fix(ci): missing v12 and gx15 simulator targets

[pfeerick]

Summary

The WASM simulator target lists had drifted from the supported radio set (fw.json / build_fw.yml, 44 targets) in both directions. Most visibly, Companion ships without a V12 simulator.

Follow-up to #7114 (HelloRadioSky V12) and #7643 (Radiomaster GX15) — both added the radio to the firmware build, but each missed one of the two simulator target lists.

Target	companion.yml matrix	build-wasm-modules.sh default
v12	❌ missing	✅	#7114 updated build_fw.yml only
gx15	✅	❌ missing	#7643 missed the local build list
mt12	⚠️ listed twice (bw-1 + bw-2)	✅	built twice per CI run
x9e-hall	⚠️ built, unusable	❌	see below

x9e-hall is not a Companion firmware — Hall sticks is a build option of x9e (horussticks), so getSimulatorId() never returns it, and findSimulatorByName() only truncates x9e-hall down to x9e. The module could never be selected, so it is dropped from both lists rather than added to the second. It remains a firmware target in fw.json / build_fw.yml.

The two lists now hold the same 43 targets, which is every build_fw.yml target except x9e-hall.

Web simulator radio list

web/public/radios.json had drifted the same way, missing c14, gx15, t22, v12 and v14lcd.

Regenerating it surfaced a bug in gen-radios-json.js: it read disp.w / disp.h / disp.depth, but the hw_defs use lcd_w / lcd_h / lcd_depth, so every entry silently fell back to the 480x272x16 default. The committed file has correct values, so this only bites on the next regeneration — it would have flattened the display size of all 30 mono and non-480x272 radios in one go.

With the field mapping fixed, regeneration is a no-op for existing entries: the only changes are the 5 added radios, plus tpros/tprov2 moving to follow current fw.json ordering.

Keeping it in sync

radios.json is a generated file that was drifting unnoticed, so it now gets the same treatment as the other generated files:

• just gen-radios (and docker-gen-radios), added to the codegen / docker-codegen aggregates
• a Codegen drift job, triggered by fw.json, hw_defs/** and the script itself
• a README note that fw.json — not the hw_defs — decides which radios are included, and that the x9e-hall skip warning is expected

Testing

• just gen-radios verified idempotent; existing entries byte-identical after the display fix
• just --list / just --show check the new recipes resolve, matching what the drift workflow does
• Not verified locally: the v12 WASM module has never been built in CI, so this PR's Companion run is the first real test of it

🤖 Generated with Claude Code