[ccoventry]

Release note: Fixed: the crosshair border is now the same thickness on every side.

Closes #1348

Cause

The border in #1348's captures isn't drawn by ds_hud_crosshair_outline_size. It's the 1-texel black border in customxhair-diceWSOD.spr's own art. That sheet's tiles 7–12 (the ones the Players table uses) are exact mirror images, centred in their 64×64 tiles. So none of the difference comes from the art.

At the game's size, a tile's pixels mirror about the middle of pixel (1280, 720) (2560×1440). ds_hud_crosshair_scale zoomed about the corner of that pixel. That moved the crosshair's middle 0.5 × (S − 1) px down and right, onto neither a pixel's middle nor its edge. Then each edge's texel was sampled at a different phase, and the outer edges are hard-cut by the alpha test (the movie cfg's gl_spriteblend 0). At 1.25 that came out as 1.5 px on the top and left and 2.3 px on the bottom and right. Scale 2 happened to land on a pixel edge, which is why it looked right.

Fix

crosshair_scale::zoom_point picks the zoom point so that the crosshair's middle lands exactly on the centre corner (W/2, H/2) at any scale. One function changed; the drawn outline (#1338) uses the same zoom and needs nothing else.

Zooming about the pixel's middle also gives a mirror image, and I tried it first. But it left each arm's two ends at 1.5 and 2.4 px. Landing on the corner gives every edge 1.7–1.85 px. The creator's crosshairs are drawn about a texel corner, so on a pixel corner all the edges fall at the same phase.

Tested

In game, both installs, Brain, wsod25_grp4_h1_m00cat_dod_railroad2_test_1 (m00cat, tile 8 of customxhair-diceWSOD.spr), 2560×1440, captured through the hook's batch. The figures are border px per side, measured along the centre row, the centre column and across each arm; each is the mean, with the smallest and largest edge after it.

Install	Scale	gl_spriteblend	Before: top / left	Before: bottom / right	After: every side
POST	1.25	0 (the WSOD25 cfg's)	2.05 (1.51–2.39)	2.25	1.74 (1.68–1.85)
PRE	1.25	0	2.05 (1.51–2.39)	2.25	1.74 (1.68–1.85)
POST	1.5	0	2.25 (1.86–2.65)	2.45	2.45 (2.19–2.61)
PRE	1.5	0	2.25	2.45	2.45 (2.19–2.61)
POST	1.25	1	~1.21	1.32–1.43	1.21
PRE	1.25	1	~1.21	1.32–1.43	1.21
POST	1.25, outline 2	0	3.56	4.36	3.85 (every side)
POST / PRE	1	0 and 1	1.47 / 1.02, mirrored	same	unchanged (no zoom at 1)

With gl_spriteblend 1 the outer edge is blended instead of cut, so the border reads thinner (about 1.2 px, the texel's true 1.27) and the old shift was smaller (about 0.2 px instead of 0.8). The pink's inner edge blends into the black the same way at either value.

The red crosshair (cl_xhair_style 0, POST 1.25) moves by at most a pixel. Its arms used to sit 3 px further right than left; now the difference is 1 px. That piece's layout is DoD's own dynamic gap.

cargo test -p goldsrc-hooks crosshair_scale, rustfmt and clippy (host and i686, 1.98.1) are clean.

Frames, zoomed crops and the before/after image: D:\dodstudio-tests\crosshair-border-1348\.

🤖 Generated with Claude Code

[ccoventry]

Combined in-game test of #1350, #1351 and #1352

Built from dev f628957 with all three branches merged locally. They merge with no conflicts. DLL: local\test-combined\. Recorded at 2560×1440 with the user's dicewsod25.cfg on the Brain account, as an HLAE 300 fps batch capture.

• Hook tests: 1228 pass. Clippy and rustfmt on 1.98.1 are clean.
• Show scope: the scope draws on the first zoomed frame of every scope-in #1351 scope: passes on both installs. On POST, 10 of 10 scope-ins and unzooms land on the frame the zoom changes. On PRE, 11 of 11 do. Demo: wsod25_ply2_m1_h2_m00cat.
• Show scope: the first scoped frame can draw a tiny scope on a black screen #933 re-run: still passes on POST. The scope is full size, and the gun-lowered frames correctly show no scope.
• Crosshair border: same thickness on every side (#1348) #1350 crosshair: passes on POST. The border is 1.68 px on all four sides, on tile 8 at scale 1.25 with gl_spriteblend 0.
• Headshots can be heard further away: new Sound setting in game #1352 far headshot: passes on PRE at 1507 units. With the setting at 0 the hit is silent; at 1 it plays at −21.3 dB (with the cfg's mirv_movie_sound_volume 0.25).
  • Not tested: POST, and PRE's second moment. Steam switched off the Brain account partway through, so launching stopped there.
  • The console_commands_data.js hint for ds_sound_headshot_far is now regenerated on this branch.
• No regressions: 6 game sessions ran with no crash, and the hook log has no errors.

Evidence is in D:\dodstudio-tests\combined-1350-1352\.