Mac-动态库搜索及加载顺序-dylib-应用程序打包

  国内一般操作系统为windows操作系统，但是mac系统在高端市场，也占据着一部分比率，开发Mac应用，需要先了解，Mac操作系统搜索动态库的顺序。

1.搜索顺序

  因为运行时动态库的搜索路径的先后顺序是：     a.编译目标代码时指定的动态库搜索路径；     b.环境变量LD_LIBRARY_PATH指定的动态库搜索路径；     c.配置文件/etc/ld.so.conf中指定的动态库搜索路径；     d.默认的动态库搜索路径/lib和/usr/lib；     /lib和/usr/lib只有root权限才可以修改，配置文件/etc/ld.so.conf也需要root权限。

2.和应用程序一同发布

  在xcode下编译出的程序，在开发机器上运行是没有问题的。但是给其他用户用，就可能出问题。因为用户不一定有这个库。有两种方法可以解决这个问题；一是给其他用户也安装依赖的库文件；二是将所有的dylib随行发布，消除依赖。 第一种方案不考虑，大部分时候这样做并不现实；下面说说如何随行发布dylib。单纯将依赖的dylib文件拷贝到可执行文件目录下一同传输过去是不能消除依赖的；执行的时候还是报错；在编译一个动态库的时候, 你需要指定 INSTALL_PATH. 也就是它的安装路径；编译完成后如果一个可执行程序使用了该动态库, 那么在编译可执行程序的时候, 动态库的 INSTALL_PATH 会被记录到可执行程序中, 用来定位这个动态库。因此我们首先需要将用到的dylib文件都拷贝到可执行文件目录下，然后改变动态库的INSTALL_PATH；将其改到可执行文件所在目录；需要注意的是：如果依赖多个动态库，用到的动态库已会依赖其他动态库，因此用到的所有的动态库的依赖动态库路径都需要修改。   以OpenCV为例子，假设最终编译出来的可执行文件为 macimgproc；执行命令：otool -L macimgproc可看到如下的输出：

macimgproc:
    /usr/local/opt/opencv/lib/libopencv_calib3d.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_contrib.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_core.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_features2d.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_flann.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_gpu.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_highgui.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_imgproc.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_legacy.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_ml.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_nonfree.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_objdetect.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_ocl.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_photo.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_stitching.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_superres.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_video.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_videostab.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/lib/libc++.1.dylib (compatibility version 1.0.0, current version 307.4.0)
    /usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 1238.0.0)

  说明macimgproc依赖所有的OpenCV动态库文件；因此首先需要将所有动态库文件拷贝到macimgproc所在目录，然后需要将macimgproc文件中的所有/usr/local/opt/opencv/lib/libopencv_xxx修改为@executable_path/libopencv_xxx；@executable_path表示可执行文件所在目录；指示所有OpenCV动态库从可执行文件所在目录查找；更多@executable_path的介绍以及其他变量参见文章 @rpath, @loader_path, @executable_path。使用命令install_name_tool -change {old.dylib} {new.dylib} {filename}修改动态库的INSTALL_PATH，例如：

install_name_tool -change /usr/local/Cellar/opencv/2.4.12/lib/libopencv_flann.2.4.dylib @executable_path/libopencv_flann.2.4.dylib macimgproc

  执行后重新otool -L macimgproc可看到如下的输出：

macimgproc:
    ...
    /usr/local/opt/opencv/lib/libopencv_features2d.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    @executable_path/libopencv_flann.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    /usr/local/opt/opencv/lib/libopencv_gpu.2.4.dylib (compatibility version 2.4.0, current version 2.4.12)
    ...

  整个OpenCV库大概有19个dylib文件，因此写了一个简单的批量修改脚本：

#!/usr/bin/ruby

Preview=0

def fix_opencv_lib_link(file,rlib)
    if(rlib.include?('libopencv'))
        name = rlib.split('/').last
        cmd = "install_name_tool -change #{rlib} @executable_path/#{name} #{file}"
        if Preview == 1
            puts "Preivew: #{cmd}"
        else
            `#{cmd}`
        end
    else 
        puts "ignore rlib: #{rlib}";
    end
end

def fix_file_rely_lib(file)
  puts "===============start change #{file}==============="
  linklibs = `otool -L #{file}`.split("\n")
  linklibs.delete_at(0)
  linklibs.each_with_index do |rlib,i|
    rlib = rlib.split()[0]
    fix_opencv_lib_link(file,rlib)
  end
end

def doopencvlist
  # puts "Preview: #{Preview}"
  `ls |grep libopencv`.split().each_with_index do |file,i|
    fix_file_rely_lib(file)
  end
end

def viewlib(file)
    puts `otool -L #{file}`
end

if __FILE__ == $0
  if ARGV[0] == 'preview'
    Preview = 1;
    doopencvlist();
  elsif ARGV[0] == 'view'
    `ls |grep libopencv`.split().each_with_index do |file,i|
      # puts "===============start view #{file}==============="
      viewlib(file)
    end
  elsif ARGV[0] == 'do'
    doopencvlist();
  elsif ARGV[0] == 'previewfile'
    Preview = 1;
    fix_file_rely_lib(ARGV[1])
  elsif ARGV[0] == 'file'
    fix_file_rely_lib(ARGV[1])
  else
    puts "please input command [preview,do,previewfile,file,view] "
  end
end

  合理的脚本代码可以有效的提高工作效率，减少重复劳动。