functions that are currently only supported in WebKit and Blink and soon in Firefox

<heycam> http://status.modern.ie/filters "Under consideration" for IE

krit: you're right we don't have a lot of tests yet
... my colleague has been working on that as part of implementing in Firefox
... so hopefully those tests can be imported
... since they should be in the right format

ChrisL: is there a repository for that?

krit: it's in the Gecko repository
... I can ask him to send the link

ChrisL: I know Opera released a lot of tests
... I wonder if any of those cover filters

ed: there might be some filter tests for I don't think there would be any for CSS filters

ChrisL: I remember that Presto did filters in HTML right?

ed: no, only internal experiments

ChrisL: well, that explains that
... Dirk you said there was something being working on?
... is this something we expect the browsers to do?

krit: there's active development in Blink but not so much in WebKit right now (probably change in the next few months)
... we want to have the same feature set
... but we are missing from the specification the order of the regions
... I have no status report from Microsoft
... it is being worked on in Blink
... currently stagnating in WebKit
... hopefully enabled soon in Firefox
... already available in the Nightly versions

ChrisL: that's better than I feared but still

Tav: does the Firefox implementation handle shorthands?

krit: yes

Tav: how about webkit?

krit: just for HTML

ed: does the Firefox support shorthands for SVG too?

krit: yes

heycam: is that working now?

krit: yes, it works

ChrisL: I was only testing Firefox beta so that would explain some test failures
... I will do another run-through with nightly

<heycam> layout.css.filters.enabled

krit: there might be a flag

<heycam> that is not enabled by default currently

ChrisL: do I have to enable any features on Chrome canary?

krit: it's enabled by default in Chrome canary but kind of broken
... same for webkit

<heycam> https://bugzilla.mozilla.org/show_bug.cgi?id=869828 -- the meta bug for CSS Filters in Firefox

ChrisL: is there any specific bugs I should track?
... should I report individual test failures on that bug? or a new bug?

heycam: you can file a new bug and make it block this (the above) bug

ChrisL: the other question is, how easy is it going to be to make tests for some of these filter functions?
... like the turbulence filter
... especially if we're trying to make it produce a green square
... maybe just a PNG image comparison

krit: yes, that's difficult

<ed> for blink: http://crbug.com/109224 (tagging bugreports with "Cr-Blink-CSS-Filters" is probably good, or ping me and I'll set that tag on the bugs)

krit: I think Firefox has reference tests, I'm not sure how they do that

heycam: I'm not sure we have reference tests for turbulence
... but the firefox reftest system does allow you to specify tolerance to account for anti-aliasing differences etc.
... but that might not be enough

ChrisL: so it seems like there are a few outstanding spec issues
... might it go back to LC?

krit: it's currently WD

ChrisL: yes, you're right
... I just put out a new timeline for a our charter
... which is how I discovered that filters was less advanced than I thought it would be
... Dirk, you said there was an issue about the order of filter regions

krit: currently we have a hard clipping region that is 10% around the object
... the proposal was to get rid of that and replace it with auto-computation of the bounds of the filter
... and I'm working on that auto-computation

heycam: was there still and outstanding issue about the filter resolution

krit: we removed filterRes but we still have the kernel units one
... but kernelUnitLength is needed
... it makes other parameters resolution-independent but they don't look good
... so there's still an issue but I don't think we can fix it with the current primitives we have

heycam: is it an issue of how the filters are defined?
... working on physical pixels rather than logical pixels?

krit: right, they work on physical pixels
... and if you make them work on logical pixels then you'd end up with pixelation

heycam: so is the spec going to stay as it is?
... were you planning on making any more changes to this part of the spec?

krit: no
... the primitives that are affected by that are feConvolveMatrix and fePointLight or one of the lighting filters

ed: diffuse lighting I think

krit: oh yes, that's right

<ed> specularlighting too

heycam: do you need any help/planning with regards to the test suite

krit: help is always welcome

heycam: do you need it?

krit: the spec definitely needs tests
... occasionally I try to write tests but the specification is still much larger than the number of tests we have
... so if someone is willing, yes please

ChrisL: I'm willing to do it but I'm trying to work out to make them into suitable reftests

krit: it might be good to start by looking at Firefox's reftests
... can Firefox use pixel tests?

heycam: you can always do that by putting a raster in the reference

krit: in WebKit and Blink we definitely use pixel tests because it's easier but it's not cross-browser

<heycam> https://github.com/mozilla/gecko-dev/tree/master/layout/reftests/svg/filters

heycam: some filters are easier to test as reftests than others
... e.g. the color transform tests just do it on a solid color fill

krit: but even that could differ on other implementations, by on RGB value

heycam: so you can't specify pixel tolerances for the web-platform-tests setup?

krit: you can exploit that pixels have to be clamped to within [0..255]
... and produce a test that puts you in that range

ChrisL: I can see the advantage of automated tests but it makes writing the tests hard
... and boring

Tav: I thought we agreed to have both (automated and manual)

krit: but then how do you deliver the tests?

heycam: we had that in the past, we asked people to eyeball the tests

ChrisL: web-platform-tests doesn't allow multiple references like the CSS tests do
... it allows you to say it should match one or more references
... I remember that when we first did filters there were a lot of images exchanged and equations specced so that we had two implementations that were roughly pixel equivalent

heycam: well ideally the definitions of the filter primitives are precise enough that you could say that
... the results should equal a certain image +/- a tolerance
... we could just add raster images and then work out what to do when it comes to automating
... since we don't have automation yet anyway
... maybe we don't need to worry about 1 pixel differences just yet

ChrisL: well there are almost no tests at the moment so it would be better to have something people can argue over
... I thought of a way to test feDisplacement by covering a red rectangle

heycam: some of the specs Adobe works on include a testing plan...

krit: not for filters yet

heycam: suppose you wanted to get help with testing, it would be helpful to have a list of what still needs testing

ChrisL: when you look at the tests, the metadata in the tests tells you which section it covers

heycam: so we could start by picking a section that doesn't have any tests

ChrisL: we should also have tests to cover SVG, not just HTML
... thanks for the status update, that's helpful

SVG charter

ChrisL: the charter went out just recently
... there's been a change in the W3C
... every charter that goes out needs to have positive responses from at least 5% of membership
... which equals about 20 members
... so please push your AC reps to respond

accessibility TF

<ed> question: Should there be a Co-Facilitator? If so, I would expect the SVG

<ed> Chairs would wish to nominate someone?

ed: there was a mail about the accessibility TF asking who should lead this work from the SVG side

heycam: so they're looking for a co-facilitator from our side?

ed: yes

heycam: if Rich is going to be heavily involved in the TF then I'd be happy for him to take it on if he feels comfortable with it
... I'll send a mail to Rich

ChrisL: who else is going to be on it?

heycam: I'll follow the mailing list

krit: I'll be on the call and join the mailing list

text and CSS boxes

Tav: CSS has blocks and inlines
... do SVG <text> elements behave as blocks and <tspans> as inlines?

heycam: that's roughly how it works internally in Firefox

<ChrisL> <text> is block, <tspan> is inline

Tav: I was just wondering how to apply some of the CSS properties like line-height to SVG

heycam: if you have specific questions about properties then they would be good things to bring up

TPAC

ChrisL: I updated the wiki so we have a page for the TPAC now

nikos_: do we know when the FXTF meeting is?

ChrisL: no, but it seems like it might be during the SVG part of the week

heycam: I recall something like that
... I'll send an email about that

RRSAgent: make minutes public

RRSAgent: make minutes

apologies: Rik, Thomas, Cyril

RRSAgent: make minutes

This is scribe.perl Revision: 1.138  of Date: 2013-04-25 13:59:11  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Found ScribeNick: birtles
Found Scribe: birtles
Inferring ScribeNick: birtles
Default Present: [IPcaller], ed, birtles, heycam, krit, stakagi, Tav, ChrisL, nikos_
Present: [IPcaller] ed birtles heycam krit stakagi Tav ChrisL nikos_
Regrets: Rik Thomas Cyril
Agenda: https://www.w3.org/Graphics/SVG/WG/wiki/Agenda
Found Date: 25 Sep 2014
Guessing minutes URL: http://www.w3.org/2014/09/25-svg-minutes.html
People with action items: