Differential Revision: https://phabricator.services.mozilla.com/D179180

Differential Revision: https://phabricator.services.mozilla.com/D179300

This required exposing GetCaretRect via XPCOM.

Differential Revision: https://phabricator.services.mozilla.com/D179346

Bug 1828295 part 2: When setting the fake a11y Windows caret, convert the screen coordinates to client coordinates. r=nlapre

Previously, we were making the screen coordinates relative to the window rect.
However, the window rect includes the non-client area of the window, which includes the title bar.
SetCaretPos expects client coordinates.
In Firefox, this normally doesn't matter because the title bar is disabled, so there is no non-client area.
However, when the title bar is disabled, the coordinates were previously incorrect.

Differential Revision: https://phabricator.services.mozilla.com/D176620

I don't quite know which part of this was breaking, but the coordinates were quite wrong in the parent process.
They were wrong enough that when we adjusted for the character height, we ended up with a negative height in the resulting rect.
This code is now more similar to how we do this in LocalAccessible::BoundsInAppUnits.
This should also take CSS transforms into account in most cases where we weren't previously, though perhaps not for some iframe transforms.

Differential Revision: https://phabricator.services.mozilla.com/D176619

Backed out changeset 52b9bf7e6344 (bug 1828295)
Backed out changeset 6e6ee5fe337f (bug 1828295)
Backed out changeset 4e8ff9c439ef (bug 1828295)

This required exposing GetCaretRect via XPCOM.

Differential Revision: https://phabricator.services.mozilla.com/D179346

Bug 1828295 part 2: When setting the fake a11y Windows caret, convert the screen coordinates to client coordinates. r=nlapre

Previously, we were making the screen coordinates relative to the window rect.
However, the window rect includes the non-client area of the window, which includes the title bar.
SetCaretPos expects client coordinates.
In Firefox, this normally doesn't matter because the title bar is disabled, so there is no non-client area.
However, when the title bar is disabled, the coordinates were previously incorrect.

Differential Revision: https://phabricator.services.mozilla.com/D176620

I don't quite know which part of this was breaking, but the coordinates were quite wrong in the parent process.
They were wrong enough that when we adjusted for the character height, we ended up with a negative height in the resulting rect.
This code is now more similar to how we do this in LocalAccessible::BoundsInAppUnits.
This should also take CSS transforms into account in most cases where we weren't previously, though perhaps not for some iframe transforms.

Differential Revision: https://phabricator.services.mozilla.com/D176619

This ensures that these are HyperTextAccessibles if they contain text, which prevents assertions and exposes formatting information.
It also gives these the correct role.

Differential Revision: https://phabricator.services.mozilla.com/D179409

Differential Revision: https://phabricator.services.mozilla.com/D179404

Differential Revision: https://phabricator.services.mozilla.com/D178578

Bug 1833313: Remove all remaining a11y cache checks and non-caching code except for Mac specific bits. r=morgan,geckoview-reviewers,calu

Differential Revision: https://phabricator.services.mozilla.com/D178146

Backed out changeset f688d9dff006 (bug 1835194) for causing mochitest failures in accessible/tests/browser/telemetry/browser_HCM_telemetry.js

Differential Revision: https://phabricator.services.mozilla.com/D179180

Bug 1826869 - Cache doc mime type and make available via DocAccessibleParent. r=morgan,ipc-reviewers,mccr8

Differential Revision: https://phabricator.services.mozilla.com/D178887

Differential Revision: https://phabricator.services.mozilla.com/D178886

We had some different results in non-caching mode. We can now unify the results and not worry about that.

Depends on D178718

Differential Revision: https://phabricator.services.mozilla.com/D178719

Bug 1821965 - P2: Remove sync IPC extension messages for text operations. r=morgan,ipc-reviewers,mccr8

We can get rid the entire local HyperTextAccessible-based implementation as well!

Depends on D178717

Differential Revision: https://phabricator.services.mozilla.com/D178718

With caching on we can remove the legacy text marker and the abstracting class that allowed us to operate in both modes.

Depends on D178724

Differential Revision: https://phabricator.services.mozilla.com/D178717

Depends on D178715

Differential Revision: https://phabricator.services.mozilla.com/D178724

Depends on D178714

Differential Revision: https://phabricator.services.mozilla.com/D178715

Differential Revision: https://phabricator.services.mozilla.com/D178714

clang is warning that C++20 expects comparison operators to be commutative: `a == b` and `b == a` should resolve to the same comparison operator function. Warnings about the comparison of const and non-const objects can be fixed by making the comparison operator function const.

accessible/base/TextAttrs.h:139:54 [-Wambiguous-reversed-operator] ISO C++20 considers use of overloaded operator '!=' (with operand types 'mozilla::a11y::TextAttrsMgr::TextDecorValue' and 'mozilla::a11y::TextAttrsMgr::TextDecorValue') to be ambiguous despite there being a unique best viable function with non-reversed arguments
...

Differential Revision: https://phabricator.services.mozilla.com/D179019

Backed out changeset 6df0877f288a (bug 1821965)
Backed out changeset e663ccc0334e (bug 1821965)
Backed out changeset 36b250225e50 (bug 1821965)
Backed out changeset 6df5d07bf19f (bug 1832353)
Backed out changeset 409966c76f0f (bug 1832353)
Backed out changeset 3be446e44436 (bug 1832353)

Differential Revision: https://phabricator.services.mozilla.com/D178578

We had some different results in non-caching mode. We can now unify the results and not worry about that.

Depends on D178718

Differential Revision: https://phabricator.services.mozilla.com/D178719

Bug 1821965 - P2: Remove sync IPC extension messages for text operations. r=morgan,ipc-reviewers,mccr8

We can get rid the entire local HyperTextAccessible-based implementation as well!

Depends on D178717

Differential Revision: https://phabricator.services.mozilla.com/D178718

With caching on we can remove the legacy text marker and the abstracting class that allowed us to operate in both modes.

Differential Revision: https://phabricator.services.mozilla.com/D178717

Differential Revision: https://phabricator.services.mozilla.com/D178724

Differential Revision: https://phabricator.services.mozilla.com/D178715

Differential Revision: https://phabricator.services.mozilla.com/D178714

Bug 1834910 - Update accessible/other/Platform.cpp to match current accessible/base/Platform.h. r=Jamie

Differential Revision: https://phabricator.services.mozilla.com/D179000

Bug 1834404: Pre-allocate the HyperText offsets cache array to minimize re-allocations and thus improve performance. r=nlapre

Differential Revision: https://phabricator.services.mozilla.com/D178730

The core, cross-platform implementation already masks passwords.
Thus, the ATK specific code has been unnecessary for a long time now.
The ATK masking code also only applied to LocalAccessible.
This means we can get rid of this differentiation between local and remote.

Differential Revision: https://phabricator.services.mozilla.com/D178610

Bug 1834006 part 1: Remove several unnecessary differentiations between local and remote Accessibles in platform specific code. r=morgan

Differential Revision: https://phabricator.services.mozilla.com/D178609

Now that we always have the core parent process cache, this is no longer necessary.

Differential Revision: https://phabricator.services.mozilla.com/D178427

Now that we always have the parent process core cache enabled, this is always handled by the Pivot.

Differential Revision: https://phabricator.services.mozilla.com/D178426