Skip to content

fullhan-opensdk: bump to 8f2079d (GC4653 day file aligned with the SDK's day tunings) - #2911

Merged
openipc-ai merged 1 commit into
masterfrom
fullhan-opensdk-gc4653-day-tuning
Nov 20, 2026
Merged

openipc-ai merged 1 commit into
masterfrom
fullhan-opensdk-gc4653-day-tuning

Conversation

@openipc-ai

@openipc-ai openipc-ai commented Nov 20, 2026 •

Copy link
Copy Markdown
Collaborator

Problem

The V100 GC4653 sensor library (#2879) uses the SDK's GC4633 files as its tuning, and the SDK ships the same GC4633 file as both its day and its night tuning. That file has 0x0040 in the 16-bit word at 0x78. The stock PB1's own GC4653 night tuning has the same value. Every day tuning in the SDK, and the stock PB1's GC4653 day tuning, has 0x0100 there. So the image's gc4653_mipi_attr.hex is a night tuning installed as the day one.

openfullhan 16d7282 sets that word to 0x0100 in the day file only; the night file keeps 0x0040. 8f2079d corrects the README there; see below.

Boards: fh8852v100_lite and fh8856v100_lite, the only ones that build fullhan-opensdk. This is untested on its own: it aligns the day file with the SDK's day tunings, and is not a fix for any symptom measured on hardware.

Hardware tested on

The board is fh8856v100_lite: the PB1 is an FH8856 V100 and runs that image. I have no V100 camera myself. The day file in this pin is the one a PB1 owner (FH8856 V100 + GC4653, board IF56N) has been running since #1610 comment 6969758974. The same sha256 is shown below for that file and for this pin. I cannot see that camera; the results below are quoted from the owner's reports.

What the hardware does not show is that this word, by itself, changes anything. The all-grey day picture on that camera turned out to come from majestic on every FH885x V100. A majestic test build fixed it, and that build ran with this day file. The old value has not been tried with the fixed majestic. 16d7282's commit message blames the old value for the grey picture. That is wrong, and 8f2079d corrects the README to say what is and is not known. The case for this bump is that the day file now matches every SDK day tuning and the stock camera's own day tuning, not a measured difference.

Evidence

Pinned tarball, checked against the file the owner installed:

$ curl -sL https://github.com/openipc/openfullhan/archive/8f2079d88ffd60a19c176a7ad1f13ad7477f4a1b.tar.gz | tar xz
$ sha256sum openfullhan-*/libraries/sensor/fh8852v100/galaxycore_gc4653/*.hex
2dd5a83efb0a13e647f151a4504af8b4b931a50c05dc7ab31badfced0e1d3f96  .../gc4653_mipi_attr.hex
b0ff6c82857b166874103eeec7c05c36834002d2ae02b268eaf3346c5b992136  .../gc4653_mipi_night_attr.hex

Before, with the current pin's day file and the current majestic: the owner's daylight snapshot had R = G = B in all 3,686,400 pixels (#1610).

After, with this day file (owner, #1610 comment 6972138565: Day file is 2dd5a...) and the majestic V100 test build (comment 7027991564): "Colour is quite good", and 2,745,630 of the 3,686,400 pixels of the 2560x1440 snapshot are not grey. Since majestic also changed between those runs, this does not isolate the word at 0x78; see above.

$ git diff --name-only origin/master | python3 .github/scripts/ci-matrix.py --stdin
ci-matrix: 2/132 boards (needs_build=True) --- narrowed to the affected boards
  fh8852v100_lite
  fh8856v100_lite

Scope

  • No kernel patches under general/package/all-patches/linux/ (those go to OpenIPC/linux)
  • No files specific to a single retail camera model (those go to OpenIPC/builder)
  • No probing or bring-up tooling (that goes to OpenIPC/ipctool)
  • Nothing under general/overlay/ or in a shared load_<vendor> script hardcodes a value specific to my board
  • Package sources come from an OpenIPC repository, and any version bump keeps at least the specificity of the pin it replaces (a new package should pin a full 40-character SHA)
  • No LD_PRELOAD, and no binaries that cannot be rebuilt from source
  • New code is selected by a defconfig, so CI actually builds it

@qodo-free-for-open-source-projects

qodo-free-for-open-source-projects Bot commented Nov 20, 2026 •

Copy link
Copy Markdown

PR Summary by Qodo

Pin openfullhan GC4653 day tuning to 8f2079d

⚙️ Configuration changes 🕐 10-20 Minutes

Grey Divider

AI Description

• Pin openfullhan to a revision whose GC4653 day tuning matches the SDK day value.
• Keep the night tuning unchanged for FH8852/FH8856 V100 lite builds.
• The tuning word’s effect on image colour remains unverified independently of majestic.
Merge Risk 🟡 Medium

Blast radius 🟢 Low 💥 1 capability

  • 1 affected end-to-end capability after fault-specific containment; implementation boundaries do not widen its functional scope.
  • The changed SDK revision is selected only for the FH8852V100 sensor family, built as the GC4653 sensor implementation, and consumed by Majestic through the installed sensor library and tuning paths.
  • These are producer, packaging, and consumer hops of one camera-capture behavior, not independent capabilities.
Sources

general/package/fullhan-opensdk/fullhan-opensdk.mk[18-20]

FULLHAN_OPENSDK_SENSORS_fh8852v100 = galaxycore_gc4653
FULLHAN_OPENSDK_SENSORS = $(FULLHAN_OPENSDK_SENSORS_$(OPENIPC_SOC_FAMILY))
FULLHAN_OPENSDK_SENSOR_DIR = $(@D)/libraries/sensor/$(OPENIPC_SOC_FAMILY)

general/package/fullhan-opensdk/fullhan-opensdk.mk[7-8]

FULLHAN_OPENSDK_SITE = $(call github,openipc,openfullhan,$(FULLHAN_OPENSDK_VERSION))
FULLHAN_OPENSDK_VERSION = 8f2079d88ffd60a19c176a7ad1f13ad7477f4a1b
Behavioral complexity 🟡 Medium 🧩 Compatibility cases

  • Before: the image consumed artifacts produced from the previous openfullhan revision.
  • After: the same build and runtime pipeline consumes artifacts from a different external revision.
  • New correctness condition: the new GC4653 library and tuning data must remain compatible with the fixed FH885x driver and Majestic's runtime loading/configuration contract.
  • The build loop, sensor selection, and installation mechanics are unchanged; the complexity comes from the changed external producer-consumer compatibility case, not from those existing loops.
Sources

general/package/fullhan-opensdk/fullhan-opensdk.mk[7-8]

FULLHAN_OPENSDK_SITE = $(call github,openipc,openfullhan,$(FULLHAN_OPENSDK_VERSION))
FULLHAN_OPENSDK_VERSION = 8f2079d88ffd60a19c176a7ad1f13ad7477f4a1b

general/package/fullhan-opensdk/fullhan-opensdk.mk[18-20]

FULLHAN_OPENSDK_SENSORS_fh8852v100 = galaxycore_gc4653
FULLHAN_OPENSDK_SENSORS = $(FULLHAN_OPENSDK_SENSORS_$(OPENIPC_SOC_FAMILY))
FULLHAN_OPENSDK_SENSOR_DIR = $(@D)/libraries/sensor/$(OPENIPC_SOC_FAMILY)
Potential damage 🟡 Medium 🩹 Temporary loss

  • A load or initialization incompatibility can prevent the camera pipeline from starting for the affected sensor.
  • Reverting the firmware/build pin and reflashing a compatible image can restore the code and artifacts, but a running device requires that deployment step; no automatic runtime repair or fallback to the previous SDK is shown.
  • No persistent user data or external transaction is implicated.
  • Recoverable by rebuilding with the prior or compatible SDK revision and redeploying/reflashing the image.
  • Reverting source alone does not repair an already flashed device.
Sources

general/package/fullhan-opensdk/fullhan-opensdk.mk[7-8]

FULLHAN_OPENSDK_SITE = $(call github,openipc,openfullhan,$(FULLHAN_OPENSDK_VERSION))
FULLHAN_OPENSDK_VERSION = 8f2079d88ffd60a19c176a7ad1f13ad7477f4a1b

general/package/fullhan-osdrv-fh8852v100/fullhan-osdrv-fh8852v100.mk[25-31]

	$(INSTALL) -m 755 -d $(TARGET_DIR)/usr/lib/sensors
	$(INSTALL) -m 644 -t $(TARGET_DIR)/usr/lib/sensors $(FULLHAN_OSDRV_FH8852V100_PKGDIR)/files/sensor/*.so

	# majestic reads tuning from /etc/sensors, as the V200 package installs
	# it; under /usr/lib/sensors/params no V100 sensor found its .hex.
	$(INSTALL) -m 755 -d $(TARGET_DIR)/etc/sensors
	$(INSTALL) -m 644 -t $(TARGET_DIR)/etc/sensors $(FULLHAN_OSDRV_FH8852V100_PKGDIR)/files/sensor/params/*
Assessed commit 90aca26
Diagram

graph TD
  Pin["Package pin"] --> Source["openfullhan source"] --> Build["Sensor package"] --> Day["Day tuning"] --> Majestic["majestic"]
  Build --> Night["Night tuning"] --> Majestic
Loading
High-Level Assessment

Keep the exact upstream SHA pin. It brings in the corrected day file through the existing source-build and installation path; patching the hex file locally would duplicate the upstream change without isolating its effect from majestic.

Files changed (1) +1 / -1

Other (1) +1 / -1
fullhan-opensdk.mkAdvance the openfullhan source pin for GC4653 day tuning +1/-1

Advance the openfullhan source pin for GC4653 day tuning

• Replaces the fullhan-opensdk commit SHA with 8f2079d. That upstream revision supplies a GC4653 day tuning file with 0x0100 at offset 0x78 while retaining 0x0040 in the night file; no local build or installation rules change.

general/package/fullhan-opensdk/fullhan-opensdk.mk

@qodo-free-for-open-source-projects

qodo-free-for-open-source-projects Bot commented Nov 20, 2026 •

Copy link
Copy Markdown

Code Review by Qodo

🐞 Bugs (1) 📘 Rule violations (1) 📎 Requirement gaps (0) 🎨 UX issues (0) 🔗 Cross-repo conflicts (0) 📜 Skill insights (0)

Grey Divider


Action required

1. A V100 camera day-tuning change ships without target-board tests 📘 Rule violation ≡ Correctness
Description
The bump of FULLHAN_OPENSDK_VERSION to 8f2079d at line 8 swaps the GC4653 day tuning file that
fh8852v100_lite and fh8856v100_lite install, reportedly changing the word at 0x78 from
0x0040 to 0x0100, but that word has not been shown to change anything on hardware. Every V100
camera built from these boards receives the new value on its next image build, yet the PR's only
hardware results come from a PB1 (FH8856 + GC4653, board IF56N) run under a different majestic
build, and the author admits "I have no V100 camera".
Code

general/package/fullhan-opensdk/fullhan-opensdk.mk[8]

+FULLHAN_OPENSDK_VERSION = 8f2079d88ffd60a19c176a7ad1f13ad7477f4a1b
Evidence
The package pin is the only change in the diff and is the sole channel by which the updated GC4653
day tuning file reaches both V100 lite boards. The PR description states that the author has no V100
camera, that the word 'by itself' changing anything is not shown, and that the old value has not
been retested with the fixed majestic. The cited before/after results come from a PB1 owner's camera
with a different majestic build, so no output from the boards this change targets supports the
claimed effect.

Rule 3: Hardware evidence is present and honest
general/package/fullhan-opensdk/fullhan-opensdk.mk[8-8]
general/package/fullhan-opensdk/fullhan-opensdk.mk[7-8]

Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The bump to `8f2079d` changes the GC4653 day tuning installed on `fh8852v100_lite` and `fh8856v100_lite`, so the bytes driving the sensor on those cameras change. The PR has no output from a V100 camera. The only evidence cited is from a PB1 (FH8856 + GC4653, board IF56N) run under a different majestic build, and the description concedes that the word at `0x78` has not been shown to change anything.
## Fix Focus Areas
- general/package/fullhan-opensdk/fullhan-opensdk.mk[7-8]
## Recommended Fix
Before merging, either:
(a) Capture before/after evidence on a V100 camera (fh8852v100_lite or fh8856v100_lite) running the old and new pin with the same majestic build, such as snapshots, dmesg, or sensor logs, and cite it in the PR description with the verification checkboxes ticked; or
(b) State plainly in the PR that the change is untested on the affected boards, and limit it to the documented rationale (matching the SDK day value) with the uncertainty stated. Alternatively, keep the old pin or carry the day file as a board overlay until the test is done.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools



Informational

2. Commit title credits the 0x78 word with a fix 🐞 Bug ⚙ Maintainability
Description
The commit title and PR title say the bump is 'GC4653 day tuning with the SDK's day value at 0x78',
which presents the tuning change as the correction. The PR description itself says the word has not
been shown to change the picture and that the earlier grey-picture attribution was wrong, so the
history will record an unverified cause.
Code

general/package/fullhan-opensdk/fullhan-opensdk.mk[8]

+FULLHAN_OPENSDK_VERSION = 8f2079d88ffd60a19c176a7ad1f13ad7477f4a1b
Evidence
The only change in this diff is the pin on line 8, so the commit message is the only place the cause
is named. The PR description's 'What the hardware does not show' section contradicts the framing in
the commit title, and a later reader using git log for the grey-picture history would take the 0x78
change as the cause.

general/package/fullhan-opensdk/fullhan-opensdk.mk[8-8]

Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution
Issue description
The commit title claims the bump fixes GC4653 day tuning at 0x78, but the PR description says that effect is not measured. Fix Focus Areas
- general/package/fullhan-opensdk/fullhan-opensdk.mk[8-8]
Recommended Fix
Retitle the commit to describe the change as an alignment of the day tuning file with the SDK day tuning, not as a fix for the grey picture, and keep the unverified status in the commit body.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools


Grey Divider

Tip of the day
💡 Did you know, you can turn on the rule miner and Qodo learns your standards from review history

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗

Grey Divider

Qodo Logo

…K's day tunings)

openfullhan 16d7282 gives the GC4653 day file the value every SDK day
tuning, and the stock PB1's GC4653 day tuning, has in the word at 0x78
(0x0100); the SDK's GC4633 file it was borrowed from has the night value
(0x0040). What the word controls has not been isolated on hardware: the
grey V100 day picture in #1610 was majestic's, and its fix ran together
with this file. 8f2079d corrects the openfullhan README to say so.
@openipc-ai
openipc-ai force-pushed the fullhan-opensdk-gc4653-day-tuning branch from a1fae5b to 90aca26 Compare November 20, 2026 03:50
@openipc-ai openipc-ai changed the title fullhan-opensdk: bump to 8f2079d (GC4653 day tuning with the SDK's day value at 0x78) fullhan-opensdk: bump to 8f2079d (GC4653 day file aligned with the SDK's day tunings) Nov 20, 2026
@openipc-ai

Copy link
Copy Markdown
Collaborator Author

Thanks. Both points are addressed:

  1. Hardware evidence. The PB1 is not another board: it is an FH8856 V100 running the fh8856v100_lite image, one of the two targets. It has been running exactly this day file (sha256 2dd5a83e…, which matches the pinned tarball). What is not isolated is the effect of the 0x78 word on its own, because the majestic fix ran in the same session. The description now leads with that ("untested on its own … not a fix for any symptom measured on hardware"), which is option (b). Keeping the old pin would leave a known night tuning installed as the day file, against every SDK day tuning and the stock camera's own, so the pin moves.
  2. The title presented the change as a fix. The commit and PR are now titled "GC4653 day file aligned with the SDK's day tunings". The commit body says the grey picture was majestic's and that the word's effect has not been isolated.

@openipc-ai
openipc-ai merged commit a0a5e6f into master Nov 20, 2026
34 checks passed
@openipc-ai
openipc-ai deleted the fullhan-opensdk-gc4653-day-tuning branch November 20, 2026 04:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant