merge_pr_58503

Use fixed viewport size for sticky elements stuck to the top level root scroll container.

Though this patch includes two wpt reftests, both are basically
identical. The only one difference is the one in mozilla directory runs
with dom.meta-viewport.enabled=true so that the test runs through the
mobile specific code path even on desktops.

Differential Revision: https://phabricator.services.mozilla.com/D286592

bugzilla-url: https://bugzilla.mozilla.org/show_bug.cgi?id=2005857
gecko-commit: 5795e0a4dad1d2a66885a711cd65257b524fc3e4
gecko-reviewers: botond

merge_pr_58502

Tweak inferred-mrow-stretchy.html and add separate tests for bug 2020658 and bug 236963.

The goal of inferred-mrow-stretchy.html is to verify that stretching
happens inside various elements using the mrow layout algorithm (either
as part of their own layout algorithm, or because they wrap their
children in an anonymous mrow), a concept called "inferred mrow" in
MathML 3. However, with how the test is currently written, some of these
elements are actually embellished operators and so what's actually tested
is the stretching during the layout of the element. Adding a
child ensures that these elements are not embellished and allows to
recover the original goal of this test. Incidentally, the two failures
in the .ini file are commented to mention the old bug 236963.

A separate test embellished-operator-004.html is added to preverse
existing testing of such embellished that is the sole child of a
element. We try various level of showing our existing
failures for 0 or 2 levels (respectively bug 236963 and bug 2020658).

Differential Revision: https://phabricator.services.mozilla.com/D287516

bugzilla-url: https://bugzilla.mozilla.org/show_bug.cgi?id=2023048
gecko-commit: 082e6f7f2319b41b018adbe6ee05f33874656a08
gecko-reviewers: emilio

merge_pr_58500

Move DidHistoryRestore check from GoToAnchor to its callers.

GoToAnchor is called twice for non-ASCII fragments (raw, then percent-decoded).
The DidHistoryRestore check inside GoToAnchor was consumed by the first failed
lookup, so the second call scrolled to the fragment instead of preserving the
restored scroll position. Move the check to Document::ScrollToRef(), matching
what nsDocShell::ScrollToAnchor already does.

Differential Revision: https://phabricator.services.mozilla.com/D287447

bugzilla-url: https://bugzilla.mozilla.org/show_bug.cgi?id=2020309
gecko-commit: 687e5c4d51698a4ba3b3349d4d895d1f17249624
gecko-reviewers: layout-reviewers, emilio

merge_pr_58493

Prompt API: Reject on invalid constraints and prefix conflicts

Surface kInvalidConstraint via new mojom Error enum.
Use this as the custom_reason for StreamingResponder disconnect.
Bubble cancellation up via Blink ModelStreamingResponseStatus.
Add unit tests and WPTs with basic Echo regex prefix checks.

Bug: 391919456, 434766400, 454698388, 488433627
Change-Id: Ie3a918c238ed02b40334a4228690d7031310cd23
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/7654881
Commit-Queue: Mike Wasserman [email protected]
Reviewed-by: Steven Holte [email protected]
Reviewed-by: Mustafa Emre Acer [email protected]
Auto-Submit: Mike Wasserman [email protected]
Cr-Commit-Position: refs/heads/main@{#1599563}

merge_pr_58469

Remove obsolete feature-policy sensor tests and rename to permissions-policy (#58469)

• Remove obsolete feature-policy tests and rename to permissions-policy

This PR cleans up remaining references to the deleted /feature-policy/
directory by:

1. Deleting obsolete sensor -feature-policy test files since their
   -permissions-policy equivalents already exist:

   • accelerometer/Accelerometer--by-feature-policy.html (5 files + 3 headers)
   • gyroscope/Gyroscope--by-feature-policy.html (5 files + 3 headers)
   • magnetometer/Magnetometer--by-feature-policy.html (5 files + 3 headers)

2. Renaming generic-sensor-feature-policy-test.sub.js to
   generic-sensor-permissions-policy-test.sub.js and updating all imports

3. Updating generic-sensor-tests.js:

   • featurePolicyNames -> permissionsPolicyNames
   • Updated test description text

4. Updating generic-sensor-helpers.js:

   • featurePolicyNames -> permissionsPolicyNames

5. Updating all sensor-data.js files:

   • accelerometer/resources/sensor-data.js
   • ambient-light/resources/sensor-data.js
   • gyroscope/resources/sensor-data.js
   • magnetometer/resources/sensor-data.js
   • orientation-sensor/resources/sensor-data.js

6. Renaming client-hints files:

   • client-hints/accept-ch-stickiness/cross-origin-subresource-with-feature-policy.https.html
     -> cross-origin-subresource-with-permissions-policy.https.html (+ resources)
   • client-hints/accept-ch/feature-policy-navigation/ -> permissions-policy-navigation/
   • client-hints/accept-ch/feature-policy.sub.https.html -> permissions-policy.sub.https.html

7. Renaming gamepad and permissions files:

   • gamepad/feature-policy-gamepad.html -> permissions-policy-gamepad.html
   • permissions/feature-policy-permissions-query.html -> permissions-policy-permissions-query.html
   • permissions/permissions-query-feature-policy-attribute.https.sub.html
     -> permissions-query-permissions-policy-attribute.https.sub.html

8. Deleting fenced-frame/client-hints-meta* files:
   These tests used which was deprecated.
   Permissions-Policy does NOT support http-equiv meta tags (only HTTP headers),
   so these tests cannot be migrated and are now obsolete.

This follows up on PR #58327 which deleted the /feature-policy/ directory.

• fixup

merge_pr_58498

Add WPT for selectstart event

The event should be fired when user extends a selection range only when
there is no selection range or the selection range is collapsed.

Additionally, if the event is canceled, the range should not be changed.

This patch just adds the tests, but does not fix the bugs.

https://w3c.github.io/selection-api/#selectstart-event

Differential Revision: https://phabricator.services.mozilla.com/D287524

bugzilla-url: https://bugzilla.mozilla.org/show_bug.cgi?id=2022751
gecko-commit: 71da8a2a9b0125c993f378d7847544a690daa596
gecko-reviewers: smaug

merge_pr_58495

WebKit export of https://bugs.webkit.org/show_bug.cgi?id=309941 (#58495)

merge_pr_58494

[Gap Decorations]: Suppress row gaps even when there is a spanning item

Per the resolution taken in csswg issue 13362 1, gaps should be
suppressed across fragment breaks, even if a spanning item is present.

In the case of a candidate for gap suppression with a spanning item, we
can't just naively take away the gap else we run the risk of taking a
chunk of the item as well. Instead, we expand the next row to fill the
space where the gap would have otherwise been if it wasn't suppressed,
maintaining a 0px gap.

Note: When track sizes or positions are altered because of suppression
or expansion, these changes are not passed from the parent grid to the
subgrid. This matches the current behavior with row expansion and
subgrid, and will be addressed when track size and position changes due
to fragmentation are propagated to subgrid, as tracked by
crbug.com/491881353.

Relevant wpts are updated and added.

Bug: 490022546
Change-Id: I3282e3d3d0b3afc2ff59b707c1e3334024999d1f
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/7644466
Reviewed-by: Alison Maher [email protected]
Commit-Queue: Sam Davis Omekara [email protected]
Cr-Commit-Position: refs/heads/main@{#1599428}

merge_pr_58022

Add wptrunner.products.Product.add_arguments(argparse.ArgumentParser)

merge_pr_58488

[Masonry] BuildGridSizingTree in Grid Lanes

Refactor GridLanesLayoutAlgorithm to use the sizing tree throughout.
This involved calling BuildGridSizingTree() to build the tree. This CL
also moves track sizing code around to methods similar to that of Grid
to split things out more by function and eliminate duplicate code around
the algorithm.

To get this working, we now store extra information (virtual items and
intrinsic track sizes) on the sizing tree and layout data to allow
information to be cached in BuildGridSizingTree() and to be reused in
later parts of the algorithm.

There are still some open TODOs to get certain pieces fully in place,
but this gets grid lanes working with the sizing tree with a single root
node.

This change should not have any behavioral impact. It is a refactor in
preparation for supporting subgrid in Grid Lanes.

All of the subgrid tests have been marked as crashing since some of the
code paths are lighting up without full support yet. These will be
investigated as further support is added for subgrid.

This CL also fixes a number of broken tests from a wpt import. These
failures weren't related to this CL, but ensure this CL is run with all
known passing tests previous to this change.

Bug: 343257585, 492268944
Change-Id: Icb12bc1144a58d3632a8fc10f259a3c34272d8b4
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/7643400
Reviewed-by: Kurt Catti-Schmidt [email protected]
Commit-Queue: Alison Maher [email protected]
Cr-Commit-Position: refs/heads/main@{#1599235}