Arch Linux User Repository

This package is also hosted on GitHub.

• Please report issues and patches at GitHub.
• Travis Build Status: <img alt="Build Status" src="https://travis-ci.org/bartoszek/AUR-colmap-git.svg?branch=travis" />

Use env vars to control build process:

• BUILD_CUDA=ON|OFF to skip cuda kernel build (default 'ON')
• CUDA_ARCH="5.2,7.2" to build for a specific Cuda arch, supports multiple values.
• CMAKE_FLAGS="xxx:yyy" to add extra CMake switches (colon-separated list)

Usage cases:

• export BUILD_CUDA=OFF before build
• BUILD_CUDA=OFF ~your-aur-helper~
• makepkg BUILD_CUDA=OFF
• yay -S colmap-git --mflags "BUILD_CUDA=OFF"

dependencies shall be changed to adapt for the new libjpeg-turbo [1], which provides the required libjpeg.so.

[1] https://archlinux.org/packages/extra/x86_64/libjpeg-turbo/

This works okay with cuda 12 btw I use the following somewhat hacky and modified scipt to get the build working. Maybe I should just do cuda < 13.

 #!/bin/hint/bash
 # Maintainer : bartus <arch-user-repoᘓbartus.33mail.com>

-check_option "debug" "y" && BUILD_TYPE=Debug || BUID_TYPE=Release
-
 ## Configuration env vars:
 _BUILD_CUDA="${BUILD_CUDA:-ON}"
 _CUDA_ARCHITECTURES="${CUDA_ARCH:-native}"
@@ -13,7 +11,7 @@
 _name=colmap
 #fragment="#commit=5bea89263bf5f3ed623b8e6e6a5f022a0ed9c1de"
 pkgname=${_name}-git
-pkgver=3.8.r219.g6eb27c01
+pkgver=3.8.r224.gc0d89268
 pkgrel=1
 pkgdesc="General-purpose Structure-from-Motion (SfM) and Multi-View Stereo (MVS) pipeline with a graphical and command-line interface."
 arch=('i686' 'x86_64')
@@ -25,7 +23,7 @@
 depends=('cgal' 'ceres-solver' 'gflags' 'metis' 'suitesparse' 'freeglut' 'glew' 'google-glog' 'freeimage' 'libjpeg' 'boost-libs' 'qt5-base' 'flann')
 makedepends=('boost' 'cmake' 'eigen' 'git' 'ninja' 'python-sphinx' )
 if [ "$_BUILD_CUDA" == "ON" ] ; then 
-  makedepends+=('cuda<12')
+  makedepends+=('cuda>=12')
   optdepends+=('libcudart.so: required for dense reconstruction')
 fi
 source=("${pkgname}::git+https://github.com/colmap/colmap.git${_fragment}"
@@ -42,9 +40,6 @@
   git -C "$pkgname" describe --long --tags | sed 's/\([^-]*-g\)/r\1/;s/-/./g'
 }

-prepare() {
-: #  git -C "$pkgname" apply -v "$srcdir"/*.patch
-}

 build() {
   # determine whether we can precompile CUDA kernels
@@ -62,7 +57,7 @@
     fi

   _CMAKE_FLAGS+=( -DTESTS_ENABLED=OFF
-                  -DCMAKE_BUILD_TYPE=${BUILD_TYPE}
+                  -DCMAKE_BUILD_TYPE=Release
                   -DCMAKE_INSTALL_PREFIX=/usr )
   cmake "${_CMAKE_FLAGS[@]}" -G Ninja -S "$pkgname" -B build
 # shellcheck disable=SC2046 # allow MAKEFLAGS to carry multiple flags.
@@ -83,4 +78,4 @@
     install -m644 "${vocab_tree}" "${pkgdir}/usr/share/${_name}/${vocab_tree##*/}"
   done
 }

Hello @bartus,

The package fails to apply the patch cuda11_gcc11.patch. It compiles though after removing the patch from PKGBUILD.

==> Starting prepare()...
Checking patch src/colmap/CMakeLists.txt...
warning: src/colmap/CMakeLists.txt has type 100644, expected 100755
error: while searching for:
       IS_GNU AND CMAKE_CXX_COMPILER_VERSION VERSION_LESS 6)
        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
    else()
        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14")
    endif()
endif()


error: patch failed: src/colmap/CMakeLists.txt:43
error: src/colmap/CMakeLists.txt: patch does not apply
==> ERROR: A failure occurred in prepare().
    Aborting...
 -> error making: colmap-git-exit status 4
 -> Failed to install the following packages. Manual intervention is required:
colmap-git - exit status 4

@hrehfeld This is because of the Auto at the start of the pkgbuild:

_CUDA_ARCH="${CUDA_ARCH:-Auto}"

I changed it to:

_CUDA_ARCH="${CUDA_ARCH:-native}"

Or here's documentation on the possible values: https://cmake.org/cmake/help/latest/prop_tgt/CUDA_ARCHITECTURES.html#prop_tgt:CUDA_ARCHITECTURES

Build is currently broken:

-- Targetting Ninja
-- Using /usr/bin/c++ compiler.
-- Found GMP: /usr/lib/libgmp.so  
-- Found MPFR: /usr/lib/libmpfr.so  
-- Found Boost: /usr/lib64/cmake/Boost-1.80.0/BoostConfig.cmake (found suitable version "1.80.0", minimum required is "1.48")  
-- Boost include dirs: /usr/include
-- Boost libraries:    
-- Using gcc version 4 or later. Adding -frounding-math
-- Found Qt
--   Module : /usr/lib64/cmake/Qt5Core
--   Module : /usr/lib64/cmake/Qt5OpenGL
--   Module : /usr/lib64/cmake/Qt5Widgets
-- Found CGAL
--   Includes : /usr/include
--   Libraries : CGAL
-- Build type specified as Release
-- Enabling SIMD support
-- Enabling OpenMP support
-- Disabling interprocedural optimization
CMake Error at /usr/share/cmake/Modules/CMakeDetermineCUDACompiler.cmake:279 (message):
  CMAKE_CUDA_ARCHITECTURES:

    Auto

  is not one of the following:

    * a semicolon-separated list of integers, each optionally
      followed by '-real' or '-virtual'
    * a special value: all, all-major, native

Call Stack (most recent call first):
  CMakeLists.txt:222 (enable_language)


-- Configuring incomplete, errors occurred!

No possible install without cuda package. I use software solver, I disable BUILD_CUDA but dependencies CUDA (1.3GB) is always installed.

https://github.com/colmap/colmap/issues/342

This package is also hosted on GitHub.

• Please report issues and patches at GitHub.
• Travis Build Status: <img alt="Build Status" src="https://travis-ci.org/bartoszek/AUR-colmap-git.svg?branch=travis" />

Use env vars to control build process:

• BUILD_CUDA=ON|OFF to skip cuda kernel build (default 'ON')
• CUDA_ARCH="5.2,7.2" to build for a specific Cuda arch, supports multiple values.
• CMAKE_FLAGS="xxx:yyy" to add extra CMake switches (colon-separated list)

Usage cases:

• export BUILD_CUDA=OFF before build
• BUILD_CUDA=OFF ~your-aur-helper~
• makepkg BUILD_CUDA=OFF
• yay -S colmap-git --mflags "BUILD_CUDA=OFF"

@herbert: Nope, support for Nvidia Fermi architecture (sm≤3.0) was dropped in cuda 9 compiler. You need to downgrade cuda to V8 for compilation to succeed.

btw. driver's capabilities still includes sm_20