This document is licensed under a Creative Commons Attribution 3.0 License.
This document is intended to be used as a guideline for the testing activities related to the CSS Regions spec [css3-regions]. Its main goal is to provide an overview of the general testing areas, possible caveats and testing aspects not immediately apparent from the spec. Also, it provides a means of tracking the progress of the CSS Regions spec testing.
This document is not meant to replace the spec in determining the normative and non-normative assertions to be tested, but rather complement it.
This document is merely a public working draft of a potential specification. It has no official standing of any kind and does not represent the support or consensus of any standards organisation.
As CSS moved away from the monolithic development of CSS 2.1 to the modular development of CSS 3, the number of proposed new features and the complexity of the layout landscape have increased dramatically. While this directly translates to increased flexibility and agility in adopting and implementing new CSS features, it also increases the complexity of testing CSS features and the need for coordinating the testing efforts. Also, the need for testing coordination increases as crowd-sourcing efforts like Test the Web Forward present people less familiar with the processes and policies of the W3C with the opportunity to contribute new tests.
Except when defining new behaviors or redefining old behaviors, the implicit assumption for new CSS modules is that they play nicely with other modules or properties defined in CSS 2.1 [CSS21]. As CSS Regions is a spec that touches many aspects of layout, styling and CSSOM, it's not unreasonable to want to test the spec against these implicit assumptions, too.
This testing strategy document is meant to complement the CSS Regions spec and the existing test suite by providing an overview of the testing areas (especially the less apparent ones) and tracking the progress of the testing activities against these test areas.
To ensure a comprehensive test suite with useful, high quality tests, a number of goals are proposed. They range from process goals (how to conduct testing) to implementation goals (how to write good tests).
An important vector in creating successfully testing CSS Regions is to enable easy test contributions, both from W3C partners and from non-W3C members that wish to contribute. This is achieved by clearly marking and explaining the areas that need testing, linked to existing tests and general testing progress.
In order to increase the quality of the test contributions, this document offers a set of guidelines for conducting testing (see section 4. Approach) and a testing progress tracker to increase the surface coverage of tests (see section 7. Test progress tracking).
In terms of actual tests produced for the CSS Regions, the main goal
is to ensure that most tests are automatable (i.e. they're either
reftests or use testharness.js). Even where manual tests
are absolutely necessary they should be written so that they can be
easily automated – as there are on-going efforts to make
WebDriver [webdriver] automated tests a first class citized in W3C
testing. This means that even if a manual test requires user
interaction, the validation or PASS/FAIL conditions should still be
clear enough as to allow automatic validation if said interaction is
later automated.
There are a number of risks associated with creating a high-quality test suite for CSS Regions. The most important ones are listed below.
The CSS Region spec introduces a significant change in the way layout
can be done and as a consequence, many of the assumptions that hold in
the context of CSS 2.1 must be re-validated in the context of CSS
Regions. In the meanwhile, a lot of new layout modules have been
proposed, with various degrees of implementor support and maturity.
Also, non-CSS specific specs introduced concepts that affect how style
is propagated (e.g. Shadow DOM [shadow-dom]) or how elements are
rendered and interact with their containing documents (e.g.
<iframe seamless> in HTML 5 [HTML5]).
All the above factors increase the testing surface and the number of the possible cases that might need an explicitly specified behavior in order to ensure intuitive and not-unexpected results as well as stable interaction with widely used browser features (that might not be yet so strictly specified).
In terms of specifying the expected behavior, the current approach is to try and specify it for specs that are final or nearly-final and to just make a note of the possible interactions and unspecified behaviors in the case of specs that are still in flux. In exceptional cases, a new spec might be created to cater for the needs of multiple specs (e.g. the CSS Fragmentation spec [css3-break]).
Given the complexity of the spec, a big number of tests will need to be created to produce a test suite that can ensure interoperability between implementations.
In this context, the main purpose of this document is to provide useful informations for creating and contributing tests in an effective manner in terms of coverage and test quality.
As spec testing cannot be realistically separated from testing a particular implementation (except for the very simple cases), the approach proposed for testing is one that tries to first cover as many areas as possible, instead of deep diving on a certain feature or aspect of the spec first. A side benefit of this approach is that the spec tests can be used at any time to gauge the level of support of a certain implementation.
Having this breadth-first approach in mind, tests will be created for the testing areas listed in section 5. Testing areas. Testing will be done in multiple passes, each aimed at covering more specific edge-cases.
These are testing areas normatively defined by the spec. They cover things explicitly or implicitly defined in the CSS Regions spec. Please note that while detailed, this list is not necessarily exhaustive and normative behaviors may not be contained in it. When in doubt, consult the CSS Regions spec or ask a question on the mailing list.
Below is the list of explicit testing areas:
flow-into and flow-from;
flow-into should be tested both with and without
the content switchregion-fragment::region() functional pseudo-elementregion-avoid,
region-always – see [css3-break] for
detailsNamedFlow interface
Region interface
regionoversetchange event
regionfragmentchange event
getClientRects() and
getBoundingClientRects()offsetTop, offsetBottom,
offsetLeft, offsetRight and
offsetParent
These are testing areas either normatively defined in other specs that explicitly refer to the CSS Regions spec (e.g. [css3-break]) or simply not explicitly defined, but implied by various aspects of the spec (e.g. processing model, CSS 2.1 compliance, etc.). Please note that while detailed, this list is not necessarily exhaustive and normative behaviors may not be contained in it. When in doubt, consult the CSS Regions spec or ask a question on the mailing list.
Below is the list of implicit testing areas:
<iframe>
<iframe> and
<iframe seamless><video><canvas><table><style scoped>:active, :hover)Below is a list of people you should reach out to if you have any questions related to this document or testing CSS Regions in general:
Currently test progress tracking is done via gitHub milestones and issues. Each testing level has a corresponding milestone with issues created for each testing area. Currently, only the milestones for smoke-level testing have been created, with milestones for the second pass of testing to be added soon.