Subj : src/doors/termgfx/focus.c focus.h src/doors/termgfx/test/test_focus.cs
To   : Git commit to main/sbbs/master
From : Rob Swindell (on Debian Linux)
Date : Mon Aug 10 2026 21:30:27

https://gitlab.synchro.net/main/sbbs/-/commit/61e0fc1b44d9b0fe71ec050f
Added Files:
	src/doors/termgfx/focus.c focus.h src/doors/termgfx/test/test_focus.c
Modified Files:
	src/doors/termgfx/CMakeLists.txt src/doors/termgfx/test/CMakeLists.txt
Log Message:
termgfx: track terminal focus, so an unfocused window stops turning

A terminal reports pointer motion whether or not its window has focus.
A door reading motion as look-around therefore turns the view while the
pointer is being moved for some other window, and the player comes back
facing somewhere new.

DECSET 1004 is the only way to know: the terminal sends CSI I on focus
gained and CSI O on focus lost. focus.c wraps the two escape strings and
the state machine, and offers a latched focus-out edge -- a release edge
that falls while the window is unfocused never arrives, so a door has to
let go of every held key when it hears about the gap, not when the next
report happens to land.

SyncTERM does not implement mode 1004 (src/conio/cterm.adoc), so nothing
here may depend on the events arriving. What is stored is therefore the
NEGATIVE, "focus has been lost", so that a zeroed struct means focused:
doors memset their terminal state, and a `focused` field would start
false and silently disable mouse look on every terminal that never
answers.

Only a bare, unparameterised final counts. Anything carrying parameters
is somebody else's sequence -- claiming CSI S for F4 turned SyncTERM's
XTSMGRAPHICS reply into a keypress that dropped the player back to the
BBS, and CSI I and CSI O sit one letter away from the same trap.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
--- SBBSecho 3.37-Linux
 * Origin: Vertrauen - [vert/cvs/bbs].synchro.net (1:103/705)

.