[css-viewport-1] Add automation support for viewport segments

[darktears]

This commits define ways to emulate display features of a device. A display feature is a hardware feature that acts as a divider and creates logically separate regions of the viewport.

Using WebDriver developers and user agents will have the opportunity to test or emulate several types of devices.

[bramus]

I don’t think I am the right person to co-review this. I would assume the authors of the https://w3c.github.io/webdriver/ spec are the people you are looking for.

[darktears]

@OrKoN would you be able to review this please?

The approach is modeled after Compute Pressure, Device Posture and Generic Sensor.

[OrKoN]

Thanks for the ping, let me try to review it:

1. Is there a rendered version that includes changes?
2. have you considered specifying it as part of https://w3c.github.io/webdriver-bidi/#command-browsingContext-setViewport instead?
3. is anything expected to happen when [=[[DisplayFeaturesOverride]]=] is set? currently, it seems the override is set for the future use by the algorithm and the command returns?

[darktears]

@OrKoN

1. If by rendered, you mean after bikeshed is run, not yet. I can find a way to throw the generated html somewhere for you to see.
2. No but I'm wondering if it's a good fit. setViewport command only allows to set the width/height and devicePixelRatio but do not seem to include any other viewport properties. If we decide to add a parameter in setViewport then where would all the explanations and diagrams go?
3. I don't understand your question. Setting [=[[DisplayFeaturesOverride]]=] is the very purpose of this. When the override is set then the segments property is recalculated using the array of display features from DisplayFeaturesOverride slot coming from the setdisplayfeatures command.

[OrKoN]

If by rendered, you mean after bikeshed is run, not yet. I can find a way to throw the generated html somewhere for you to see.

Yes, usually https://github.com/w3c/spec-prod is used to publish a preview but looks like it is not set up in this repo.

No but I'm wondering if it's a good fit. setViewport command only allows to set the width/height and devicePixelRatio but do not seem to include any other viewport properties. If we decide to add a parameter in setViewport then where would all the explanations and diagrams go?

I assume it works the same way (it is also Bikeshed-based), in a non-normative section, with files being put to a folder. Note that I am not saying it is required.

I don't understand your question. Setting [=[[DisplayFeaturesOverride]]=] is the very purpose of this. When the override is set then the segments property is recalculated using the array of display features from DisplayFeaturesOverride slot coming from the setdisplayfeatures command.

Probably I made an assumption that is not correct. If this command is only meant to override the segments getter available to the page, the current steps look sufficient. Or does the command imply a way to emulate the viewport with segments visually? Is the command expected to also trigger relevant Screen Orientation, Window Resize or Posture change events mentioned in the spec or any other events? And to confirm the DOM attribute cannot be overriden from JavaScript for testing?

[darktears]

Yes, usually https://github.com/w3c/spec-prod is used to publish a preview but looks like it is not set up in this repo.

Yes, I'm not entirely sure why. I'm not really a regular contributor to the CSS WG. I just happen to work on adding the viewport segments functionality.

I assume it works the same way (it is also Bikeshed-based), in a non-normative section, with files being put to a folder. Note that I am not saying it is required.

I meant more the fact that the BIDI spec doesn't really cover the concepts but just defines the commands. For me it felt natural to have the Automation story sit next to the property it covers. Also the viewport segments concept goes hand in hand with the Device Posture API and we have dedicated commands for the Device Posture API. I don't have a strong opinion.

If this command is only meant to override the segments getter available to the page, the current steps look sufficient.

This is correct. Please note that the entire viewport segments concept also defines MQs and env variables which will also be invalidated and recalculated. After this patch lands, I was planning to reference this addition on the MQ spec and CSS variable spec to cover the [DisplayFeaturesOverride] case.

Or does the command imply a way to emulate the viewport with segments visually? Is the command expected to also trigger relevant Screen Orientation, Window Resize or Posture change events mentioned in the spec or any other events?

Screen orientation, Window Resize and posture change are not expected to change when the internal slot is set.

And to confirm the DOM attribute cannot be overriden from JavaScript for testing?

The segments property is a read only frozen array so no (unless I'm mistaken), MQs can't be overridden but CSS env variables could I believe. Besides the automation aspect for authors to try and automate testing of their websites on different foldable devices the intention of this patch is also for testing purposes for user agents. If we can override the display features coming from the Windows and Android native APIs we can also verify the calculation of the segments property, the CSS MQs and the CSS Env variables through various configurations (more than one fold for example). This will be very useful for conformance.

[OrKoN]

That sounds good.The PR looks good to me % minor comments.
Do you plan to add WPTs for these WebDriver methods in https://github.com/web-platform-tests/wpt/tree/master/webdriver/tests/classic?

[OrKoN]

LGTM

[darktears]

That sounds good.The PR looks good to me % minor comments. Do you plan to add WPTs for these WebDriver methods in https://github.com/web-platform-tests/wpt/tree/master/webdriver/tests/classic?

I intend to do something similar to https://github.com/web-platform-tests/wpt/tree/master/device-posture and add tests to verify the segments property, CSS MQs and environment variables.

[OrKoN]

It would be also nice to add WebDriver-specific tests to test error conditions, wrong input, some happy path cases. Some browsers might not actually implement testdriver.js using WebDriver so it is also a good indication for external users for the level of support in the corresponding driver.

[darktears]

It would be also nice to add WebDriver-specific tests to test error conditions, wrong input, some happy path cases. Some browsers might not actually implement testdriver.js using WebDriver so it is also a good indication for external users for the level of support in the corresponding driver.

I can investigate that.

[darktears]

I will still need a reviewer like @emilio to approve and merge this change. Thank you!

[emilio]

Is there a precedent to put these kind of spec text inside the CSS spec? I haven't seen it before... cc @fantasai @tabatkins

[OrKoN]

I have not seen a precedent with CSS specs but here other specs defining WebDriver extension commands https://dontcallmedom.github.io/webdex/e.html#extension%20commands%40%40webdriver%25%25dfn (not sure if the tool omits some specs)

[darktears]

WebAuth, Device Posture, Compute Pressure, Generic Sensors, Web Bluetooth for example define the automation/testing story in their respective specs.

Some specs like WebUSB are defining them into a dedicated spec, some specs into an explainer like WebXR.

So, I don't think there is a W3C guidance around that.