EOS R: HDR exposure bracketing + manual shutter control

[jed2nd]

What this does

Enables working exposure (HDR) bracketing on the EOS R (FW 1.8.0), plus the shutter-control and capture pieces it needs.

The core fix: PROP_SHUTTER encoding

On the R, PROP_SHUTTER is a 2-byte value whose high byte is the shutter as a signed Tv in 1/3-stop units (0 = 1", +3 = 0.5", -3 = 2", verified on camera). ML had been writing the value in the low byte, so the camera always saw Tv 0 and every frame came out 1" — which is why bracketing produced no exposure variation. This converts at the property boundary (write and read), leaving ML's internal raw scale unchanged.

Included

• lens.c — hi-byte-Tv encode/decode; CONFIG_R capture via the CameraConductor remote-release event with an early return so bracket frames sequence cleanly; coerce out-of-range shutter writes to the camera limit.
• shoot.c — detect the user's shot via the card file_number (the R's job_state notifications are unreliable) to fire the bracket; read the M-mode dial via a brief half-press so the bracket centers on the user's shutter; drop the MENU_MODE cancel check on R; auto frame count → fixed 5-frame symmetric bracket.
• features.h / property_whitelist.h — enable HDR bracketing + intervalometer, allow PROP_SHUTTER.

Comparison with #279, and why this one

#279 fixes the single-capture "saving…" hang on the R using the same CameraConductor remote-release event. This PR carries an equivalent capture fix and goes further:

	#279	This PR
Single-capture saving-hang fix	✅ remote-release	✅ same remote-release event
Multi-frame sequencing	falls through to the generic post-capture waits	adds a settle + early return that skips those waits — without it the R stalls after frame 1, so a bracket/intervalometer can't fire a sequence
PROP_SHUTTER encoding fix	—	✅ hi-byte Tv (without it every frame is 1")
HDR exposure bracketing	—	✅
Dial-centered bracket + intervalometer	—	✅
Hardware test coverage	single capture	full 11-point matrix (comment below)

So this is a superset: an equivalent remote-release capture fix, plus the early-return refinement and the shutter-encoding fix that multi-frame bracketing requires. It can replace #279, or be rebased onto it if you'd rather land #279 first — happy to do whichever.

(Minor: #279 wires the remote-release as a named stub SetEventIrRemoteReleaseBtn; this PR currently calls it by address 0xE0190215 with a comment — trivial to switch to the named stub if you prefer.)

Testing

Validated on an EOS R (FW 1.8.0) — full functional matrix in the comment below (single capture, 5-frame HDR bracket centered on the dial with varying exposures, intervalometer, clean power-off).

Known limitations

• The R doesn't report the shutter dial to ML at idle, so the bracket reads it via a brief half-press at the start (a short pause before the first frame).
• Bracketing from a very fast base (e.g. 1/2000) clamps the fast-end frames at 1/8000 (a physical limit), handled so no frame is wasted.
• Auto frame-count is a fixed 5-frame bracket on the R (the histogram auto-metering path is unreliable on this body).

[jed2nd]

Test matrix — EOS R, firmware 1.8.0 (7.3.9)

All run on hardware. Setup: M mode, manual focus, scene with tonal range.

#	Feature	Expected	Result
1	Single photo + clean finalize	Saved; no "saving…" hang at power-off	✅ PASS
2	Shutter value applies	EXIF matches the set shutter	✅ PASS
3	HDR bracket — centered, 1/125 base, 5 frames	5 images stepping dark→bright, centered on 1/125	✅ PASS
4	HDR bracket — 3 frames	3 images	✅ PASS
5	HDR bracket — fast base (1/2000) clamp	Fast frames clamp at 1/8000, no wasted duplicate	✅ PASS — two frames landed at 1/8000 (−2EV and −4EV both exceed the limit; expected)
6	HDR bracket — centers on the dial (1/500)	Centers on 1/500, not a fixed default	✅ PASS
7	HDR bracket — reliability ×3	Full set each time, clean power-off	✅ PASS
8	Exposure ordering	All exposures present and correct	✅ PASS — fired in ML's sequence order (center, −, +, −−, ++), not monotonic brightness; HDR tools sort by EXIF
9	Intervalometer	~5 frames spaced ~3 s, Leave-Menu trigger	✅ PASS — 4 frames, triggered on menu leave
10	Single photo after bracketing	Normal capture still works	✅ PASS
11	Clean shutdown	No saving hang	✅ PASS

Net: single capture, correctly-centered 5-frame HDR bracket with varying exposures (verified across the 1/8000–30" range), intervalometer, and clean power-off all working.

[jed2nd]

Re-ran a smoke test on the final stripped branch build (the exact PR content, with the dev probes removed): single capture + clean power-off, 5-frame HDR bracket (centered on the dial, exposures stepping dark→bright), and intervalometer all confirmed working on the EOS R. No behavioral change from the matrix above.

[jed2nd]

@kitor @reticulatedpines

Prefer this PR over #279

[jed2nd]

Pushed a follow-up to reduce the R special-casing:

• The shutter-encoding and HDR-bracket adaptations are now gated on semantic capability flags (CONFIG_PROP_SHUTTER_HI_BYTE, CONFIG_HDR_BRACKET_VIA_FILENUMBER) defined in platform/R.180/internals.h, instead of scattered CONFIG_R. CONFIG_R now appears only in the capture #elif (the existing per-body release chain alongside CONFIG_5D2/CONFIG_40D).
• Removed a leftover BRKCANCEL.TXT debug log that had slipped through.
• Restored the shared prop_set_rawshutter length to 4 (the R takes its own early-return path, so that change to the generic path was unintended).
• Replaced the raw 0xE0190215 address with a named SetEventIrRemoteReleaseBtn stub.

Behavior-preserving for the R (the flags compile the same blocks); builds clean under -Werror.

[jed2nd]

Expanded this PR to cover all three manual exposure controls on the R, not just shutter — ISO and aperture read/write now work too.

The interesting part: the R encodes all three exposure properties the same unusual way. The value sits in a high byte as an offset code at 3 units per stop, rather than ML's usual low-byte APEX raw — which is why naïve writes silently did nothing (ISO/aperture stuck at one value) or came out as a fixed 1" exposure (shutter):

Property	Length	Code location	Encoding
PROP_SHUTTER	2 bytes	byte 1	signed Tv·3, 0 = 1"
PROP_ISO	4 bytes	byte 1	15 = ISO 100, +3/stop
PROP_APERTURE	2 bytes	byte 1	9 = f/2.8, +3/stop

ML keeps its own APEX raw scale internally and converts only at the property boundary ((ml_raw − offset)·3/8 → high byte; the inverse on read). Each is behind its own capability flag — CONFIG_PROP_SHUTTER_HI_BYTE, CONFIG_PROP_ISO_BYTE1, CONFIG_PROP_APERTURE_BYTE1 — so nothing changes for other bodies. ISO and aperture also needed un-gating in property_whitelist.h: ISO was defensively deny-listed (a runtime slave proved the handler is safe on the R), and both were added to prop_write_allow. Aperture deliberately skips the usual min/max COERCE because the R doesn't report a lens range to ML (raw_aperture_min/max stay 0) — the camera clamps to the mounted lens itself.

Validated on an EOS R (FW 1.8.0): all three track the dial correctly and set correctly through the production lens_set_raw* path, with exact read-back (e.g. ISO set 100/400/800/3200 and aperture set f/4·f/8·f/11 both round-trip exactly).