Implementation Report for the CSS Writing Modes Level 3

1. Results

1.1. Summary

The test suite contains 1159 tests for mandatory requirements (i.e. RFC2119 “MUST”, and equivalents), and 27 tests for optional requirements (i.e. RFC2119 “SHOULD” or “MAY”, and equivalents).

1127 of 1159 tests for mandatory requirements pass in two or more implementations; 32 do not. 14 of the failing tests are not needed to meet CR exit criteria for CSS Writing Modes Level 3 (see “Tests not Needed for CSS Writing Modes Level 3 Exit Criteria” for why). 9 of the tests that do pass in two implementations do so by taking into account non browser implementations. The CSS-WG has examined the remaining failing tests, and based on the following considerations, believes that the Candidate Recommendation exit criteria are fulfilled:

• The success criteria established in the working group charter specify that In order to advance to Proposed Recommendation, each module is expected to have at least two independent implementations of each of feature defined in the specification. With multiple browser engines (as well as non browsers) implementing the specification, and each feature being implemented by two or more, this criterion is fulfilled, even though there remains a few tests for certain specificities of some features that do not yet have two passing implementations.
• Quite a few of the failing tests have work-in-progress implementations.
• CSS Writing Modes Level 3 is a particularly complex specification, affecting all aspects of layout. For a project of that complexity, no matter how much time we spend on it, there will always be a few more bugs to fix, and a few more nuances to get right. We should not let the fact that we're being thorough with testing impede the progress of specification, and we believe that it is appropriate to signal to the broader community the degree of stability of this specification by publishing it as a Proposed Recommendation, then as a W3C Recommendation.

9 of 27 optional tests pass in two or more browser implementations; 18 do not.

1.2. Detailed Analysis

1.2.1 Tests for Mandatory Requirements

1.2.1.1. Tests Lacking two Implementations

1.2.1.2. Noteworthy Tests with two Implementations

1.2.1.3. Tests not Needed for CSS Writing Modes Level 3 Exit Criteria

1.2.2. Tests for Optional Requirements

The following tests for RFC2119 “Should” or “May” behavior lack two passing implementations. They are not required to pass in order to meet CR exit criteria for CSS Writing Modes Level 3.

2. Methodology

This implementation report was prepared to document the passing of the Candidate Recommendation exit criteria by the CSS Basic Writing Modes Level 3.

The test suite is considered to be complete with no known or suspected bugs at the time of its publication. If you find any errors, would like to contribute additional tests, or have other feedback, please report it on the web-platform-tests project or to the archived public discussion mailing list for CSS test suites, public-css-testsuite@w3.org.

A continuously updated version of the test suite is available. This report was prepared using the tests in the test suite as of 16 September 2019.

Note: Many of the tests in this test suite are somewhat old, and interact poorly with the automation mechanism of the web-platform-test project. Issues such as antialiasing problems or slow loading or external resources may cause automated screenshot comparison to fail, causing false negatives. Test results reported here have been manually verified.

This report was prepared on the basis of testing 4 browser engines:

• Firefox (engine: Gecko, version 70.0a1; except a few tests on version 72.0a1)
• Chrome (engine: Blink, version 78.0.3887.0; except a few tests on version 79.0.3942.0)
• Safari (engine: Webkit, version 13.1 Release 89)
• Edge (engine: EdgeHTML, version 18)

Microsoft has recently discontinued the EdgeHTML engine and moved Edge to Blink, but EdgeHTML-based Edge is included for included as evidence that the features of this specification can be implemented by multiple independent parties.

There exists non-browser implementations of this specification as well, notably CSS print formatters. The entire test suite was not run for these implementations; only tests which lacked two browser implementations were verified, and cases were they did pass are mentioned in this report. The following were tested:

• Prince (version 12.5)
• Antenna House Formatter (version6.6 MR7)

3. Raw Results

The following table gives the results for browser implementations, as of the 16 October 2019.