Meeting minutes
Scott has volunteered to help out with the conformance test suite.
He's currently working on adding conformance spans to the text spec.
Garret will work with him to work through setting up a basic test case.
Conformance spans are identified by a unique id so we can track coverage programatically.
Vlad: conformance tests are next up because implementers will be using them to test implementations.
Vlad: looking back at woff2 was useful starting point to develop a test plan.
Vlad: our initial thinking about tests was influenced by the test plans.
Vlad: little more work, but in the long run that helped us decide what can be tested and how (especially if more than one person is involved).
Garret: sounds good, I'll work with scott and develop a plan and check that into the client test repo.
Garret: we will use the wasm polyfill until we have a browser implementation.
Garret: planning to follow the approach with woff2.
Chris: may have trouble with things like VF extension.
Chris: can do ref subsets were we check rendering. via web platform tests. Design to be run automatically by browsers.
Chris: once a test is accepted then the browsers will keep running those.
Skef: state of text rendering may cause issues. eg. chrome picks different rendering engines for different platforms.
Skef: rendering can be different.
Garret: when testing with VF's we can use very large deltas so that the with/without renderings are very different then use fuzzy matching.
Skef: does the spec allow the initial font to be static?
Garret: yes, can be upgraded to VF by patch.
Skef: in the case of CFF2 this might be the first case of static CFF2.
Chris: can be also be a vf from the start but the ranges aren't fully covered.
Vlad: we might want to have separate functional tests that aren't conformance tests (since those are all tied to specific conformance statements).
Garret: could use the framework we develop for conformance testing to develop additional tests outside the scope of conformance tests.
Skef: in woff2 how does brotli get tested?
Vlad: brotli was an external reference and not tested directly.
Skef: for IFT we have new brotli patching capabilities.
Garret: that's similar in that it's also an external reference.
Skef: how much of shared brotli was developed for IFT?
Garret: shared brotli existed prior to IFT
Skef: there's conformance tests and then there's tests that the client will work. With IFT there's more flexibility then woff and if we don't test all of those corners we will end up with wonky clients.
Chris: risk that tests made with one generator will break.
Chris: clients will break on fonts generated by a different encoder.
Skef: integration challenge were browsers already have brotli from woff2 but not yet the shared brotli.
<ChrisL> RFC2119
<ChrisL> https://
Skef: have a lot of failure cases, how is that handled?
Garret: on error clients can continue to using the font with whatever has been loaded so far, but extension stops.
Skef: how do we handle the optional behaviour?
Garret: tests would only check the failure codepoints and not probe the ones with optional behaviour
Skef: characters should render correctly in the font we're subsetting or in the fallback font.
Vlad: with fallback fonts in woff2 we had a set of fallback fonts provided with the test suite.
Scott: could fallback font be a base64 encoded?
Garret: yeah seems like that should work.
TPAC
Vlad: registration for TPAC has recently been opened
Vlad: hotel has limited number of rooms available so consider registering early.
Vlad: for participation, Skef have you had updates on attending?
Skef: will not be coming.
Vlad: making arrangements to attend in person.
Chris: hope I can attend.
Garret: I have been approved to attend.
<ChrisL> w3c/
Skef: when doing the PR's don't update the HTML until just before the merge so that others can make edits
Discussed the possibility of generating html automatically via CI, Garret will look into if that's possible.
<ChrisL> https://
Vlad: let's plan the next call around the test plan once it's available.