an opinionated terminal multiplexer in C with programmable pane chrome https://git.sl0p.foo/slosh.git/atom/tests/test_mouse_pixels.py?h=master 2026-10-09T10:31:53Z 2026-10-09T10:31:53Z user user@clank 2026-10-09T10:24:53Z urn:sha1:d04228e9124eca3117843dfb5dfcf531d2107d07 No behaviour: argument wrapping in path_test, and one check() call in test_mouse_pixels that was over the line length. Both are what the pre-commit hook would have done anyway. 2026-09-29T12:25:34Z user user@clank 2026-09-29T12:25:34Z urn:sha1:3760c4e99aff06d09edb0575d7596d0607298e71 A pane is told how big a cell is twice -- in mode 2048's resize report and in the XTWINOPS answers -- and the mouse encoder was told something else: a 1x1 cell, so that cell coordinates and "pixels" were the same number. That held until a program turned on SGR-pixel mouse (mode 1016) by itself. Textual does, the moment in-band resize works, and then divides every coordinate it is sent by the pixels-per-cell we advertised: a click 30 cells across and 9 down arrived as (3, 0), so every click in a Textual app landed in the top-left corner. Hand the encoder the cell size the pane already has, and put the pointer in the middle of its cell. A size that contradicts what the same pane was told about itself is the lie, not the precision.