[RayTrunk]

What changed
While Wisp hides the taskbars, each one now also gets an empty window region (SetWindowRgn). If Explorer shows a taskbar, it draws nothing. The region is set only once the taskbars are hidden, and removed before a peek from the chevron menu and on restore (also from the watchdog, which uses restore_snapshot).

Why
Switching between autohide and Keep the dock visible adds or removes the dock's app bar, which changes the work area. Explorer then shows its taskbars again, roughly 250–350 ms later and several times in a row. The show hook hides them, but only after they have drawn, which is the flash in #18. A direct test confirmed the trigger: SPI_SETWORKAREA with SPIF_SENDCHANGE makes Explorer show its taskbars immediately; without the broadcast it doesn't.

On the primary screen the shown taskbar is mostly the collapsed auto-hide strip, so the flash is most visible on secondary screens, where Shell_SecondaryTrayWnd sits fully on screen when shown.

Clipping a taskbar while it's still visible made Explorer show it again, so the region goes on only after the hide has been verified.

How I checked it
Windows 11, two screens (2560×1440 primary and 1920×1080 secondary, both at 125%), hide_taskbar = true.

• A script toggled dock_autohide 10 times (5 cycles) through WM_APP + 6 and captured the bottom strip of the secondary screen every ~12 ms. Frames that showed the taskbar: 56 on unmodified main (every switch), 0 with this change.
• Logged taskbar visibility and region over the switches: Explorer still shows the windows (IsWindowVisible toggles as before), but the region stays empty. Explorer sometimes drops the region on a secondary taskbar while it's hidden; the next reapply sets it again.
• Quit Wisp: the taskbar comes back with its full shape. With Wisp not running, clearing and setting the region on a visible taskbar hid and restored it cleanly.
• Startup: the first hide sometimes needs a few retries on my machine, but unmodified main does too, so it isn't caused by this change.
• cargo fmt --check and both clippy runs pass. On my machine (German Windows), cargo test fails commands::time::tests::finds_cities_in_the_windows_zone_list and commands::tests::finds_power_commands_time_and_settings; both fail the same way on unmodified main.

Not tested: Windows 10, virtual desktops, a peek from the chevron menu after the change.

Fixes #18

🤖 Generated with Claude Code

[SumirSeth]

Appreciate the PR. I have some local code for this fix, I'll see how our approaches overlap (or don't) and get this merged as appropriate.

[RayTrunk]

Correction pushed (6ba85ba). The first version assumed Explorer's taskbars have no window region of their own. They do: a simple region the size of the window (0,0,2560,60 and 0,0,1920,48 here). On restore, the first version set no region, which left the second screen's taskbar without one until Explorer laid it out again.

Now Bar keeps the taskbar's own region in the snapshot, taken before Wisp changes anything. A peek and restore_snapshot set it again, and the watchdog gets it through --bar.

Checked after the change:

• Normal quit and a forced kill (so the watchdog restores): both taskbars end with exactly their original region and are visible.
• 10 mode switches, captures of the second screen: 0 frames with the taskbar.

Startup. While checking this, I noticed the first hide at startup sometimes fails ("A taskbar remained visible after SW_HIDE"). The hider is then restored and retried, and the taskbar stays visible for 3.8–6.6 s. This also happens without this PR:

Build	Starts	First hide failed
v0.5.0 release	20	2
Local build of unmodified main	12	3
This PR	22	7
This PR with the clipping switched off	8	1

The samples are too small to say whether this PR changes the rate, so I'm not claiming it doesn't. I'll open a separate issue for the startup failure itself.