Depthaiv3 alpha15 by phodina · Pull Request #402019 · NixOS/nixpkgs

phodina · 2025-04-26T13:10:31Z

Add DepthAI package which is basically OpenCV running on Edge devices made by Luxonis.

Currently work in progress therefore marked as Draft, would like to decrease the amoutn of patches by upstreaming some of the changes like

cmake: Add option to enable Hunter to fetch data luxonis/depthai-core#1303

Package request

Github repo link

Upstreaming the flake environment

WIP Dependencies:

CC: @jayrahdevore @tmayoff @hacker1024 @Pandapip1

Things done

Add a 👍 reaction to pull requests you find important.

phodina · 2025-04-30T22:40:16Z

deptahai-core:

nixpkgs:

rtabmap: 0.21.4.1 -> 0.21.13 #403078

phodina · 2025-05-01T17:42:26Z

backward-cpp: use cmake as a build system #403366

The libnop and mp4v2 don't populate the lib/cmake so they are not found.

Porting basalt is still WIP, biggest issue is the large dependency on third party packages from ROS.

Pandapip1 · 2025-05-01T22:20:56Z

Porting basalt is still WIP, biggest issue is the large dependency on third party packages from ROS.

I was working on ROS stuff at #355629, but it stalled because it was noted that ROS is pretty actively developed. If you want I can restart the effort.

phodina · 2025-05-02T16:57:14Z

I was working on ROS stuff at #355629, but it stalled because it was noted that ROS is pretty actively developed. If you want I can restart the effort.

AFAIK basalt is not required. It's optional dependency, though would love to make it work. However, first I'd focus just on DAI and then on the ROS ecosystem and finding way to package basalt.

Would like to do one more pass of the DAI package and mark the optional stuff and split package into multiple outputs.

Also it might be work to make two definitions - one that does not have dependencies on GUI stuff (good for containers) and then full blown implementation.

phodina · 2025-05-03T11:00:52Z

FIY oakctl Not required as dependency but useful tool for OAK4 line

#403889

phodina · 2025-05-18T07:10:33Z

Rebased and removed the already merged commits for libnop and cpr

Should I also squash the commits for depthai-data and depthai-core`?

Pandapip1 · 2025-05-18T07:36:37Z

Should I also squash the commits for depthai-data and depthai-core`?

Nope! These are different packages and should keep their separate commits. The individual commits for each package probably should be squashed into one commit per package, but this is no means a requirement. See #347586

phodina · 2025-05-18T08:49:00Z

Nope! These are different packages and should keep their separate commits. The individual commits for each package probably should be squashed into one commit per package, but this is no means a requirement. See #347586

Agreed on that.

Well meant just per package as each has several changes so that there's just one depthai-core and another for depthai-data. IMHO I don't see benefit in keeping both the alpha14 and alpha15 versions as there are alpha.

phodina · 2025-05-18T09:07:12Z

There are additional points. IMHO I'd split the definition into multiple subpackages:

depthai-core-minimal (OpenCV without GTK support, optional CMake flags turned off)
depthai-core (OpenCV with GTK support, more CMake flags turned on)
depthai-core-full (turn on examples and tests, for C++ and Python)

Also there are tests but they require HW attached or available on either the USB or network. Therefore they can't be just enabled and executed as part of the derivation, right? Or is there as it's similar to GPU as they are also not available in the sandbox environment but do have test suites.

If so could set up hardware to run the tests to see if everything is working (I just use subset of the DAI features that in my case so far work).

phodina · 2025-05-22T07:54:59Z

So pushed changes tracking the v3_develop branch as there were many changes that allow more CMake config options and more offline build.

mweinelt · 2025-05-22T13:07:49Z

Looks like you grabbed an unrelated home-assistant commit when rebasing.

66c0c09

phodina · 2025-05-28T10:51:24Z

@Pandapip1 switched to v3.0.0-beta.1 and squashed the previous commits. Also put all the nixpkgs changes into single patch. Rest of the patches is either merged in the v3_develop branch or pending review.

phodina · 2025-05-31T10:21:40Z

There are additional points. IMHO I'd split the definition into multiple subpackages:

depthai-core-minimal (OpenCV without GTK support, optional CMake flags turned off)
depthai-core (OpenCV with GTK support, more CMake flags turned on)
depthai-core-full (turn on examples and tests, for C++ and Python)

Split the package into multiple based on enabled features.

Also there are tests but they require HW attached or available on either the USB or network. Therefore they can't be just enabled and executed as part of the derivation, right? Or is there as it's similar to GPU as they are also not available in the sandbox environment but do have test suites.

Also the nix package builds as a part of depthai-core repo Github CI. The tests are next step to implement.

Lastly I'd like to also split further the outputs - probably dev, out, python and share.

…ai-core-full

phodina · 2025-05-31T10:46:22Z

What needs to improve/minimize is the pkgs/by-name/de/depthai-core/0001-nixos-specific-patches.patch. The patched out lines are mostly due to VCPKG that has CMake configs for the dependencies but they are not present in Nix so needs revert to pkg-config.

phodina · 2025-05-31T10:50:37Z

Any idea how to handle the check if the package is split?

- Because pkgs/by-name/de/depthai-core exists, the attribute `pkgs.depthai-core` must be defined like

    depthai-core = callPackage ./../by-name/de/depthai-core/package.nix { /* ... */ };

  However, in this PR, it isn't defined that way. See the definition in pkgs/top-level/all-packages.nix:363

    depthai-core

phodina · 2025-07-10T11:46:25Z

@Pandapip1 could I ask you for in-depth review?

Pandapip1

First pass. Please excuse any typos, this was written on my phone over a train ride.

Pandapip1 · 2025-07-10T12:34:29Z

@@ -0,0 +1 @@
+{ lib, callPackage }: callPackage ./package.nix { inherit lib; }


Why is this here? This seemingly serves no purpose.

Pandapip1 · 2025-07-10T12:37:39Z

+
+  catch2_3WithSharedLibs = catch2_3.overrideAttrs (oldAttrs: {
+    cmakeFlags = (oldAttrs.cmakeFlags or [ ]) ++ [
+      "-DBUILD_SHARED_LIBS=ON"


This should be stdenv.hostPlatform.isStatic, not just always on. Also, consider using the helper for readability:

Suggested change

"-DBUILD_SHARED_LIBS=ON"

(lib.cmakeBool "BUILD_SHARED_LIBS" (!(stdenv.hostPlatform.isStatic)))

Also, this seems like an oversight and you should feel free to add this to the actual catch2_3 package.

Pandapip1 · 2025-07-10T12:40:40Z

+  # Latest commits are not compatible, use older version
+  ws-protocolCompat = ws-protocol.overrideAttrs (old: {


Non-blocking, but:

Has this been reported yet;

If it has, can you add a comment with a link to the issue?

Also, is there a specific commit that broke it? If so, consider adding a patch that reverts that commit.

Pandapip1 · 2025-07-10T12:41:32Z

+    patches = [ ./001-Ws-protocol-add-missing-libs.patch ];
+  });
+
+  prefixVar = "\${" + "prefix}";


Suggested change

prefixVar = "\${" + "prefix}";

prefixVar = "''${prefix}";

This is apparently how you properly escape this.

Pandapip1 · 2025-07-10T12:42:15Z

+  # Latest commits are not compatible, use older version
+  xtensorCompat = xtensor.overrideAttrs (old: {


Likewise, if there's an existing issue please link to it, else please create one.

Pandapip1 · 2025-07-10T13:11:28Z

+  src = fetchFromGitHub {
+    owner = "foxglove";
+    repo = "ws-protocol";
+    tag = "releases/cpp/v${finalAttrs.version}";


⁉️

~~Is there any reason this can't be used?~~

~~tag = "v${finalAttrs.version}";~~

Never mind, this is actually correct. That's actually horrible.

Pandapip1 · 2025-07-10T13:11:50Z

+    hash = "sha256-JMCwxShOOv1PSrlKiPGtsCgKJVR+7ds9otKrlPHfIio=";
+  };
+
+  # On Darwin the `openssl` and `zlib` are not found during linking


Has this been reported upstream?

Pandapip1 · 2025-07-10T13:12:46Z

+
+  sourceRoot = "${finalAttrs.src.name}/cpp/foxglove-websocket";
+
+  cmakeFlags = [ (lib.cmakeBool "BUILD_SHARED_LIBS" true) ];


You know the drill

Suggested change

cmakeFlags = [ (lib.cmakeBool "BUILD_SHARED_LIBS" true) ];

cmakeFlags = [ (lib.cmakeBool "BUILD_SHARED_LIBS" (!(stdenv.hostPlatform.isStatic))) ];

Pandapip1 · 2025-07-10T13:13:44Z

+        cat > $out/lib/pkgconfig/foxglove-websocket.pc << EOF
+    Name: foxglove-websocket
+    Description: Foxglove WebSocket Protocol Implementation
+    Version: ${finalAttrs.version}


Consider:

Suggested change

Version: ${finalAttrs.version}

Version: $version

I'm pretty sure this is preferred. I could be wrong though.

Pandapip1 · 2025-07-10T13:16:35Z

+  inherit (callPackage ../by-name/de/depthai-core/package.nix { })
+    depthai-core
+    depthai-core-minimal
+    depthai-core-full
+    ;


Okay, a few things:

default.nix is seemingly unused then and should be removed;

package.nixes in by-name should resolve to a derivation, not an attrset. The fact that it's not and then you're masking it in all-packages.nix is very cursed and will likely break something at some point.

To fix 2, just use depthai-core-full = depthai-core.override { ... } in all-packages.nix

Pandapip1

2nd pass

Pandapip1 · 2025-07-10T13:46:35Z

Consider adding pytestCheckHook, and a pythonImportsCheck

Pandapip1 · 2025-07-10T13:48:32Z

+  nativeBuildInputs = [ flit-core ];
+  propagatedBuildInputs = [
+    docutils
+    libclang
+  ];


Suggested change

nativeBuildInputs = [ flit-core ];

propagatedBuildInputs = [

docutils

libclang

];

build-system = [ flit-core ];

dependencies = [

docutils

libclang

];

python packages are weird and use different naming. I think there's some additional handling that happens when you name things correctly.

Pandapip1 · 2025-07-10T13:49:26Z

+
+        # Create foxglove-websocketConfigVersion.cmake
+        cat > $out/lib/cmake/foxglove-websocket/foxglove-websocketConfigVersion.cmake << EOF
+    set(PACKAGE_VERSION "${finalAttrs.version}")


Suggested change

set(PACKAGE_VERSION "${finalAttrs.version}")

set(PACKAGE_VERSION "$version")

Missed this too. Again, not required but I seem to recall this is recommended over string templates.

Pandapip1 · 2025-07-10T13:50:00Z

+    # foxglove-websocket CMake configuration file
+
+    # Compute installation prefix relative to this file
+    get_filename_component(_IMPORT_PREFIX "\''${CMAKE_CURRENT_LIST_DIR}/../../.." ABSOLUTE)


I don't think that \ is required. Could be wrong, but:

Suggested change

get_filename_component(_IMPORT_PREFIX "\''${CMAKE_CURRENT_LIST_DIR}/../../.." ABSOLUTE)

get_filename_component(_IMPORT_PREFIX "''${CMAKE_CURRENT_LIST_DIR}/../../.." ABSOLUTE)

Pandapip1 · 2025-07-10T13:50:54Z

+
+  cmakeFlags = [
+    (lib.cmakeBool "BUILD_EXAMPLE" true)
+    (lib.cmakeBool "BUILD_TESTS" true)


These should be run in checkPhase.

Pandapip1 · 2025-07-10T14:04:51Z

+  });
+
+  # Latest commits are not compatible, use older version
+  ws-protocolCompat = ws-protocol.overrideAttrs (old: {


old is seemingly not used

Suggested change

ws-protocolCompat = ws-protocol.overrideAttrs (old: {

ws-protocolCompat = ws-protocol.overrideAttrs {

Pandapip1 · 2025-07-10T14:05:20Z

+  });
+
+  # Latest commits are not compatible, use older version
+  xlinkCompat = xlink.overrideAttrs (old: {


old is seemingly not used

Suggested change

xlinkCompat = xlink.overrideAttrs (old: {

xlinkCompat = xlink.overrideAttrs {

Pandapip1 · 2025-07-10T14:05:44Z

+  pyOpencv4WithGtk = python3Packages.opencv4.override {
+    enableGtk2 = true; # For GTK2 support
+    enableGtk3 = true; # For GTK3 support
+  };


Suggested change

pyOpencv4WithGtk = python3Packages.opencv4.override {

enableGtk2 = true; # For GTK2 support

enableGtk3 = true; # For GTK3 support

};

Unused

Pandapip1 · 2025-07-10T14:06:11Z

+  rerun_sdk = fetchzip {
+    url = "https://github.com/rerun-io/rerun/releases/download/0.16.1/rerun_cpp_sdk.zip";
+    hash = "sha256-tYwcRjlU8m+OHkwdbQeM+X5outF0NFbj8Ee3y5h9+l0=";
+    stripRoot = false; # Important for zip files to preserve the directory structure
+  };


Can't this be a fetchFromGitHub?

Pandapip1 · 2025-07-10T14:07:51Z

+
+    preConfigure = ''
+      echo "Fixing xtensor.pc.in for Nix compatibility..."
+      find . -name xtensor.pc.in -exec sed -i 's|${prefixVar}/@CMAKE_INSTALL_LIBDIR@|@CMAKE_INSTALL_FULL_LIBDIR@|g' {} \;


prefixVar is only ever used here. Consider:

Suggested change

find . -name xtensor.pc.in -exec sed -i 's|${prefixVar}/@CMAKE_INSTALL_LIBDIR@|@CMAKE_INSTALL_FULL_LIBDIR@|g' {} \;

find . -name xtensor.pc.in -exec sed -i 's|''${prefix}/@CMAKE_INSTALL_LIBDIR@|@CMAKE_INSTALL_FULL_LIBDIR@|g' {} \;

and dropping the variable

phodina · 2025-07-15T09:22:06Z

First pass. Please excuse any typos, this was written on my phone over a train ride.

No porblem. Thanks, started on the fixes

phodina · 2026-01-28T09:36:53Z

As I'm no longer affiliated with Luxonis I must put this into a draft mode.

If there would be anyone willing to take up the mantel feel free :)

@MaticTonin would you take on the challenge?

Amronos · 2026-02-08T17:42:53Z

I have been very slowly going through the code and cleaning it up for a while now. It will take a lot more time to have a mergeable PR ready, but if somebody wants to take a look/collaborate, I can put up a draft PR.

phodina force-pushed the depthaiv3_alpha15_draft branch from 170f800 to b902544 Compare April 26, 2025 13:18

github-actions Bot added 10.rebuild-darwin: 1-10 This PR causes between 1 and 10 packages to rebuild on Darwin. 10.rebuild-linux: 1-10 This PR causes between 1 and 10 packages to rebuild on Linux. labels Apr 26, 2025

phodina force-pushed the depthaiv3_alpha15_draft branch from b902544 to 4406da9 Compare April 26, 2025 13:37

phodina changed the title ~~DRAFT: Depthaiv3 alpha15~~ Depthaiv3 alpha15 May 1, 2025

phodina marked this pull request as ready for review May 1, 2025 17:43

Pandapip1 reviewed May 1, 2025

View reviewed changes

Comment thread pkgs/by-name/ba/backward-cpp/package.nix Outdated

ofborg Bot added the 2.status: merge conflict This PR has merge conflicts with the target branch label May 3, 2025

phodina force-pushed the depthaiv3_alpha15_draft branch from a0a6b55 to 0bff900 Compare May 4, 2025 15:41

ofborg Bot removed the 2.status: merge conflict This PR has merge conflicts with the target branch label May 4, 2025

wegank added the 2.status: merge conflict This PR has merge conflicts with the target branch label May 17, 2025

phodina force-pushed the depthaiv3_alpha15_draft branch from b0d4a30 to f5db0dd Compare May 18, 2025 07:09

ofborg Bot removed the 2.status: merge conflict This PR has merge conflicts with the target branch label May 18, 2025

github-actions Bot added the 6.topic: python Python is a high-level, general-purpose programming language. label May 18, 2025

nix-owners Bot requested a review from natsukium May 18, 2025 10:33

phodina force-pushed the depthaiv3_alpha15_draft branch 2 times, most recently from c399786 to 82d9152 Compare May 22, 2025 07:51

nix-owners Bot requested a review from mweinelt May 22, 2025 07:53

mweinelt removed their request for review May 22, 2025 13:07

nix-owners Bot requested a review from mweinelt May 28, 2025 10:50

phodina and others added 6 commits May 28, 2025 12:55

ws-protocol: init at 1.4.0

109983a

mcap: init at 0.0.53

c24455d

nanorpc: init at 1.1.0-unstable-2021-04-29

4d51025

backward-cpp: add phodina as maintainer

59dadfa

pybind-11-mkdoc: init at 0-unstable-2023-02-08

a59f0a6

depthai-data: 0-unstable-2025-05-22

a4796c3

phodina force-pushed the depthaiv3_alpha15_draft branch from 82d9152 to 9cf32e7 Compare May 28, 2025 10:55

depthai-core: init at 3.0.0-beta.1

61bb2e6

phodina force-pushed the depthaiv3_alpha15_draft branch from 9cf32e7 to 61bb2e6 Compare May 28, 2025 11:06

phodina force-pushed the depthaiv3_alpha15_draft branch from 21f1b34 to 86d47f1 Compare May 31, 2025 10:35

depthai-core: Split into depthai-core-minimal, depthai-core and depth…

b86b6db

…ai-core-full

phodina force-pushed the depthaiv3_alpha15_draft branch from 86d47f1 to b86b6db Compare May 31, 2025 10:44

Pandapip1 mentioned this pull request Jun 3, 2025

depthai-core: init at 2.28.0 #343281

Closed

13 tasks

hacker1024 mentioned this pull request Jul 7, 2025

depthai* packages fail to build downloading hunter dependencies to /homeless-shelter lopsided98/nix-ros-overlay#412

Open

Pandapip1 suggested changes Jul 10, 2025

View reviewed changes

nixpkgs-ci Bot added the 2.status: merge conflict This PR has merge conflicts with the target branch label Oct 6, 2025

nixpkgs-ci Bot added the 2.status: stale https://github.com/NixOS/nixpkgs/blob/master/.github/STALE-BOT.md label Jan 12, 2026

phodina marked this pull request as draft January 28, 2026 09:37

nixpkgs-ci Bot removed the 2.status: stale https://github.com/NixOS/nixpkgs/blob/master/.github/STALE-BOT.md label Jan 28, 2026

Amronos mentioned this pull request Mar 15, 2026

depthai-core: init at 3.5.0 #500215

Draft

20 tasks

		@@ -0,0 +1 @@
		{ lib, callPackage }: callPackage ./package.nix { inherit lib; }

	"-DBUILD_SHARED_LIBS=ON"
	(lib.cmakeBool "BUILD_SHARED_LIBS" (!(stdenv.hostPlatform.isStatic)))

		# Latest commits are not compatible, use older version
		ws-protocolCompat = ws-protocol.overrideAttrs (old: {

		# Latest commits are not compatible, use older version
		xtensorCompat = xtensor.overrideAttrs (old: {


		sourceRoot = "${finalAttrs.src.name}/cpp/foxglove-websocket";

		cmakeFlags = [ (lib.cmakeBool "BUILD_SHARED_LIBS" true) ];

	set(PACKAGE_VERSION "${finalAttrs.version}")
	set(PACKAGE_VERSION "$version")

	get_filename_component(_IMPORT_PREFIX "\''${CMAKE_CURRENT_LIST_DIR}/../../.." ABSOLUTE)
	get_filename_component(_IMPORT_PREFIX "''${CMAKE_CURRENT_LIST_DIR}/../../.." ABSOLUTE)

	xlinkCompat = xlink.overrideAttrs (old: {
	xlinkCompat = xlink.overrideAttrs {

	find . -name xtensor.pc.in -exec sed -i 's\|${prefixVar}/@CMAKE_INSTALL_LIBDIR@\|@CMAKE_INSTALL_FULL_LIBDIR@\|g' {} \;
	find . -name xtensor.pc.in -exec sed -i 's\|''${prefix}/@CMAKE_INSTALL_LIBDIR@\|@CMAKE_INSTALL_FULL_LIBDIR@\|g' {} \;

Uh oh!

Uh oh!

Conversation

phodina commented Apr 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Things done

Uh oh!

phodina commented Apr 30, 2025

Uh oh!

phodina commented May 1, 2025

Uh oh!

Uh oh!

Pandapip1 commented May 1, 2025

Uh oh!

phodina commented May 2, 2025

Uh oh!

phodina commented May 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

phodina commented May 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Pandapip1 commented May 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

phodina commented May 18, 2025

Uh oh!

phodina commented May 18, 2025

Uh oh!

phodina commented May 22, 2025

Uh oh!

mweinelt commented May 22, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

phodina commented May 28, 2025

Uh oh!

phodina commented May 31, 2025

Uh oh!

phodina commented May 31, 2025

Uh oh!

phodina commented May 31, 2025

Uh oh!

phodina commented Jul 10, 2025

Uh oh!

Pandapip1 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Pandapip1 Jul 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Pandapip1 Jul 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Pandapip1 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

phodina commented Apr 26, 2025 •

edited

Loading

phodina commented May 3, 2025 •

edited

Loading

phodina commented May 18, 2025 •

edited

Loading

Pandapip1 commented May 18, 2025 •

edited

Loading

mweinelt commented May 22, 2025 •

edited

Loading

Pandapip1 Jul 10, 2025 •

edited

Loading

Pandapip1 Jul 10, 2025 •

edited

Loading