Five fixes from analysis of the latest user log:
A. HeadsetMicMute.build() now matches device.product_id as the hex
string ("0B18", "0B19") rather than ints. product_id is set as a
str by hidapi_impl.f"{pid:04X}", so the int comparison was always
False and the suppression never fired. The G522 was still being
asked to write mic-mute on each connect and erroring out 0x0A.
B. RGB-effects probe sub-device feature dump tolerates "unknown:HHHH"
string features. _format_feature now detects that shape explicitly
(and renders 0xHHHH from the suffix) instead of letting int(feat)
raise ValueError mid-iteration and abort the whole table dump.
C. HeadsetActiveEQPreset.write replaces the bare `_value = None` cache
invalidation with a synchronous read(cached=False) so _value is a
real dict before returning. Prevents a UI band-click crash with
'NoneType' object is not subscriptable when a user clicks an EQ
band before the panel re-reads after a preset switch.
D. New probe_advanced_eq_slots() iterates every advertised slot via
getCustomEQ at build time, logs which respond, and caches a list
of (slot, name, bands) on device._advanced_eq_working_slots. The
HeadsetActiveEQPreset selector builds choices only from working
slots and returns None from build() if ≤1 slot responds — G522's
firmware advertises 16 slots but only honors slot 0, so the user
no longer sees a 16-option dropdown they can't actually use.
HeadsetAdvancedEQ.build now reuses the same probe (cached) rather
than the old probe_all_presets path. The legacy alias is kept in
the hidpp20 facade for any external callers still on it.
E. (capture only — no parse change yet) get_advanced_eq_params and
get_advanced_eq_defaults now log raw=<hex> on success too. The
two functions decode the same slot to wildly different bands on
G522 (likely a different header size between getCustomEQ's reply
and getEQDefaults's reply); raw bytes will let us pin down the
exact framing difference and adjust the parser in a follow-up.
Fix path will land alongside docs/features.md notes documenting
the per-call format variation, since other 0x020D headsets may
diverge differently from G522.