v1.1.0-dev.90 — A confirm you can see, and the message dev.89 never showed

v1.1.0-dev.90 — A confirm you can see, and the message dev.89 never showed

Pre-release off the dev branch. Builds on v1.1.0-dev.89.

Restore Defaults Asked Behind The Window

dev.89 put a confirm on Restore defaults, which drops the display
settings saved with a profile and writes that to disk. The confirm rendered
behind the editor that opened it.

Both are dialogs at the same stacking level. The confirm is declared first
in the markup, which loses on source order. It was invisible and it still
took the keyboard. The next Enter or Space answered a question nobody could
see. The answer was yes.

The stylesheet has a note about this exact trap, written when the same
thing happened to the arrangement history dialog. That did not stop it
happening again, so a check now presses the button in a real window and
asks the document what is actually on top.

The Message dev.89 Was Named For

Four of the eight arrow-key combinations on a two-monitor desk do nothing.
That is Windows' rule: along the axis two displays already share an edge
on, one direction overlaps the neighbour and the other opens a gap. Windows
keeps neither.

dev.89 said it had made the editor explain that. It had not. The message
was written and then overwritten by the very next line. It never once
reached the screen. In one of the two directions you got the plan's own
warning instead, which said a monitor "could not stay where it was" when
nothing had moved at all. Worse than silence.

It reaches the screen now:

These displays sit side by side. Up and down are the only way they can
move without breaking contact, which Windows will not keep.

And A Correction

dev.89's notes said six of eight combinations did nothing and that the
primary display was immovable in all four directions. Both wrong.

It was four of eight, and the primary has always moved. The measurement
behind that claim asked whether the nudged display's own coordinates
changed. Windows pins the primary to the origin, so for that one the answer
is no whether the desk moved or not. Asked properly, the two releases
behave identically: 10,992 combinations compared across desks of two to
four displays, zero differing outcomes.

The code that claimed to fix it did nothing, and is gone. The test written
to protect it passed against the release it was supposed to be protecting
against, which is the tell that should have been caught first. It asks
whether the arrangement changed now. It fails when the anchoring it
depends on breaks.

Smaller

Pressing ? in a dialog other than the display editor still left the
keyboard stranded. The fallback focused an element that cannot take focus.

Every checkbox and radio in the app sits on the spacing scale now, not just
the one that happened to get measured. Sweeping the Settings tabs the
spacing check had never opened found two more.

That check only ever looked at the tab Settings opens on. It walks all
seven now.

Install

Auto-updates from dev.53 and later. Otherwise unzip and run.