[bryanjeal]

For #1537 and #1547. Overlaps a little of #1459: it was written independently from that branch, and if #1459 takes this branch's test row or this takes #1459's regex line, either merge drops cleanly.

What and why

The launcher's Game's upscaler choice DM_UPSCALER=metalfx lists only agl_ndsr.pck's compute shaders (UPSCALER_SOURCE in dm/games/game.py), but a game's upscaler network ships in agl_ndsr_vr.pck and agl_ndsr_vr_light.pck too: The Legend of Zelda: Ocarina of Time (Switch 2) upscales with agl_ndsr_vr.pck's fa3033d05c0a41d6 and friends, which the option never offered, so switching to MetalFX left the game's network running and told the player nothing about it (DM_GAME_UPSCALER_KEYS was empty on 0400486024DEC000).

The same shotgun #1459 applies on its branch. This does not move the launcher: against #1459's current head it shows no game.py diff beyond the branch's own two comment lines, and the test change is not there at all (git show pull/1459/head:dm/tools/tests/test_game_profile.py has no vr_light row).

Checked

• test_game_profile.py on the change above main 339d9f2: MAP gains one agl_ndsr_vr_light.pck row (the pack Ocarina of Time's FA3033D05C0A41D6 comes from), and DM_GAME_UPSCALER_KEYS for the fixture goes from 2 keys to 3 with DM_GAME_UPSCALER_N 3. The new row fails against main's /agl_ndsr\.pck# regex (fixture asserts 3 keys, gets 2) and passes with this branch's.
• test_game_profile 7 tests, test_dmsh_hmma 9, OK on 27.5.1 / M5 Max, Xcode 27.2.

Docs

• None changed. UPSCALER_SOURCE's comment in dm/games/game.py states where the option refuses a variant it cannot draw, nothing else.

[bryanjeal]

Test row now offered on #1459 where the regex line it depends on lives; closing this to keep the fix review on one branch. The Harness additions to test_dmsh_hmma.py are described on #1547.