[mryel00]

This fixes the logging of the str, rect and bitmask types in the v4l2 module.
The wrong logging can be seen here: #366 (comment)
The logging is expected to look like this now:

region_of_interest_rectangle 0x00981ae1 (rect)   : value=(0,0)/3844x2164 flags=has-payload, has-min-max
region_of_interest_auto_ctrls 0x00981ae2 (bitmask): max=0x00000001 default=0x00000001 value=0x00000001 flags=has-min-max

A side effect of this PR is a fix for the logging of min and max for the menu type, similar to #366, as they are needed for str and bitmask too.

[mryel00]

@8023 could you please test this PR and share the controls output of the log?

[Copilot]

🟡 Changes recommended

Critical Python compatibility and ctypes pointer issues remain unresolved.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

Fixes V4L2 logging for string, rectangle, bitmask, and menu controls.

Changes:

• Adds rectangle and extended-control payload support.
• Formats compound and hexadecimal control values.
• Generalizes UVC control metadata logging.

File summaries

File	Summary
crowsnest/v4l2/utils.py	Adds rectangle mapping and hex formatting.
crowsnest/v4l2/raw.py	Adds rectangle and extended-control definitions.
crowsnest/v4l2/ctl.py	Parses metadata and retrieves compound control values.
crowsnest/v4l2/constants.py	Defines the rectangle control type.
crowsnest/camera/types/uvc.py	Updates control logging and value typing.

Review details

Suppressed comments (3)

crowsnest/v4l2/ctl.py:269

• Formatting every non-payload value as a string changes the established numeric return type. _brokenfocus() compares this result with int(focus_absolute_conf) (components/streamer/ustreamer.py:142), so a configured focus_absolute is now treated as different on every startup and the corrective branch always runs. Preserve integers for scalar controls and use a mixed return annotation for payload controls.

            return f"{ctrl.value}"

crowsnest/v4l2/ctl.py:246

• ioctl_safe() catches OSError and returns -1; it never raises into this try. Since this return value is ignored, a failed VIDIOC_G_EXT_CTRLS leaves the zero-initialized payload buffer and the code logs an empty or zero rectangle/string as if it were the current value. Check the ioctl result and return None on -1 before decoding.

            utils.ioctl_safe(fd, raw.VIDIOC_G_EXT_CTRLS, ctrls)

crowsnest/v4l2/ctl.py:50

• A string control's default_value is not a printable/default setting; the V4L2 control logger prints min, max, and step for strings and omits default. With this addition every string is logged as default=0 while step is missing, which makes the new string logging inaccurate. Add STRING to the step condition instead and keep it out of this default whitelist.

        constants.V4L2_CTRL_TYPE_STRING,

• Files reviewed: 5/5 changed files
• Comments generated: 2
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[8023]

Thanks for following up on this properly, instead of leaving the guard from #366 as the fix. Heads up: I used AI to help me write this comment.

Tested on the same machine as #366 . The full log in attached. The rect and bitmask controls now match v4l2-ctl.

A small difference: v4l2-ctl on this machine prints the bitmask value as decimal, only max and default are hex:

region_of_interest_auto_ctrls 0x00981ae2 (bitmask): max=0x00000001 default=0x00000001 value=1 flags=has-min-max

Regression: brokenfocus now triggers on every start

With v4l2ctl: focus_absolute=200,... in the config I get this on every start:

focus_absolute (int)               : min=0 max=800 step=1 default=150 value=200 flags=has-min-max
WARN: [cam 1]: Detected 'brokenfocus' device.
INFO: [cam 1]: Try to set to configured Value.

The value is already 200. I had several earlier starts with the same config and no warning. The cause is that get_control_cur_value_with_qc() now returns f"{ctrl.value}", while _brokenfocus() still compares the result against int(focus_absolute_conf). "200" != 200 is always true.

Suggestion: keep returning int for scalar controls and do the string formatting in get_controls_string(). The same goes for the hex strings stored in parse_qc. That way callers always get raw values back.

Two more things from reading the code. My camera has no controls of these types, so I couldn't test either:

• ioctl_safe() swallows the OSError, and its return value is ignored here. The except OSError only covers os.open. If VIDIOC_G_EXT_CTRLS fails, we log a zeroed buffer like (0,0)/0x0 as if it were the real value. The same happens for write-only controls such as UVC pan_reset/tilt_reset: G_CTRL returns EACCES, and they get logged as value=0.
• For str controls, v4l2-ctl prints min max step and no default. This PR adds default (always 0) and leaves out step.

3dprinter-60cm_crowsnest.txt

[mryel00]

Thanks for the small review and the output. Seems like I have to fix the intendation (currently just a magic number 😂)

v4l2-ctl on this machine prints the bitmask value as decimal, only max and default are hex

I saw this in your output too. But if you look at the v4l2-ctl code, it should print it as hex too, at least in the latest version:
https://git.linuxtv.org/v4l-utils.git/tree/utils/v4l2-ctl/v4l2-ctl-common.cpp#n644
Therefore I just copied the latest behavior for this.

Regression: brokenfocus now triggers on every start

Thanks for pointing out the regression. I already thought that I introduced a bug, but I couldn't find it 😅

For str controls, v4l2-ctl prints min max step and no default. This PR adds default (always 0) and leaves out step.

Good catch ^^

ioctl_safe() swallows the OSError

This is true but this is nothing new introduced by this PR. This would need a separate fix.
But this is only a best effort implementation for v4l2. I do not intend to fully mimic v4l2-ctl. Just the basic things that most cameras got and maybe some exotic stuff that people report as a problem with their owned camera.

[mryel00]

The string and brokenfocus should be fixed now. Please test them again.
I decided against an integer for bitmask as it would result in more complex logging code without any real use case that I can think of for the integer value (at least within the scope of Crowsnest)

[8023]

Retested with c44d142. The config and v4l2-ctl output are the same as last time, so I'm not pasting them again. The startup log is attached: 3dprinter-60cm_crowsnest.txt

• brokenfocus: fixed. No warning on startup, and focus_absolute stays at 200.
• Alignment: fixed. All colons now line up, including region_of_interest_auto_ctrls (bitmask).
• str controls: not tested. My camera doesn't have any.

One small thing in get_controls_string(): max() has no default=0. If a UVC device reports no controls at all, control_values stays empty and max() raises ValueError: max() arg is an empty sequence. That's rare, but libcamera.py:72 already handles the empty case, so it would be nice to cover it here too:

max_len = max(
    (len(f"{c} ({d['type']})") for controls in self.control_values.values() for c, d in controls.items()),
    default=0,
)

Thanks for the quick fixes!

[mryel00]

hope you don't mind me sharing

Please keep off-topic stuff out of the issues. You can share timelapses and similar things in our Discord in the i-made-a-timelapse channel.

If a UVC device reports no controls at all

I'm not sure if there are any UVC devices without any controls at all, but it's a good point to add some fallback, even if it just occurs due to some weird communication error.

Thanks for the quick fixes!

Thanks for the quick review and testing!