Fix screen curtain reporting a failure when the magnifier cannot be restarted, and reduce verbosity (#21099)
Fixes #20518
Fixes #19528
Summary of the issue:
When the Screen Curtain is disabled, it restarts the Magnifier if it was active before. If the
Magnifier then failed to determine what to follow — the log in #20518 shows a COMError from
the review position — the exception propagated out of ScreenCurtain.disable(). The Screen
Curtain script then reported "Could not disable screen curtain", although the curtain had been
disabled. The Magnifier was left marked active without actually magnifying.
Separately, toggling the Screen Curtain with the Magnifier active announced "Disabling magnifier"
and "Re-enabling magnifier", which #19528 found too verbose.
Description of user facing changes:
Disabling the Screen Curtain reports "Screen curtain disabled" whenever it actually is disabled.
The Magnifier is restarted even if it cannot immediately determine what to follow; it starts at
the centre of the screen.
If the Magnifier cannot be restarted, the reason is still reported.
Toggling the Screen Curtain no longer announces that the Magnifier is being disabled or
re-enabled.
Description of developer facing changes:
None.
Description of development approach:
Three commits, matching the analysis in #20518, plus the changelog:
Magnifier._startMagnifier handles failures to determine the tracked position. It catches
the same exception types that Stop review position errors from crashing the magnifier #20992 turns into LocationUnavailableError. Stop review position errors from crashing the magnifier #20992 absorbs the
Error when disabling screen curtain and reenabling Magnifier #20518 review position error inside the focus manager; this guard fixes it on master today and
remains a safety net afterwards. Programming errors still propagate.
ScreenCurtain.enable() and disable() isolate their Magnifier notifications. A Magnifier
failure is logged and does not affect the Screen Curtain's outcome; in particular it can no
longer prevent the screen from being hidden.
The "Disabling magnifier" and "Re-enabling magnifier" announcements are removed. The existing
MagnifierStartError messages are kept, so a failed restart is still reported with its reason.
No new strings.
Depends on #20992. This PR makes _startMagnifier() survive a failure to determine what to
track, but the update loop that runs immediately afterwards still reads the same position. On
master that loop counts a tracking failure as a magnification failure, so a persistent failure
can still stop the magnifier a few ticks after the restart. #20992 rewrites that loop and fixes
exactly this, so it is not duplicated here. The failure reported in #20518 is a transient one and
is fixed by this PR alone.
Compatible with #21034: its "Magnification API unavailable" case raises MagnifierStartError,
which this path already reports.