跳到主要內容

Take a look at CMake


Overview

CMake is a cross-platform build tool like GNU build system, but it's much easier to use and faster than GNU build system. It is designed to build software written by C, C++ or Java. It adds a lot of implicit rules to help programmer with few configurations. This article will explain how to use this powerful tool step by step with workable examples. We will talk about the details of each syntax we use here but not try to exhaust all syntax CMake has. If you want to dig into CMake, please refer to its document
The version of CMake we use is 2.8.7 on Ubuntu 12.04. However, we just use the basic syntax. It sould be okay if you use other versions and on other platforms.

First example - say hello                                                                                    

Following the great tradion, our first example will use CMake to build a simple C++ program to say hello. Before we add any CMake configuration files, we have following files:
mars@dream$ pwd
/home/mars/try/cmake/C++Hello
mars@dream$ tree
.
└── main.cxx
 
0 directories, 1 file
mars@dream$ cat main.cxx
#include <iostream>
 
int main(void)
{
        std::cout << "Hello, SMAP." << std::endl;
}
  
To get CMake's great capabilities, we should add a special file named "CMakeLists.txt" in the directory. Here comes the file with comments:
#CMake's instructions are not case-sensitive, but the convention is to use the upper case
 
#CMAKE_MINIMUM_REQUIRED requires the minimum version of CMake
#if you do not specify this, cmake will complain you
CMAKE_MINIMUM_REQUIRED(VERSION 2.8)
 
#give this project a name
PROJECT(hello)
 
#use main.cxx to build an executable file named hello
ADD_EXECUTABLE(hello main.cxx)
OK, that's all. use the following commands to trigger CMake and make things happen:
mars@dream$ ls
CMakeLists.txt  main.cxx
mars@dream$ cmake .
-- The C compiler identification is GNU
-- The CXX compiler identification is GNU
-- Check for working C compiler: /usr/bin/gcc
-- Check for working C compiler: /usr/bin/gcc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Configuring done
-- Generating done
-- Build files have been written to: /home/mars/try/cmake/C++Hello
mars@dream$ ls
CMakeCache.txt  CMakeFiles  cmake_install.cmake  CMakeLists.txt  main.cxx  Makefile
mars@dream$ make
Scanning dependencies of target hello
[100%] Building CXX object CMakeFiles/hello.dir/main.cxx.o
Linking CXX executable hello
[100%] Built target hello
mars@dream$ ls
CMakeCache.txt  CMakeFiles  cmake_install.cmake  CMakeLists.txt  hello  main.cxx  Makefile
mars@dream$ ./hello
Hello, SMAP.
mars@dream$ make clean
mars@dream$ ls
CMakeCache.txt  CMakeFiles  cmake_install.cmake  CMakeLists.txt  main.cxx  Makefile
mars@dream$
As we observed, cmake detects what toolchain should be used and generates a makefile after using "cmake .". The makefile it generates will contain many makefile targets, like all, clean...etc. 

Second example - say hello with hierarchy directories

Previous example is too simple to be true. When we develop a software, we usually need to break functionalities into different directories. One of the powers of CMake is its flexibility to handle this. We will seperate some functions from main function and try to build a library. The main program will include the library's declarations(header file) and be linked against it. In addition, the previous example shows that CMake will generate some files which might mess with our source tree. We will see how to get all of those things done.
Before CMakeLists.txt comes in, we have the following directories and files:
mars@dream$ pwd
/home/mars/try/cmake/C++HelloHierarchy
mars@dream$ tree
.
├── build
├── include
│      └── hello.hxx
├── lib
│      └── hello.cxx
└── src
         └── main.cxx
 
4 directories, 3 files
mars@dream$ cat include/hello.hxx
#ifndef _HELLO_HXX_
#define _HELLO_HXX_
 
void hello(void);
void smap(void);
 
#endif //end of _HELLO_HXX_
mars@dream$ cat lib/hello.cxx
#include <iostream>
#include <hello.hxx>
 
void hello(void)
{
        std::cout << "Hello, ";
}
 
void smap(void)
{
        std::cout << "SMAP." << std::endl;
}
mars@dream$ cat src/main.cxx
#include <hello.hxx>
 
int main(void)
{
        hello();
        smap();
}

The build directory is for clean build. CMake can not remove the extra files it generates due to some reasons. CMake solves this problem by an elegant way - out-of-source build. Briefly speaking, out-of-source means you can build the software in any directory without building software in source tree directory. Therefore the files CMake generates will only reside in the build directory. Our empty "build" directory is for this approach.
Now we can add some CMakeLists.txt files into source tree: 
mars@dream$ pwd
/home/mars/try/cmake/C++HelloHierarchy
mars@dream$ tree
.
├── build
├── CMakeLists.txt
├── include
│      └── hello.hxx
├── lib
│      ├── CMakeLists.txt
│      └── hello.cxx
└── src
        ├── CMakeLists.txt
        └── main.cxx
 
4 directories, 6 files
mars@dream$ cat ./CMakeLists.txt
#I just add comments with new instructions.
 
CMAKE_MINIMUM_REQUIRED(VERSION 2.8)
PROJECT(hello_hierarchy)
 
#CMake's way to define a variable.
#we define a list variable containing each subdirectories to build
SET(SUBDIRS lib src)
 
#use foreach instruction to tell CMake what subdirectories it should build
foreach(dir ${SUBDIRS})
  ADD_SUBDIRECTORY(${dir})
endforeach()
mars@dream$ cat lib/CMakeLists.txt
#${PROJECT_SOURCE_DIR} would be the top level of the source tree
#in this case, it is /home/mars/try/cmake/C++HelloHierarchy
#INCLUDE_DIRECTORIES instruction means CMake should add an include path
#before doing the following process
INCLUDE_DIRECTORIES(${PROJECT_SOURCE_DIR}/include)
 
#ask CMake to generate both shared and static version of this library
ADD_LIBRARY(hello_shared SHARED hello.cxx)
ADD_LIBRARY(hello_static STATIC hello.cxx)
 
#By default, CMake will generate library based on the first argument of ADD_LIBRARY
#However, that's not we want. What we want is libhello.a and libhello.so
#not libhello_static.a nor libhello_shared.so
#here comes the idiom to solve the odds
SET_TARGET_PROPERTIES(hello_static PROPERTIES OUTPUT_NAME "hello")
SET_TARGET_PROPERTIES(hello_shared PROPERTIES OUTPUT_NAME "hello")
 
#in addition, shared library in Unix system usually has version with it
#we can mark its version by the following instruction
SET_TARGET_PROPERTIES(hello_shared PROPERTIES VERSION 1.1 SOVERSION 1)
mars@dream$ cat src/CMakeLists.txt
INCLUDE_DIRECTORIES (${PROJECT_SOURCE_DIR}/include)
 
#${PROJECT_BINARY_DIR}'s value is the build directory's path
#since we use out-of-source build,
#it might be /home/mars/try/cmake/C++HelloHierarchy
#LINK_DIRECTORIES instruction will add the path to search link path
LINK_DIRECTORIES(${PROJECT_BINARY_DIR}/lib)
 
ADD_EXECUTABLE(hello_hie main.cxx)
 
#since we have the link path of libhello.[a|so], we should ask CMake to link it
TARGET_LINK_LIBRARIES(hello_hie hello)

Now, change to build directory and type as the following:
mars@dream$ cmake ..
-- The C compiler identification is GNU
-- The CXX compiler identification is GNU
-- Check for working C compiler: /usr/bin/gcc
-- Check for working C compiler: /usr/bin/gcc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Configuring done
-- Generating done
-- Build files have been written to: /home/mars/try/cmake/C++HelloHierarchy/build
mars@dream$ make
Scanning dependencies of target hello_shared
[ 33%] Building CXX object lib/CMakeFiles/hello_shared.dir/hello.cxx.o
Linking CXX shared library libhello.so
[ 33%] Built target hello_shared
Scanning dependencies of target hello_static
[ 66%] Building CXX object lib/CMakeFiles/hello_static.dir/hello.cxx.o
Linking CXX static library libhello.a
[ 66%] Built target hello_static
Scanning dependencies of target hello_hie
[100%] Building CXX object src/CMakeFiles/hello_hie.dir/main.cxx.o
Linking CXX executable hello_hie
[100%] Built target hello_hie
mars@dream$ ./src/hello_hie
Hello, SMAP.
mars@dream$ ls lib/
CMakeFiles  cmake_install.cmake  libhello.a  libhello.so  libhello.so.1  libhello.so.1.1  Makefile

Next step

You might be so excited as I first happened to CMake. You want to know more. How to integrate installation process? How to automate test? ...etc. There is no end of this article if we add those things. However, after reading this article, you already know the most important concpets of CMake. I hope this will help you to try more with CMake. I suggest the following links:

留言

這個網誌中的熱門文章

誰在呼叫我?不同的backtrace實作說明好文章

今天下班前一個同事問到:如何在Linux kernel的function中主動印出backtrace以方便除錯? 寫過kernel module的人都知道,基本上就是用dump_stack()之類的function就可以作到了。但是dump_stack()的功能是如何作到的呢?概念上其實並不難,慣用手法就是先觀察stack在function call時的變化(一般OS或計組教科書都有很好的說明,如果不想翻書,可以參考 這篇 ),然後將對應的return address一層一層找出來後,再將對應的function名稱印出即可(透過執行檔中的section去讀取函式名稱即可,所以要將KALLSYM選項打開)。在userspace的實作可參考Jserv介紹過的 whocallme 或對岸好手實作過的 backtrace() ,都是針對x86架構的很好說明文章。 不過從前面兩篇文章可以知道,只要知道編譯器的calling convention,就可以實作出backtrace,所以是否GCC有提供現成的機制呢?Yes, that is what __builtin_return_address() for!! 可以參考這篇 文章 。該篇文章還提到了其他可以拿來實作功能更齊全的backtrace的 程式庫 ,在了解了運作原理後,用那些東西還蠻方便的。 OK,那Linux kernel是怎麼做的呢?就是用頭兩篇文章的方式啦~ 每個不同的CPU架構各自手工實作一份dump_stack()。 為啥不用GCC的機制?畢竟...嗯,我猜想,除了backtrace以外,開發者還會想看其他register的值,還有一些有的沒的,所以光是GCC提供的介面是很難印出全部所要的資訊,與其用半套GCC的機制,不如全都自己來~ arm的實作 大致上長這樣,可以看到基本上就只是透過迭代fp, lr, pc來完成: 352 void unwind_backtrace (struct pt_regs * regs , struct task_struct *tsk) 353 { 354 struct stackframe frame ; 355 register unsigned long current_sp asm ( "

淺讀Linux root file system初始化流程

在Unix的世界中,file system佔據一個極重要的抽象化地位。其中,/ 所代表的rootfs更是所有後續新增file system所必須依賴前提條件。以Linux為例,黑客 Jserv 就曾經詳細說明過 initramfs的背後設計考量 。本篇文章不再重複背景知識,主要將追蹤rootfs初始化的流程作點整理,免得自己日後忘記。 :-) file system與特定CPU架構無關,所以我觀察的起點從init/main.c的start_kernel()開始,這是Linux作完基本CPU初始化後首先跳進的C function(我閱讀的版本為 3.12 )。跟root file system有關的流程羅列如下: start_kernel()         -> vfs_caches_init_early()         -> vfs_caches_init()                 -> mnt_init()                         -> init_rootfs()                         -> init_mount_tree()         -> rest_init()                 -> kernel_thread(kernel_init,...) 其中比較重要的是mnt_int()中的init_rootfs()與init_mout_tree()。init_rootfs()實作如下: int __init init_rootfs(void) {         int err = register_filesystem(&rootfs_fs_type);         if (err)                 return err;         if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&                 (!root_fs_names || strstr(root_fs_names, "tmpfs"))) {          

kernel panic之後怎麼辦?

今天同事在處理一個陌生的模組時遇到kernel panic,Linux印出了backtrace,同事大致上可以知道是在哪個function中,但該function的長度頗長,短時間無法定位在哪個位置,在這種情況下,要如何收斂除錯範圍呢?更糟的是,由於加入printk會改變模組行為,所以printk基本上無法拿來檢查參數的值是否正常。 一般這樣的問題會backtrace的資訊來著手。從這個資訊我們可以知道在function中的多少offset發生錯誤,以x86為例(從 LDD3 借來的例子): Unable to handle kernel NULL pointer dereference at virtual address 00000000 printing eip: d083a064 Oops: 0002 [#1] SMP CPU:    0 EIP:    0060:[<d083a064>]    Not tainted EFLAGS: 00010246   (2.6.6) EIP is at faulty_write+0x4/0x10 [faulty] eax: 00000000   ebx: 00000000   ecx: 00000000   edx: 00000000 esi: cf8b2460   edi: cf8b2480   ebp: 00000005   esp: c31c5f74 ds: 007b   es: 007b   ss: 0068 Process bash (pid: 2086, threadinfo=c31c4000 task=cfa0a6c0) Stack: c0150558 cf8b2460 080e9408 00000005 cf8b2480 00000000 cf8b2460 cf8b2460        fffffff7 080e9408 c31c4000 c0150682 cf8b2460 080e9408 00000005 cf8b2480        00000000 00000001 00000005 c0103f8f 00000001 080e9408 00000005 00000005 Call Trace:  [<c0150558>] vfs