SCons 2.3.1

User Guide

Steven Knight

Steven Knight

version 2.3.1

2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014


Table of Contents

Preface
1. SCons Principles
2. A Caveat About This Guide's Completeness
3. Acknowledgements
4. Contact
1. Building and Installing SCons
1. Installing Python
2. Installing SCons From Pre-Built Packages
2.1. Installing SCons on Red Hat (and Other RPM-based) Linux Systems
2.2. Installing SCons on Debian Linux Systems
2.3. Installing SCons on Windows Systems
3. Building and Installing SCons on Any System
3.1. Building and Installing Multiple Versions of SCons Side-by-Side
3.2. Installing SCons in Other Locations
3.3. Building and Installing SCons Without Administrative Privileges
2. Simple Builds
1. Building Simple C / C++ Programs
2. Building Object Files
3. Simple Java Builds
4. Cleaning Up After a Build
5. The SConstruct File
5.1. SConstruct Files Are Python Scripts
5.2. SCons Functions Are Order-Independent
6. Making the SCons Output Less Verbose
3. Less Simple Things to Do With Builds
1. Specifying the Name of the Target (Output) File
2. Compiling Multiple Source Files
3. Making a list of files with Glob
4. Specifying Single Files Vs. Lists of Files
5. Making Lists of Files Easier to Read
6. Keyword Arguments
7. Compiling Multiple Programs
8. Sharing Source Files Between Multiple Programs
4. Building and Linking with Libraries
1. Building Libraries
1.1. Building Libraries From Source Code or Object Files
1.2. Building Static Libraries Explicitly: the StaticLibrary Builder
1.3. Building Shared (DLL) Libraries: the SharedLibrary Builder
2. Linking with Libraries
3. Finding Libraries: the $LIBPATH Construction Variable
5. Node Objects
1. Builder Methods Return Lists of Target Nodes
2. Explicitly Creating File and Directory Nodes
3. Printing Node File Names
4. Using a Node's File Name as a String
5. GetBuildPath: Getting the Path From a Node or String
6. Dependencies
1. Deciding When an Input File Has Changed: the Decider Function
1.1. Using MD5 Signatures to Decide if a File Has Changed
1.2. Using Time Stamps to Decide If a File Has Changed
1.3. Deciding If a File Has Changed Using Both MD Signatures and Time Stamps
1.4. Writing Your Own Custom Decider Function
1.5. Mixing Different Ways of Deciding If a File Has Changed
2. Older Functions for Deciding When an Input File Has Changed
2.1. The SourceSignatures Function
2.2. The TargetSignatures Function
3. Implicit Dependencies: The $CPPPATH Construction Variable
4. Caching Implicit Dependencies
4.1. The --implicit-deps-changed Option
4.2. The --implicit-deps-unchanged Option
5. Explicit Dependencies: the Depends Function
6. Dependencies From External Files: the ParseDepends Function
7. Ignoring Dependencies: the Ignore Function
8. Order-Only Dependencies: the Requires Function
9. The AlwaysBuild Function
7. Environments
1. Using Values From the External Environment
2. Construction Environments
2.1. Creating a Construction Environment: the Environment Function
2.2. Fetching Values From a Construction Environment
2.3. Expanding Values From a Construction Environment: the subst Method
2.4. Handling Problems With Value Expansion
2.5. Controlling the Default Construction Environment: the DefaultEnvironment Function
2.6. Multiple Construction Environments
2.7. Making Copies of Construction Environments: the Clone Method
2.8. Replacing Values: the Replace Method
2.9. Setting Values Only If They're Not Already Defined: the SetDefault Method
2.10. Appending to the End of Values: the Append Method
2.11. Appending Unique Values: the AppendUnique Method
2.12. Appending to the Beginning of Values: the Prepend Method
2.13. Prepending Unique Values: the PrependUnique Method
3. Controlling the Execution Environment for Issued Commands
3.1. Propagating PATH From the External Environment
3.2. Adding to PATH Values in the Execution Environment
8. Automatically Putting Command-line Options into their Construction Variables
1. Merging Options into the Environment: the MergeFlags Function
2. Separating Compile Arguments into their Variables: the ParseFlags Function
3. Finding Installed Library Information: the ParseConfig Function
9. Controlling Build Output
1. Providing Build Help: the Help Function
2. Controlling How SCons Prints Build Commands: the $*COMSTR Variables
3. Providing Build Progress Output: the Progress Function
4. Printing Detailed Build Status: the GetBuildFailures Function
10. Controlling a Build From the Command Line
1. Command-Line Options
1.1. Not Having to Specify Command-Line Options Each Time: the SCONSFLAGS Environment Variable
1.2. Getting Values Set by Command-Line Options: the GetOption Function
1.3. Setting Values of Command-Line Options: the SetOption Function
1.4. Strings for Getting or Setting Values of SCons Command-Line Options
1.5. Adding Custom Command-Line Options: the AddOption Function
2. Command-Line variable=value Build Variables
2.1. Controlling Command-Line Build Variables
2.2. Providing Help for Command-Line Build Variables
2.3. Reading Build Variables From a File
2.4. Pre-Defined Build Variable Functions
2.5. Adding Multiple Command-Line Build Variables at Once
2.6. Handling Unknown Command-Line Build Variables: the UnknownVariables Function
3. Command-Line Targets
3.1. Fetching Command-Line Targets: the COMMAND_LINE_TARGETS Variable
3.2. Controlling the Default Targets: the Default Function
3.3. Fetching the List of Build Targets, Regardless of Origin: the BUILD_TARGETS Variable
11. Installing Files in Other Directories: the Install Builder
1. Installing Multiple Files in a Directory
2. Installing a File Under a Different Name
3. Installing Multiple Files Under Different Names
12. Platform-Independent File System Manipulation
1. Copying Files or Directories: The Copy Factory
2. Deleting Files or Directories: The Delete Factory
3. Moving (Renaming) Files or Directories: The Move Factory
4. Updating the Modification Time of a File: The Touch Factory
5. Creating a Directory: The Mkdir Factory
6. Changing File or Directory Permissions: The Chmod Factory
7. Executing an action immediately: the Execute Function
13. Controlling Removal of Targets
1. Preventing target removal during build: the Precious Function
2. Preventing target removal during clean: the NoClean Function
3. Removing additional files during clean: the Clean Function
14. Hierarchical Builds
1. SConscript Files
2. Path Names Are Relative to the SConscript Directory
3. Top-Level Path Names in Subsidiary SConscript Files
4. Absolute Path Names
5. Sharing Environments (and Other Variables) Between SConscript Files
5.1. Exporting Variables
5.2. Importing Variables
5.3. Returning Values From an SConscript File
15. Separating Source and Build Directories
1. Specifying a Variant Directory Tree as Part of an SConscript Call
2. Why SCons Duplicates Source Files in a Variant Directory Tree
3. Telling SCons to Not Duplicate Source Files in the Variant Directory Tree
4. The VariantDir Function
5. Using VariantDir With an SConscript File
6. Using Glob with VariantDir
16. Variant Builds
17. Internationalization and localization with gettext
1. Prerequisites
2. Simple project
18. Writing Your Own Builders
1. Writing Builders That Execute External Commands
2. Attaching a Builder to a Construction Environment
3. Letting SCons Handle The File Suffixes
4. Builders That Execute Python Functions
5. Builders That Create Actions Using a Generator
6. Builders That Modify the Target or Source Lists Using an Emitter
7. Where To Put Your Custom Builders and Tools
19. Not Writing a Builder: the Command Builder
20. Pseudo-Builders: the AddMethod function
21. Writing Scanners
1. A Simple Scanner Example
2. Adding a search path to a scanner: FindPathDirs
22. Building From Code Repositories
1. The Repository Method
2. Finding source files in repositories
3. Finding #include files in repositories
3.1. Limitations on #include files in repositories
4. Finding the SConstruct file in repositories
5. Finding derived files in repositories
6. Guaranteeing local copies of files
23. Multi-Platform Configuration (Autoconf Functionality)
1. Configure Contexts
2. Checking for the Existence of Header Files
3. Checking for the Availability of a Function
4. Checking for the Availability of a Library
5. Checking for the Availability of a typedef
6. Adding Your Own Custom Checks
7. Not Configuring When Cleaning Targets
24. Caching Built Files
1. Specifying the Shared Cache Directory
2. Keeping Build Output Consistent
3. Not Using the Shared Cache for Specific Files
4. Disabling the Shared Cache
5. Populating a Shared Cache With Already-Built Files
6. Minimizing Cache Contention: the --random Option
25. Alias Targets
26. Java Builds
1. Building Java Class Files: the Java Builder
2. How SCons Handles Java Dependencies
3. Building Java Archive (.jar) Files: the Jar Builder
4. Building C Header and Stub Files: the JavaH Builder
5. Building RMI Stub and Skeleton Class Files: the RMIC Builder
27. Miscellaneous Functionality
1. Verifying the Python Version: the EnsurePythonVersion Function
2. Verifying the SCons Version: the EnsureSConsVersion Function
3. Explicitly Terminating SCons While Reading SConscript Files: the Exit Function
4. Searching for Files: the FindFile Function
5. Handling Nested Lists: the Flatten Function
6. Finding the Invocation Directory: the GetLaunchDir Function
28. Troubleshooting
1. Why is That Target Being Rebuilt? the --debug=explain Option
2. What's in That Construction Environment? the Dump Method
3. What Dependencies Does SCons Know About? the --tree Option
4. How is SCons Constructing the Command Lines It Executes? the --debug=presub Option
5. Where is SCons Searching for Libraries? the --debug=findlibs Option
6. Where is SCons Blowing Up? the --debug=stacktrace Option
7. How is SCons Making Its Decisions? the --taskmastertrace Option
8. Watch SCons prepare targets for building: the --debug=prepare Option
9. Why is a file disappearing? the --debug=duplicate Option
A. Construction Variables
B. Builders
C. Tools
D. Functions and Environment Methods
E. Handling Common Tasks

List of Examples

E.1. Wildcard globbing to create a list of filenames
E.2. Filename extension substitution
E.3. Appending a path prefix to a list of filenames
E.4. Substituting a path prefix with another one
E.5. Filtering a filename list to exclude/retain only a specific set of extensions
E.6. The "backtick function": run a shell command and capture the output
E.7. Generating source code: how code can be generated and used by SCons

Thank you for taking the time to read about SCons. SCons is a next-generation software construction tool, or make tool--that is, a software utility for building software (or other files) and keeping built software up-to-date whenever the underlying input files change.

The most distinctive thing about SCons is that its configuration files are actually scripts, written in the Python programming language. This is in contrast to most alternative build tools, which typically invent a new language to configure the build. SCons still has a learning curve, of course, because you have to know what functions to call to set up your build properly, but the underlying syntax used should be familiar to anyone who has ever looked at a Python script.

Paradoxically, using Python as the configuration file format makes SCons easier for non-programmers to learn than the cryptic languages of other build tools, which are usually invented by programmers for other programmers. This is in no small part due to the consistency and readability that are hallmarks of Python. It just so happens that making a real, live scripting language the basis for the configuration files makes it a snap for more accomplished programmers to do more complicated things with builds, as necessary.

SCons would not exist without a lot of help from a lot of people, many of whom may not even be aware that they helped or served as inspiration. So in no particular order, and at the risk of leaving out someone:

First and foremost, SCons owes a tremendous debt to Bob Sidebotham, the original author of the classic Perl-based Cons tool which Bob first released to the world back around 1996. Bob's work on Cons classic provided the underlying architecture and model of specifying a build configuration using a real scripting language. My real-world experience working on Cons informed many of the design decisions in SCons, including the improved parallel build support, making Builder objects easily definable by users, and separating the build engine from the wrapping interface.

Greg Wilson was instrumental in getting SCons started as a real project when he initiated the Software Carpentry design competition in February 2000. Without that nudge, marrying the advantages of the Cons classic architecture with the readability of Python might have just stayed no more than a nice idea.

The entire SCons team have been absolutely wonderful to work with, and SCons would be nowhere near as useful a tool without the energy, enthusiasm and time people have contributed over the past few years. The "core team" of Chad Austin, Anthony Roach, Bill Deegan, Charles Crain, Steve Leblanc, Greg Noel, Gary Oberbrunner, Greg Spencer and Christoph Wiedemann have been great about reviewing my (and other) changes and catching problems before they get in the code base. Of particular technical note: Anthony's outstanding and innovative work on the tasking engine has given SCons a vastly superior parallel build model; Charles has been the master of the crucial Node infrastructure; Christoph's work on the Configure infrastructure has added crucial Autoconf-like functionality; and Greg has provided excellent support for Microsoft Visual Studio.

Special thanks to David Snopek for contributing his underlying "Autoscons" code that formed the basis of Christoph's work with the Configure functionality. David was extremely generous in making this code available to SCons, given that he initially released it under the GPL and SCons is released under a less-restrictive MIT-style license.

Thanks to Peter Miller for his splendid change management system, Aegis, which has provided the SCons project with a robust development methodology from day one, and which showed me how you could integrate incremental regression tests into a practical development cycle (years before eXtreme Programming arrived on the scene).

And last, thanks to Guido van Rossum for his elegant scripting language, which is the basis not only for the SCons implementation, but for the interface itself.

This chapter will take you through the basic steps of installing SCons on your system, and building SCons if you don't have a pre-built package available (or simply prefer the flexibility of building it yourself). Before that, however, this chapter will also describe the basic steps involved in installing Python on your system, in case that is necessary. Fortunately, both SCons and Python are very easy to install on almost any system, and Python already comes installed on many systems.

SCons comes pre-packaged for installation on a number of systems, including Linux and Windows systems. You do not need to read this entire section, you should need to read only the section appropriate to the type of system you're running on.

SCons comes in RPM (Red Hat Package Manager) format, pre-built and ready to install on Red Hat Linux, Fedora, or any other Linux distribution that uses RPM. Your distribution may already have an SCons RPM built specifically for it; many do, including SUSE, Mandrake and Fedora. You can check for the availability of an SCons RPM on your distribution's download servers, or by consulting an RPM search site like http://www.rpmfind.net/ or http://rpm.pbone.net/.

If your distribution supports installation via yum, you should be able to install SCons by running:

# yum install scons
      

If your Linux distribution does not already have a specific SCons RPM file, you can download and install from the generic RPM provided by the SCons project. This will install the SCons script(s) in /usr/bin, and the SCons library modules in /usr/lib/scons.

To install from the command line, simply download the appropriate .rpm file, and then run:

# rpm -Uvh scons-2.3.1-1.noarch.rpm
      

Or, you can use a graphical RPM package manager. See your package manager application's documention for specific instructions about how to use it to install a downloaded RPM.

SCons provides a Windows installer that makes installation extremely easy. Download the scons-2.3.1.win32.exe file from the SCons download page at http://www.scons.org/download.php. Then all you need to do is execute the file (usually by clicking on its icon in Windows Explorer). These will take you through a small sequence of windows that will install SCons on your system.

If a pre-built SCons package is not available for your system, then you can still easily build and install SCons using the native Python distutils package.

The first step is to download either the scons-2.3.1.tar.gz or scons-2.3.1.zip, which are available from the SCons download page at http://www.scons.org/download.html.

Unpack the archive you downloaded, using a utility like tar on Linux or UNIX, or WinZip on Windows. This will create a directory called scons-2.3.1, usually in your local directory. Then change your working directory to that directory and install SCons by executing the following commands:

# cd scons-2.3.1
# python setup.py install
    

This will build SCons, install the scons script in the python which is used to run the setup.py's scripts directory (/usr/local/bin or C:\Python25\Scripts), and will install the SCons build engine in the corresponding library directory for the python used (/usr/local/lib/scons or C:\Python25\scons). Because these are system directories, you may need root (on Linux or UNIX) or Administrator (on Windows) privileges to install SCons like this.

In this chapter, you will see several examples of very simple build configurations using SCons, which will demonstrate how easy it is to use SCons to build programs from several different programming languages on different types of systems.

Here's the famous "Hello, World!" program in C:

int
main()
{
    printf("Hello, world!\n");
}
   

And here's how to build it using SCons. Enter the following into a file named SConstruct:

Program('hello.c')
      

This minimal configuration file gives SCons two pieces of information: what you want to build (an executable program), and the input file from which you want it built (the hello.c file). Program is a builder_method, a Python call that tells SCons that you want to build an executable program.

That's it. Now run the scons command to build the program. On a POSIX-compliant system like Linux or UNIX, you'll see something like:

% scons
scons: Reading SConscript files ...
scons: done reading SConscript files.
scons: Building targets ...
cc -o hello.o -c hello.c
cc -o hello hello.o
scons: done building targets.

On a Windows system with the Microsoft Visual C++ compiler, you'll see something like:

C:\>scons
scons: Reading SConscript files ...
scons: done reading SConscript files.
scons: Building targets ...
cl /Fohello.obj /c hello.c /nologo
link /nologo /OUT:hello.exe hello.obj
embedManifestExeCheck(target, source, env)
scons: done building targets.

First, notice that you only need to specify the name of the source file, and that SCons correctly deduces the names of the object and executable files to be built from the base of the source file name.

Second, notice that the same input SConstruct file, without any changes, generates the correct output file names on both systems: hello.o and hello on POSIX systems, hello.obj and hello.exe on Windows systems. This is a simple example of how SCons makes it extremely easy to write portable software builds.

(Note that we won't provide duplicate side-by-side POSIX and Windows output for all of the examples in this guide; just keep in mind that, unless otherwise specified, any of the examples should work equally well on both types of systems.)

The Program builder method is only one of many builder methods that SCons provides to build different types of files. Another is the Object builder method, which tells SCons to build an object file from the specified source file:

Object('hello.c')
      

Now when you run the scons command to build the program, it will build just the hello.o object file on a POSIX system:

% scons
scons: Reading SConscript files ...
scons: done reading SConscript files.
scons: Building targets ...
cc -o hello.o -c hello.c
scons: done building targets.

And just the hello.obj object file on a Windows system (with the Microsoft Visual C++ compiler):

C:\>scons
scons: Reading SConscript files ...
scons: done reading SConscript files.
scons: Building targets ...
cl /Fohello.obj /c hello.c /nologo
scons: done building targets.

SCons also makes building with Java extremely easy. Unlike the Program and Object builder methods, however, the Java builder method requires that you specify the name of a destination directory in which you want the class files placed, followed by the source directory in which the .java files live:

Java('classes', 'src')
     

If the src directory contains a single hello.java file, then the output from running the scons command would look something like this (on a POSIX system):

% scons
scons: Reading SConscript files ...
scons: done reading SConscript files.
scons: Building targets ...
javac -d classes -sourcepath src src/hello.java
scons: done building targets.

We'll cover Java builds in more detail, including building Java archive (.jar) and other types of file, in Chapter 26, Java Builds.

If you're used to build systems like Make you've already figured out that the SConstruct file is the SCons equivalent of a Makefile. That is, the SConstruct file is the input file that SCons reads to control the build.

One important way in which the SConstruct file is not exactly like a normal Python script, and is more like a Makefile, is that the order in which the SCons functions are called in the SConstruct file does not affect the order in which SCons actually builds the programs and object files you want it to build.[1] In other words, when you call the Program builder (or any other builder method), you're not telling SCons to build the program at the instant the builder method is called. Instead, you're telling SCons to build the program that you want, for example, a program built from a file named hello.c, and it's up to SCons to build that program (and any other files) whenever it's necessary. (We'll learn more about how SCons decides when building or rebuilding a file is necessary in Chapter 6, Dependencies, below.)

SCons reflects this distinction between calling a builder method like Program and actually building the program by printing the status messages that indicate when it's "just reading" the SConstruct file, and when it's actually building the target files. This is to make it clear when SCons is executing the Python statements that make up the SConstruct file, and when SCons is actually executing the commands or other actions to build the necessary files.

Let's clarify this with an example. Python has a print statement that prints a string of characters to the screen. If we put print statements around our calls to the Program builder method:

print "Calling Program('hello.c')"
Program('hello.c')
print "Calling Program('goodbye.c')"
Program('goodbye.c')
print "Finished calling Program()"
       

Then when we execute SCons, we see the output from the print statements in between the messages about reading the SConscript files, indicating that that is when the Python statements are being executed:

% scons
scons: Reading SConscript files ...
Calling Program('hello.c')
Calling Program('goodbye.c')
Finished calling Program()
scons: done reading SConscript files.
scons: Building targets ...
cc -o goodbye.o -c goodbye.c
cc -o goodbye goodbye.o
cc -o hello.o -c hello.c
cc -o hello hello.o
scons: done building targets.

Notice also that SCons built the goodbye program first, even though the "reading SConscript" output shows that we called Program('hello.c') first in the SConstruct file.



[1] In programming parlance, the SConstruct file is declarative, meaning you tell SCons what you want done and let it figure out the order in which to do it, rather than strictly imperative, where you specify explicitly the order in which to do things.

In this chapter, you will see several examples of very simple build configurations using SCons, which will demonstrate how easy it is to use SCons to build programs from several different programming languages on different types of systems.

You've seen that when you call the Program builder method, it builds the resulting program with the same base name as the source file. That is, the following call to build an executable program from the hello.c source file will build an executable program named hello on POSIX systems, and an executable program named hello.exe on Windows systems:

Program('hello.c')
    

If you want to build a program with a different name than the base of the source file name, you simply put the target file name to the left of the source file name:

Program('new_hello', 'hello.c')
       

(SCons requires the target file name first, followed by the source file name, so that the order mimics that of an assignment statement in most programming languages, including Python: "program = source files".)

Now SCons will build an executable program named new_hello when run on a POSIX system:

% scons -Q
cc -o hello.o -c hello.c
cc -o new_hello hello.o

And SCons will build an executable program named new_hello.exe when run on a Windows system:

C:\>scons -Q
cl /Fohello.obj /c hello.c /nologo
link /nologo /OUT:new_hello.exe hello.obj
embedManifestExeCheck(target, source, env)

You've just seen how to configure SCons to compile a program from a single source file. It's more common, of course, that you'll need to build a program from many input source files, not just one. To do this, you need to put the source files in a Python list (enclosed in square brackets), like so:

Program(['prog.c', 'file1.c', 'file2.c'])
       

A build of the above example would look like:

% scons -Q
cc -o file1.o -c file1.c
cc -o file2.o -c file2.c
cc -o prog.o -c prog.c
cc -o prog prog.o file1.o file2.o

Notice that SCons deduces the output program name from the first source file specified in the list--that is, because the first source file was prog.c, SCons will name the resulting program prog (or prog.exe on a Windows system). If you want to specify a different program name, then (as we've seen in the previous section) you slide the list of source files over to the right to make room for the output program file name. (SCons puts the output file name to the left of the source file names so that the order mimics that of an assignment statement: "program = source files".) This makes our example:

Program('program', ['prog.c', 'file1.c', 'file2.c'])
       

On Linux, a build of this example would look like:

% scons -Q
cc -o file1.o -c file1.c
cc -o file2.o -c file2.c
cc -o prog.o -c prog.c
cc -o program prog.o file1.o file2.o

Or on Windows:

C:\>scons -Q
cl /Fofile1.obj /c file1.c /nologo
cl /Fofile2.obj /c file2.c /nologo
cl /Foprog.obj /c prog.c /nologo
link /nologo /OUT:program.exe prog.obj file1.obj file2.obj
embedManifestExeCheck(target, source, env)

One drawback to the use of a Python list for source files is that each file name must be enclosed in quotes (either single quotes or double quotes). This can get cumbersome and difficult to read when the list of file names is long. Fortunately, SCons and Python provide a number of ways to make sure that the SConstruct file stays easy to read.

To make long lists of file names easier to deal with, SCons provides a Split function that takes a quoted list of file names, with the names separated by spaces or other white-space characters, and turns it into a list of separate file names. Using the Split function turns the previous example into:

Program('program', Split('main.c file1.c file2.c'))
    

(If you're already familiar with Python, you'll have realized that this is similar to the split() method in the Python standard string module. Unlike the split() member function of strings, however, the Split function does not require a string as input and will wrap up a single non-string object in a list, or return its argument untouched if it's already a list. This comes in handy as a way to make sure arbitrary values can be passed to SCons functions without having to check the type of the variable by hand.)

Putting the call to the Split function inside the Program call can also be a little unwieldy. A more readable alternative is to assign the output from the Split call to a variable name, and then use the variable when calling the Program function:

src_files = Split('main.c file1.c file2.c')
Program('program', src_files)
    

Lastly, the Split function doesn't care how much white space separates the file names in the quoted string. This allows you to create lists of file names that span multiple lines, which often makes for easier editing:

src_files = Split("""main.c
                     file1.c
                     file2.c""")
Program('program', src_files)
    

(Note in this example that we used the Python "triple-quote" syntax, which allows a string to contain multiple lines. The three quotes can be either single or double quotes.)

It's common to re-use code by sharing source files between multiple programs. One way to do this is to create a library from the common source files, which can then be linked into resulting programs. (Creating libraries is discussed in Chapter 4, Building and Linking with Libraries, below.)

A more straightforward, but perhaps less convenient, way to share source files between multiple programs is simply to include the common files in the lists of source files for each program:

Program(Split('foo.c common1.c common2.c'))
Program('bar', Split('bar1.c bar2.c common1.c common2.c'))
       

SCons recognizes that the object files for the common1.c and common2.c source files each need to be built only once, even though the resulting object files are each linked in to both of the resulting executable programs:

% scons -Q
cc -o bar1.o -c bar1.c
cc -o bar2.o -c bar2.c
cc -o common1.o -c common1.c
cc -o common2.o -c common2.c
cc -o bar bar1.o bar2.o common1.o common2.o
cc -o foo.o -c foo.c
cc -o foo foo.o common1.o common2.o

If two or more programs share a lot of common source files, repeating the common files in the list for each program can be a maintenance problem when you need to change the list of common files. You can simplify this by creating a separate Python list to hold the common file names, and concatenating it with other lists using the Python + operator:

common = ['common1.c', 'common2.c']
foo_files = ['foo.c'] + common
bar_files = ['bar1.c', 'bar2.c'] + common
Program('foo', foo_files)
Program('bar', bar_files)
    

This is functionally equivalent to the previous example.

It's often useful to organize large software projects by collecting parts of the software into one or more libraries. SCons makes it easy to create libraries and to use them in the programs.

You build your own libraries by specifying Library instead of Program:

Library('foo', ['f1.c', 'f2.c', 'f3.c'])
      

SCons uses the appropriate library prefix and suffix for your system. So on POSIX or Linux systems, the above example would build as follows (although ranlib may not be called on all systems):

% scons -Q
cc -o f1.o -c f1.c
cc -o f2.o -c f2.c
cc -o f3.o -c f3.c
ar rc libfoo.a f1.o f2.o f3.o
ranlib libfoo.a

On a Windows system, a build of the above example would look like:

C:\>scons -Q
cl /Fof1.obj /c f1.c /nologo
cl /Fof2.obj /c f2.c /nologo
cl /Fof3.obj /c f3.c /nologo
lib /nologo /OUT:foo.lib f1.obj f2.obj f3.obj

The rules for the target name of the library are similar to those for programs: if you don't explicitly specify a target library name, SCons will deduce one from the name of the first source file specified, and SCons will add an appropriate file prefix and suffix if you leave them off.

The previous example shows building a library from a list of source files. You can, however, also give the Library call object files, and it will correctly realize In fact, you can arbitrarily mix source code files and object files in the source list:

Library('foo', ['f1.c', 'f2.o', 'f3.c', 'f4.o'])
        

And SCons realizes that only the source code files must be compiled into object files before creating the final library:

% scons -Q
cc -o f1.o -c f1.c
cc -o f3.o -c f3.c
ar rc libfoo.a f1.o f2.o f3.o f4.o
ranlib libfoo.a

Of course, in this example, the object files must already exist for the build to succeed. See Chapter 5, Node Objects, below, for information about how you can build object files explicitly and include the built files in a library.

The Library function builds a traditional static library. If you want to be explicit about the type of library being built, you can use the synonym StaticLibrary function instead of Library:

StaticLibrary('foo', ['f1.c', 'f2.c', 'f3.c'])
        

There is no functional difference between the StaticLibrary and Library functions.

If you want to build a shared library (on POSIX systems) or a DLL file (on Windows systems), you use the SharedLibrary function:

SharedLibrary('foo', ['f1.c', 'f2.c', 'f3.c'])
        

The output on POSIX:

% scons -Q
cc -o f1.os -c f1.c
cc -o f2.os -c f2.c
cc -o f3.os -c f3.c
cc -o libfoo.so -shared f1.os f2.os f3.os

And the output on Windows:

C:\>scons -Q
cl /Fof1.obj /c f1.c /nologo
cl /Fof2.obj /c f2.c /nologo
cl /Fof3.obj /c f3.c /nologo
link /nologo /dll /out:foo.dll /implib:foo.lib f1.obj f2.obj f3.obj
RegServerFunc(target, source, env)
embedManifestDllCheck(target, source, env)

Notice again that SCons takes care of building the output file correctly, adding the -shared option for a POSIX compilation, and the /dll option on Windows.

Usually, you build a library because you want to link it with one or more programs. You link libraries with a program by specifying the libraries in the $LIBS construction variable, and by specifying the directory in which the library will be found in the $LIBPATH construction variable:

Library('foo', ['f1.c', 'f2.c', 'f3.c'])
Program('prog.c', LIBS=['foo', 'bar'], LIBPATH='.')
      

Notice, of course, that you don't need to specify a library prefix (like lib) or suffix (like .a or .lib). SCons uses the correct prefix or suffix for the current system.

On a POSIX or Linux system, a build of the above example would look like:

% scons -Q
cc -o f1.o -c f1.c
cc -o f2.o -c f2.c
cc -o f3.o -c f3.c
ar rc libfoo.a f1.o f2.o f3.o
ranlib libfoo.a
cc -o prog.o -c prog.c
cc -o prog prog.o -L. -lfoo -lbar

On a Windows system, a build of the above example would look like:

C:\>scons -Q
cl /Fof1.obj /c f1.c /nologo
cl /Fof2.obj /c f2.c /nologo
cl /Fof3.obj /c f3.c /nologo
lib /nologo /OUT:foo.lib f1.obj f2.obj f3.obj
cl /Foprog.obj /c prog.c /nologo
link /nologo /OUT:prog.exe /LIBPATH:. foo.lib bar.lib prog.obj
embedManifestExeCheck(target, source, env)

As usual, notice that SCons has taken care of constructing the correct command lines to link with the specified library on each system.

Note also that, if you only have a single library to link with, you can specify the library name in single string, instead of a Python list, so that:

Program('prog.c', LIBS='foo', LIBPATH='.')
    

is equivalent to:

Program('prog.c', LIBS=['foo'], LIBPATH='.')
    

This is similar to the way that SCons handles either a string or a list to specify a single source file.

By default, the linker will only look in certain system-defined directories for libraries. SCons knows how to look for libraries in directories that you specify with the $LIBPATH construction variable. $LIBPATH consists of a list of directory names, like so:

Program('prog.c', LIBS = 'm',
                  LIBPATH = ['/usr/lib', '/usr/local/lib'])
      

Using a Python list is preferred because it's portable across systems. Alternatively, you could put all of the directory names in a single string, separated by the system-specific path separator character: a colon on POSIX systems:

LIBPATH = '/usr/lib:/usr/local/lib'
    

or a semi-colon on Windows systems:

LIBPATH = 'C:\\lib;D:\\lib'
    

(Note that Python requires that the backslash separators in a Windows path name be escaped within strings.)

When the linker is executed, SCons will create appropriate flags so that the linker will look for libraries in the same directories as SCons. So on a POSIX or Linux system, a build of the above example would look like:

% scons -Q
cc -o prog.o -c prog.c
cc -o prog prog.o -L/usr/lib -L/usr/local/lib -lm

On a Windows system, a build of the above example would look like:

C:\>scons -Q
cl /Foprog.obj /c prog.c /nologo
link /nologo /OUT:prog.exe /LIBPATH:\usr\lib /LIBPATH:\usr\local\lib m.lib prog.obj
embedManifestExeCheck(target, source, env)

Note again that SCons has taken care of the system-specific details of creating the right command-line options.

Internally, SCons represents all of the files and directories it knows about as Nodes. These internal objects (not object files) can be used in a variety of ways to make your SConscript files portable and easy to read.

All builder methods return a list of Node objects that identify the target file or files that will be built. These returned Nodes can be passed as arguments to other builder methods.

For example, suppose that we want to build the two object files that make up a program with different options. This would mean calling the Object builder once for each object file, specifying the desired options:

Object('hello.c', CCFLAGS='-DHELLO')
Object('goodbye.c', CCFLAGS='-DGOODBYE')
    

One way to combine these object files into the resulting program would be to call the Program builder with the names of the object files listed as sources:

Object('hello.c', CCFLAGS='-DHELLO')
Object('goodbye.c', CCFLAGS='-DGOODBYE')
Program(['hello.o', 'goodbye.o'])
    

The problem with specifying the names as strings is that our SConstruct file is no longer portable across operating systems. It won't, for example, work on Windows because the object files there would be named hello.obj and goodbye.obj, not hello.o and goodbye.o.

A better solution is to assign the lists of targets returned by the calls to the Object builder to variables, which we can then concatenate in our call to the Program builder:

hello_list = Object('hello.c', CCFLAGS='-DHELLO')
goodbye_list = Object('goodbye.c', CCFLAGS='-DGOODBYE')
Program(hello_list + goodbye_list)
      

This makes our SConstruct file portable again, the build output on Linux looking like:

% scons -Q
cc -o goodbye.o -c -DGOODBYE goodbye.c
cc -o hello.o -c -DHELLO hello.c
cc -o hello hello.o goodbye.o

And on Windows:

C:\>scons -Q
cl /Fogoodbye.obj /c goodbye.c -DGOODBYE
cl /Fohello.obj /c hello.c -DHELLO
link /nologo /OUT:hello.exe hello.obj goodbye.obj
embedManifestExeCheck(target, source, env)

We'll see examples of using the list of nodes returned by builder methods throughout the rest of this guide.

So far we've seen how SCons handles one-time builds. But one of the main functions of a build tool like SCons is to rebuild only what is necessary when source files change--or, put another way, SCons should not waste time rebuilding things that don't need to be rebuilt. You can see this at work simply by re-invoking SCons after building our simple hello example:

% scons -Q
cc -o hello.o -c hello.c
cc -o hello hello.o
% scons -Q
scons: `.' is up to date.

The second time it is executed, SCons realizes that the hello program is up-to-date with respect to the current hello.c source file, and avoids rebuilding it. You can see this more clearly by naming the hello program explicitly on the command line:

% scons -Q hello
cc -o hello.o -c hello.c
cc -o hello hello.o
% scons -Q hello
scons: `hello' is up to date.

Note that SCons reports "...is up to date" only for target files named explicitly on the command line, to avoid cluttering the output.

Another aspect of avoiding unnecessary rebuilds is the fundamental build tool behavior of rebuilding things when an input file changes, so that the built software is up to date. By default, SCons keeps track of this through an MD5 signature, or checksum, of the contents of each file, although you can easily configure SCons to use the modification times (or time stamps) instead. You can even specify your own Python function for deciding if an input file has changed.

By default, SCons keeps track of whether a file has changed based on an MD5 checksum of the file's contents, not the file's modification time. This means that you may be surprised by the default SCons behavior if you are used to the Make convention of forcing a rebuild by updating the file's modification time (using the touch command, for example):

% scons -Q hello
cc -o hello.o -c hello.c
cc -o hello hello.o
% touch hello.c
% scons -Q hello
scons: `hello' is up to date.

Even though the file's modification time has changed, SCons realizes that the contents of the hello.c file have not changed, and therefore that the hello program need not be rebuilt. This avoids unnecessary rebuilds when, for example, someone rewrites the contents of a file without making a change. But if the contents of the file really do change, then SCons detects the change and rebuilds the program as required:

% scons -Q hello
cc -o hello.o -c hello.c
cc -o hello hello.o
%     [CHANGE THE CONTENTS OF hello.c]
% scons -Q hello
cc -o hello.o -c hello.c
cc -o hello hello.o

Note that you can, if you wish, specify this default behavior (MD5 signatures) explicitly using the Decider function as follows:

Program('hello.c')
Decider('MD5')
      

You can also use the string 'content' as a synonym for 'MD5' when calling the Decider function.

If you prefer, you can configure SCons to use the modification time of a file, not the file contents, when deciding if a target needs to be rebuilt. SCons gives you two ways to use time stamps to decide if an input file has changed since the last time a target has been built.

The most familiar way to use time stamps is the way Make does: that is, have SCons decide that a target must be rebuilt if a source file's modification time is newer than the target file. To do this, call the Decider function as follows:

Object('hello.c')
Decider('timestamp-newer')
        

This makes SCons act like Make when a file's modification time is updated (using the touch command, for example):

% scons -Q hello.o
cc -o hello.o -c hello.c
% touch hello.c
% scons -Q hello.o
cc -o hello.o -c hello.c

And, in fact, because this behavior is the same as the behavior of Make, you can also use the string 'make' as a synonym for 'timestamp-newer' when calling the Decider function:

Object('hello.c')
Decider('make')
      

One drawback to using times stamps exactly like Make is that if an input file's modification time suddenly becomes older than a target file, the target file will not be rebuilt. This can happen if an old copy of a source file is restored from a backup archive, for example. The contents of the restored file will likely be different than they were the last time a dependent target was built, but the target won't be rebuilt because the modification time of the source file is not newer than the target.

Because SCons actually stores information about the source files' time stamps whenever a target is built, it can handle this situation by checking for an exact match of the source file time stamp, instead of just whether or not the source file is newer than the target file. To do this, specify the argument 'timestamp-match' when calling the Decider function:

Object('hello.c')
Decider('timestamp-match')
        

When configured this way, SCons will rebuild a target whenever a source file's modification time has changed. So if we use the touch -t option to change the modification time of hello.c to an old date (January 1, 1989), SCons will still rebuild the target file:

% scons -Q hello.o
cc -o hello.o -c hello.c
% touch -t 198901010000 hello.c
% scons -Q hello.o
cc -o hello.o -c hello.c

In general, the only reason to prefer timestamp-newer instead of timestamp-match, would be if you have some specific reason to require this Make-like behavior of not rebuilding a target when an otherwise-modified source file is older.

As a performance enhancement, SCons provides a way to use MD5 checksums of file contents but to read those contents only when the file's timestamp has changed. To do this, call the Decider function with 'MD5-timestamp' argument as follows:

Program('hello.c')
Decider('MD5-timestamp')
        

So configured, SCons will still behave like it does when using Decider('MD5'):

% scons -Q hello
cc -o hello.o -c hello.c
cc -o hello hello.o
% touch hello.c
% scons -Q hello
scons: `hello' is up to date.
% edit hello.c
    [CHANGE THE CONTENTS OF hello.c]
% scons -Q hello
cc -o hello.o -c hello.c
cc -o hello hello.o
      

However, the second call to SCons in the above output, when the build is up-to-date, will have been performed by simply looking at the modification time of the hello.c file, not by opening it and performing an MD5 checksum calcuation on its contents. This can significantly speed up many up-to-date builds.

The only drawback to using Decider('MD5-timestamp') is that SCons will not rebuild a target file if a source file was modified within one second of the last time SCons built the file. While most developers are programming, this isn't a problem in practice, since it's unlikely that someone will have built and then thought quickly enough to make a substantive change to a source file within one second. Certain build scripts or continuous integration tools may, however, rely on the ability to apply changes to files automatically and then rebuild as quickly as possible, in which case use of Decider('MD5-timestamp') may not be appropriate.

The different string values that we've passed to the Decider function are essentially used by SCons to pick one of several specific internal functions that implement various ways of deciding if a dependency (usually a source file) has changed since a target file has been built. As it turns out, you can also supply your own function to decide if a dependency has changed.

For example, suppose we have an input file that contains a lot of data, in some specific regular format, that is used to rebuild a lot of different target files, but each target file really only depends on one particular section of the input file. We'd like to have each target file depend on only its section of the input file. However, since the input file may contain a lot of data, we want to open the input file only if its timestamp has changed. This could be done with a custom Decider function that might look something like this:

Program('hello.c')
def decide_if_changed(dependency, target, prev_ni):
    if self.get_timestamp() != prev_ni.timestamp:
        dep = str(dependency)
        tgt = str(target)
        if specific_part_of_file_has_changed(dep, tgt):
            return True
    return False
Decider(decide_if_changed)
        

Note that in the function definition, the dependency (input file) is the first argument, and then the target. Both of these are passed to the functions as SCons Node objects, which we convert to strings using the Python str().

The third argument, prev_ni, is an object that holds the signature or timestamp information that was recorded about the dependency the last time the target was built. A prev_ni object can hold different information, depending on the type of thing that the dependency argument represents. For normal files, the prev_ni object has the following attributes:

.csig

The content signature, or MD5 checksum, of the contents of the dependency file the list time the target was built.

.size

The size in bytes of the dependency file the list time the target was built.

.timestamp

The modification time of the dependency file the list time the target was built.

Note that ignoring some of the arguments in your custom Decider function is a perfectly normal thing to do, if they don't impact the way you want to decide if the dependency file has changed.

Another thing to look out for is the fact that the three attributes above may not be present at the time of the first run. Without any prior build, no targets have been created and no .sconsign DB file exists yet. So, you should always check whether the prev_ni attribute in question is available.

We finally present a small example for a csig-based decider function. Note how the signature information for the dependency file has to get initialized via get_csig during each function call (this is mandatory!).

env = Environment()

def config_file_decider(dependency, target, prev_ni):
    import os.path

    # We always have to init the .csig value...
    dep_csig = dependency.get_csig()
    # .csig may not exist, because no target was built yet...
    if 'csig' not in dir(prev_ni):
        return True
    # Target file may not exist yet
    if not os.path.exists(str(target.abspath)):
        return True
    if dep_csig != prev_ni.csig:
        # Some change on source file => update installed one
        return True
    return False

def update_file():
    f = open("test.txt","a")
    f.write("some line\n")
    f.close()

update_file()

# Activate our own decider function
env.Decider(config_file_decider)

env.Install("install","test.txt")
      

SCons still supports two functions that used to be the primary methods for configuring the decision about whether or not an input file has changed. These functions have been officially deprecated as SCons version 2.0, and their use is discouraged, mainly because they rely on a somewhat confusing distinction between how source files and target files are handled. These functions are documented here mainly in case you encounter them in older SConscript files.

The TargetSignatures function specifies how SCons decides when a target file has changed when it is used as a dependency of (input to) another target--that is, the TargetSignatures function configures how the signatures of "intermediate" target files are used when deciding if a "downstream" target file must be rebuilt. [2]

The TargetSignatures function supports the same 'MD5' and 'timestamp' argument values that are supported by the SourceSignatures, with the same meanings, but applied to target files. That is, in the example:

Program('hello.c')
TargetSignatures('MD5')
      

The MD5 checksum of the hello.o target file will be used to decide if it has changed since the last time the "downstream" hello target file was built. And in the example:

Program('hello.c')
TargetSignatures('timestamp')
      

The modification time of the hello.o target file will be used to decide if it has changed since the last time the "downstream" hello target file was built.

The TargetSignatures function supports two additional argument values: 'source' and 'build'. The 'source' argument specifies that decisions involving whether target files have changed since a previous build should use the same behavior for the decisions configured for source files (using the SourceSignatures function). So in the example:

Program('hello.c')
TargetSignatures('source')
SourceSignatures('timestamp')
      

All files, both targets and sources, will use modification times when deciding if an input file has changed since the last time a target was built.

Lastly, the 'build' argument specifies that SCons should examine the build status of a target file and always rebuild a "downstream" target if the target file was itself rebuilt, without re-examining the contents or timestamp of the newly-built target file. If the target file was not rebuilt during this scons invocation, then the target file will be examined the same way as configured by the SourceSignature call to decide if it has changed.

This mimics the behavior of build signatures in earlier versions of SCons. A build signature re-combined signatures of all the input files that went into making the target file, so that the target file itself did not need to have its contents read to compute an MD5 signature. This can improve performance for some configurations, but is generally not as effective as using Decider('MD5-timestamp').

Now suppose that our "Hello, World!" program actually has an #include line to include the hello.h file in the compilation:

#include <hello.h>
int
main()
{
    printf("Hello, %s!\n", string);
}
      

And, for completeness, the hello.h file looks like this:

#define string    "world"
      

In this case, we want SCons to recognize that, if the contents of the hello.h file change, the hello program must be recompiled. To do this, we need to modify the SConstruct file like so:

Program('hello.c', CPPPATH = '.')
      

The $CPPPATH value tells SCons to look in the current directory ('.') for any files included by C source files (.c or .h files). With this assignment in the SConstruct file:

% scons -Q hello
cc -o hello.o -c -I. hello.c
cc -o hello hello.o
% scons -Q hello
scons: `hello' is up to date.
%     [CHANGE THE CONTENTS OF hello.h]
% scons -Q hello
cc -o hello.o -c -I. hello.c
cc -o hello hello.o

First, notice that SCons added the -I. argument from the $CPPPATH variable so that the compilation would find the hello.h file in the local directory.

Second, realize that SCons knows that the hello program must be rebuilt because it scans the contents of the hello.c file for the #include lines that indicate another file is being included in the compilation. SCons records these as implicit dependencies of the target file, Consequently, when the hello.h file changes, SCons realizes that the hello.c file includes it, and rebuilds the resulting hello program that depends on both the hello.c and hello.h files.

Like the $LIBPATH variable, the $CPPPATH variable may be a list of directories, or a string separated by the system-specific path separation character (':' on POSIX/Linux, ';' on Windows). Either way, SCons creates the right command-line options so that the following example:

Program('hello.c', CPPPATH = ['include', '/home/project/inc'])
      

Will look like this on POSIX or Linux:

% scons -Q hello
cc -o hello.o -c -Iinclude -I/home/project/inc hello.c
cc -o hello hello.o

And like this on Windows:

C:\>scons -Q hello.exe
cl /Fohello.obj /c hello.c /nologo /Iinclude /I\home\project\inc
link /nologo /OUT:hello.exe hello.obj
embedManifestExeCheck(target, source, env)

Scanning each file for #include lines does take some extra processing time. When you're doing a full build of a large system, the scanning time is usually a very small percentage of the overall time spent on the build. You're most likely to notice the scanning time, however, when you rebuild all or part of a large system: SCons will likely take some extra time to "think about" what must be built before it issues the first build command (or decides that everything is up to date and nothing must be rebuilt).

In practice, having SCons scan files saves time relative to the amount of potential time lost to tracking down subtle problems introduced by incorrect dependencies. Nevertheless, the "waiting time" while SCons scans files can annoy individual developers waiting for their builds to finish. Consequently, SCons lets you cache the implicit dependencies that its scanners find, for use by later builds. You can do this by specifying the --implicit-cache option on the command line:

% scons -Q --implicit-cache hello
cc -o hello.o -c hello.c
cc -o hello hello.o
% scons -Q hello
scons: `hello' is up to date.

If you don't want to specify --implicit-cache on the command line each time, you can make it the default behavior for your build by setting the implicit_cache option in an SConscript file:

SetOption('implicit_cache', 1)
    

SCons does not cache implicit dependencies like this by default because the --implicit-cache causes SCons to simply use the implicit dependencies stored during the last run, without any checking for whether or not those dependencies are still correct. Specifically, this means --implicit-cache instructs SCons to not rebuild "correctly" in the following cases:

  • When --implicit-cache is used, SCons will ignore any changes that may have been made to search paths (like $CPPPATH or $LIBPATH,). This can lead to SCons not rebuilding a file if a change to $CPPPATH would normally cause a different, same-named file from a different directory to be used.

  • When --implicit-cache is used, SCons will not detect if a same-named file has been added to a directory that is earlier in the search path than the directory in which the file was found last time.

SCons has built-in scanners for a number of languages. Sometimes these scanners fail to extract certain implicit dependencies due to limitations of the scanner implementation.

The following example illustrates a case where the built-in C scanner is unable to extract the implicit dependency on a header file.

#define FOO_HEADER <foo.h>
#include FOO_HEADER

int main() {
    return FOO;
}
      
% scons -Q
cc -o hello.o -c -I. hello.c
cc -o hello hello.o
%    [CHANGE CONTENTS OF foo.h]
% scons -Q
scons: `.' is up to date.

Apparently, the scanner does not know about the header dependency. Being not a full-fledged C preprocessor, the scanner does not expand the macro.

In these cases, you may also use the compiler to extract the implicit dependencies. ParseDepends can parse the contents of the compiler output in the style of Make, and explicitly establish all of the listed dependencies.

The following example uses ParseDepends to process a compiler generated dependency file which is generated as a side effect during compilation of the object file:

obj = Object('hello.c', CCFLAGS='-MD -MF hello.d', CPPPATH='.')
SideEffect('hello.d', obj)
ParseDepends('hello.d')
Program('hello', obj)
      
% scons -Q
cc -o hello.o -c -MD -MF hello.d -I. hello.c
cc -o hello hello.o
%    [CHANGE CONTENTS OF foo.h]
% scons -Q
cc -o hello.o -c -MD -MF hello.d -I. hello.c

Parsing dependencies from a compiler-generated .d file has a chicken-and-egg problem, that causes unnecessary rebuilds:

% scons -Q
cc -o hello.o -c -MD -MF hello.d -I. hello.c
cc -o hello hello.o
% scons -Q --debug=explain
scons: rebuilding `hello.o' because `foo.h' is a new dependency
cc -o hello.o -c -MD -MF hello.d -I. hello.c
% scons -Q
scons: `.' is up to date.
    

In the first pass, the dependency file is generated while the object file is compiled. At that time, SCons does not know about the dependency on foo.h. In the second pass, the object file is regenerated because foo.h is detected as a new dependency.

ParseDepends immediately reads the specified file at invocation time and just returns if the file does not exist. A dependency file generated during the build process is not automatically parsed again. Hence, the compiler-extracted dependencies are not stored in the signature database during the same build pass. This limitation of ParseDepends leads to unnecessary recompilations. Therefore, ParseDepends should only be used if scanners are not available for the employed language or not powerful enough for the specific task.

Sometimes it makes sense to not rebuild a program, even if a dependency file changes. In this case, you would tell SCons specifically to ignore a dependency as follows:

hello_obj=Object('hello.c')
hello = Program(hello_obj)
Ignore(hello_obj, 'hello.h')
      
% scons -Q hello
cc -c -o hello.o hello.c
cc -o hello hello.o
% scons -Q hello
scons: `hello' is up to date.
% edit hello.h
  [CHANGE THE CONTENTS OF hello.h]
% scons -Q hello
scons: `hello' is up to date.
    

Now, the above example is a little contrived, because it's hard to imagine a real-world situation where you wouldn't want to rebuild hello if the hello.h file changed. A more realistic example might be if the hello program is being built in a directory that is shared between multiple systems that have different copies of the stdio.h include file. In that case, SCons would notice the differences between the different systems' copies of stdio.h and would rebuild hello each time you change systems. You could avoid these rebuilds as follows:

hello = Program('hello.c', CPPPATH=['/usr/include'])
Ignore(hello, '/usr/include/stdio.h')
    

Ignore can also be used to prevent a generated file from being built by default. This is due to the fact that directories depend on their contents. So to ignore a generated file from the default build, you specify that the directory should ignore the generated file. Note that the file will still be built if the user specifically requests the target on scons command line, or if the file is a dependency of another file which is requested and/or is built by default.

hello_obj=Object('hello.c')
hello = Program(hello_obj)
Ignore('.',[hello,hello_obj])
      
% scons -Q
scons: `.' is up to date.
% scons -Q hello
cc -o hello.o -c hello.c
cc -o hello hello.o
% scons -Q hello
scons: `hello' is up to date.

Occasionally, it may be useful to specify that a certain file or directory must, if necessary, be built or created before some other target is built, but that changes to that file or directory do not require that the target itself be rebuilt. Such a relationship is called an order-only dependency because it only affects the order in which things must be built--the dependency before the target--but it is not a strict dependency relationship because the target should not change in response to changes in the dependent file.

For example, suppose that you want to create a file every time you run a build that identifies the time the build was performed, the version number, etc., and which is included in every program that you build. The version file's contents will change every build. If you specify a normal dependency relationship, then every program that depends on that file would be rebuilt every time you ran SCons. For example, we could use some Python code in a SConstruct file to create a new version.c file with a string containing the current date every time we run SCons, and then link a program with the resulting object file by listing version.c in the sources:

import time

version_c_text = """
char *date = "%s";
""" % time.ctime(time.time())
open('version.c', 'w').write(version_c_text)

hello = Program(['hello.c', 'version.c'])
      

If we list version.c as an actual source file, though, then the version.o file will get rebuilt every time we run SCons (because the SConstruct file itself changes the contents of version.c) and the hello executable will get re-linked every time (because the version.o file changes):

% scons -Q hello
cc -o hello.o -c hello.c
cc -o version.o -c version.c
cc -o hello hello.o version.o
% sleep 1
% scons -Q hello
cc -o version.o -c version.c
cc -o hello hello.o version.o
% sleep 1
% scons -Q hello
cc -o version.o -c version.c
cc -o hello hello.o version.o

(Note that for the above example to work, we sleep for one second in between each run, so that the SConstruct file will create a version.c file with a time string that's one second later than the previous run.)

One solution is to use the Requires function to specify that the version.o must be rebuilt before it is used by the link step, but that changes to version.o should not actually cause the hello executable to be re-linked:

import time

version_c_text = """
char *date = "%s";
""" % time.ctime(time.time())
open('version.c', 'w').write(version_c_text)

version_obj = Object('version.c')

hello = Program('hello.c',
                LINKFLAGS = str(version_obj[0]))

Requires(hello, version_obj)
      

Notice that because we can no longer list version.c as one of the sources for the hello program, we have to find some other way to get it into the link command line. For this example, we're cheating a bit and stuffing the object file name (extracted from version_obj list returned by the Object call) into the $LINKFLAGS variable, because $LINKFLAGS is already included in the $LINKCOM command line.

With these changes, we get the desired behavior of only re-linking the hello executable when the hello.c has changed, even though the version.o is rebuilt (because the SConstruct file still changes the version.c contents directly each run):

% scons -Q hello
cc -o version.o -c version.c
cc -o hello.o -c hello.c
cc -o hello version.o hello.o
% sleep 1
% scons -Q hello
cc -o version.o -c version.c
scons: `hello' is up to date.
% sleep 1
%     [CHANGE THE CONTENTS OF hello.c]
% scons -Q hello
cc -o version.o -c version.c
cc -o hello.o -c hello.c
cc -o hello version.o hello.o
% sleep 1
% scons -Q hello
cc -o version.o -c version.c
scons: `hello' is up to date.


[2] This easily-overlooked distinction between how SCons decides if the target itself must be rebuilt and how the target is then used to decide if a different target must be rebuilt is one of the confusing things that has led to the TargetSignatures and SourceSignatures functions being replaced by the simpler Decider function.

An environment is a collection of values that can affect how a program executes. SCons distinguishes between three different types of environments that can affect the behavior of SCons itself (subject to the configuration in the SConscript files), as well as the compilers and other tools it executes:

External Environment

The external environment is the set of variables in the user's environment at the time the user runs SCons. These variables are available within the SConscript files through the Python os.environ dictionary. See Section 1, “Using Values From the External Environment”, below.

Construction Environment

A construction environment is a distinct object creating within a SConscript file and and which contains values that affect how SCons decides what action to use to build a target, and even to define which targets should be built from which sources. One of the most powerful features of SCons is the ability to create multiple construction environments, including the ability to clone a new, customized construction environment from an existing construction environment. See Section 2, “Construction Environments”, below.

Execution Environment

An execution environment is the values that SCons sets when executing an external command (such as a compiler or linker) to build one or more targets. Note that this is not the same as the external environment (see above). See Section 3, “Controlling the Execution Environment for Issued Commands”, below.

Unlike Make, SCons does not automatically copy or import values between different environments (with the exception of explicit clones of construction environments, which inherit values from their parent). This is a deliberate design choice to make sure that builds are, by default, repeatable regardless of the values in the user's external environment. This avoids a whole class of problems with builds where a developer's local build works because a custom variable setting causes a different compiler or build option to be used, but the checked-in change breaks the official build because it uses different environment variable settings.

Note that the SConscript writer can easily arrange for variables to be copied or imported between environments, and this is often very useful (or even downright necessary) to make it easy for developers to customize the build in appropriate ways. The point is not that copying variables between different environments is evil and must always be avoided. Instead, it should be up to the implementer of the build system to make conscious choices about how and when to import a variable from one environment to another, making informed decisions about striking the right balance between making the build repeatable on the one hand and convenient to use on the other.

It is rare that all of the software in a large, complicated system needs to be built the same way. For example, different source files may need different options enabled on the command line, or different executable programs need to be linked with different libraries. SCons accommodates these different build requirements by allowing you to create and configure multiple construction environments that control how the software is built. A construction environment is an object that has a number of associated construction variables, each with a name and a value. (A construction environment also has an attached set of Builder methods, about which we'll learn more later.)

You can fetch individual construction variables using the normal syntax for accessing individual named items in a Python dictionary:

env = Environment()
print "CC is:", env['CC']
        

This example SConstruct file doesn't build anything, but because it's actually a Python script, it will print the value of $CC for us:

% scons -Q
CC is: cc
scons: `.' is up to date.

A construction environment, however, is actually an object with associated methods, etc. If you want to have direct access to only the dictionary of construction variables, you can fetch this using the Dictionary method:

env = Environment(FOO = 'foo', BAR = 'bar')
dict = env.Dictionary()
for key in ['OBJSUFFIX', 'LIBSUFFIX', 'PROGSUFFIX']:
    print "key = %s, value = %s" % (key, dict[key])
         

This SConstruct file will print the specified dictionary items for us on POSIX systems as follows:

% scons -Q
key = OBJSUFFIX, value = .o
key = LIBSUFFIX, value = .a
key = PROGSUFFIX, value = 
scons: `.' is up to date.

And on Windows:

C:\>scons -Q
key = OBJSUFFIX, value = .obj
key = LIBSUFFIX, value = .lib
key = PROGSUFFIX, value = .exe
scons: `.' is up to date.

If you want to loop and print the values of all of the construction variables in a construction environment, the Python code to do that in sorted order might look something like:

env = Environment()
for item in sorted(env.Dictionary().items()):
    print "construction variable = '%s', value = '%s'" % item
      

Another way to get information from a construction environment is to use the subst method on a string containing $ expansions of construction variable names. As a simple example, the example from the previous section that used env['CC'] to fetch the value of $CC could also be written as:

env = Environment()
print "CC is:", env.subst('$CC')
      

One advantage of using subst to expand strings is that construction variables in the result get re-expanded until there are no expansions left in the string. So a simple fetch of a value like $CCCOM:

env = Environment(CCFLAGS = '-DFOO')
print "CCCOM is:", env['CCCOM']
      

Will print the unexpanded value of $CCCOM, showing us the construction variables that still need to be expanded:

% scons -Q
CCCOM is: $CC $CCFLAGS $CPPFLAGS $_CPPDEFFLAGS $_CPPINCFLAGS -c -o $TARGET $SOURCES
scons: `.' is up to date.
      

Calling the subst method on $CCOM, however:

env = Environment(CCFLAGS = '-DFOO')
print "CCCOM is:", env.subst('$CCCOM')
      

Will recursively expand all of the construction variables prefixed with $ (dollar signs), showing us the final output:

% scons -Q
CCCOM is: gcc -DFOO -c -o
scons: `.' is up to date.
      

Note that because we're not expanding this in the context of building something there are no target or source files for $TARGET and $SOURCES to expand.

If a problem occurs when expanding a construction variable, by default it is expanded to '' (a null string), and will not cause scons to fail.

env = Environment()
print "value is:", env.subst( '->$MISSING<-' )
        
% scons -Q
value is: -><-
scons: `.' is up to date.

This default behaviour can be changed using the AllowSubstExceptions function. When a problem occurs with a variable expansion it generates an exception, and the AllowSubstExceptions function controls which of these exceptions are actually fatal and which are allowed to occur safely. By default, NameError and IndexError are the two exceptions that are allowed to occur: so instead of causing scons to fail, these are caught, the variable expanded to '' and scons execution continues. To require that all construction variable names exist, and that indexes out of range are not allowed, call AllowSubstExceptions with no extra arguments.

AllowSubstExceptions()
env = Environment()
print "value is:", env.subst( '->$MISSING<-' )
        
% scons -Q
value is:
scons: *** NameError `MISSING' trying to evaluate `$MISSING'
File "/home/my/project/SConstruct", line 3, in <module>

This can also be used to allow other exceptions that might occur, most usefully with the ${...} construction variable syntax. For example, this would allow zero-division to occur in a variable expansion in addition to the default exceptions allowed

AllowSubstExceptions(IndexError, NameError, ZeroDivisionError)
env = Environment()
print "value is:", env.subst( '->${1 / 0}<-' )
        
% scons -Q
value is: -><-
scons: `.' is up to date.

If AllowSubstExceptions is called multiple times, each call completely overwrites the previous list of allowed exceptions.

All of the Builder functions that we've introduced so far, like Program and Library, actually use a default construction environment that contains settings for the various compilers and other tools that SCons configures by default, or otherwise knows about and has discovered on your system. The goal of the default construction environment is to make many configurations to "just work" to build software using readily available tools with a minimum of configuration changes.

You can, however, control the settings in the default construction environment by using the DefaultEnvironment function to initialize various settings:

DefaultEnvironment(CC = '/usr/local/bin/gcc')
      

When configured as above, all calls to the Program or Object Builder will build object files with the /usr/local/bin/gcc compiler.

Note that the DefaultEnvironment function returns the initialized default construction environment object, which can then be manipulated like any other construction environment. So the following would be equivalent to the previous example, setting the $CC variable to /usr/local/bin/gcc but as a separate step after the default construction environment has been initialized:

env = DefaultEnvironment()
env['CC'] = '/usr/local/bin/gcc'
      

One very common use of the DefaultEnvironment function is to speed up SCons initialization. As part of trying to make most default configurations "just work," SCons will actually search the local system for installed compilers and other utilities. This search can take time, especially on systems with slow or networked file systems. If you know which compiler(s) and/or other utilities you want to configure, you can control the search that SCons performs by specifying some specific tool modules with which to initialize the default construction environment:

env = DefaultEnvironment(tools = ['gcc', 'gnulink'],
                         CC = '/usr/local/bin/gcc')
      

So the above example would tell SCons to explicitly configure the default environment to use its normal GNU Compiler and GNU Linker settings (without having to search for them, or any other utilities for that matter), and specifically to use the compiler found at /usr/local/bin/gcc.

The real advantage of construction environments is that you can create as many different construction environments as you need, each tailored to a different way to build some piece of software or other file. If, for example, we need to build one program with the -O2 flag and another with the -g (debug) flag, we would do this like so:

opt = Environment(CCFLAGS = '-O2')
dbg = Environment(CCFLAGS = '-g')

opt.Program('foo', 'foo.c')

dbg.Program('bar', 'bar.c')
        
% scons -Q
cc -o bar.o -c -g bar.c
cc -o bar bar.o
cc -o foo.o -c -O2 foo.c
cc -o foo foo.o

We can even use multiple construction environments to build multiple versions of a single program. If you do this by simply trying to use the Program builder with both environments, though, like this:

opt = Environment(CCFLAGS = '-O2')
dbg = Environment(CCFLAGS = '-g')

opt.Program('foo', 'foo.c')

dbg.Program('foo', 'foo.c')
        

Then SCons generates the following error:

% scons -Q

scons: *** Two environments with different actions were specified for the same target: foo.o
File "/home/my/project/SConstruct", line 6, in <module>

This is because the two Program calls have each implicitly told SCons to generate an object file named foo.o, one with a $CCFLAGS value of -O2 and one with a $CCFLAGS value of -g. SCons can't just decide that one of them should take precedence over the other, so it generates the error. To avoid this problem, we must explicitly specify that each environment compile foo.c to a separately-named object file using the Object builder, like so:

opt = Environment(CCFLAGS = '-O2')
dbg = Environment(CCFLAGS = '-g')

o = opt.Object('foo-opt', 'foo.c')
opt.Program(o)

d = dbg.Object('foo-dbg', 'foo.c')
dbg.Program(d)
        

Notice that each call to the Object builder returns a value, an internal SCons object that represents the object file that will be built. We then use that object as input to the Program builder. This avoids having to specify explicitly the object file name in multiple places, and makes for a compact, readable SConstruct file. Our SCons output then looks like:

% scons -Q
cc -o foo-dbg.o -c -g foo.c
cc -o foo-dbg foo-dbg.o
cc -o foo-opt.o -c -O2 foo.c
cc -o foo-opt foo-opt.o

Sometimes you want more than one construction environment to share the same values for one or more variables. Rather than always having to repeat all of the common variables when you create each construction environment, you can use the Clone method to create a copy of a construction environment.

Like the Environment call that creates a construction environment, the Clone method takes construction variable assignments, which will override the values in the copied construction environment. For example, suppose we want to use gcc to create three versions of a program, one optimized, one debug, and one with neither. We could do this by creating a "base" construction environment that sets $CC to gcc, and then creating two copies, one which sets $CCFLAGS for optimization and the other which sets $CCFLAGS for debugging:

env = Environment(CC = 'gcc')
opt = env.Clone(CCFLAGS = '-O2')
dbg = env.Clone(CCFLAGS = '-g')

env.Program('foo', 'foo.c')

o = opt.Object('foo-opt', 'foo.c')
opt.Program(o)

d = dbg.Object('foo-dbg', 'foo.c')
dbg.Program(d)
        

Then our output would look like:

% scons -Q
gcc -o foo.o -c foo.c
gcc -o foo foo.o
gcc -o foo-dbg.o -c -g foo.c
gcc -o foo-dbg foo-dbg.o
gcc -o foo-opt.o -c -O2 foo.c
gcc -o foo-opt foo-opt.o

You can replace existing construction variable values using the Replace method:

env = Environment(CCFLAGS = '-DDEFINE1')
env.Replace(CCFLAGS = '-DDEFINE2')
env.Program('foo.c')
        

The replacing value (-DDEFINE2 in the above example) completely replaces the value in the construction environment:

% scons -Q
cc -o foo.o -c -DDEFINE2 foo.c
cc -o foo foo.o

You can safely call Replace for construction variables that don't exist in the construction environment:

env = Environment()
env.Replace(NEW_VARIABLE = 'xyzzy')
print "NEW_VARIABLE =", env['NEW_VARIABLE']
        

In this case, the construction variable simply gets added to the construction environment:

% scons -Q
NEW_VARIABLE = xyzzy
scons: `.' is up to date.

Because the variables aren't expanded until the construction environment is actually used to build the targets, and because SCons function and method calls are order-independent, the last replacement "wins" and is used to build all targets, regardless of the order in which the calls to Replace() are interspersed with calls to builder methods:

env = Environment(CCFLAGS = '-DDEFINE1')
print "CCFLAGS =", env['CCFLAGS']
env.Program('foo.c')

env.Replace(CCFLAGS = '-DDEFINE2')
print "CCFLAGS =", env['CCFLAGS']
env.Program('bar.c')
        

The timing of when the replacement actually occurs relative to when the targets get built becomes apparent if we run scons without the -Q option:

% scons
scons: Reading SConscript files ...
CCFLAGS = -DDEFINE1
CCFLAGS = -DDEFINE2
scons: done reading SConscript files.
scons: Building targets ...
cc -o bar.o -c -DDEFINE2 bar.c
cc -o bar bar.o
cc -o foo.o -c -DDEFINE2 foo.c
cc -o foo foo.o
scons: done building targets.

Because the replacement occurs while the SConscript files are being read, the $CCFLAGS variable has already been set to -DDEFINE2 by the time the foo.o target is built, even though the call to the Replace method does not occur until later in the SConscript file.

When SCons builds a target file, it does not execute the commands with the same external environment that you used to execute SCons. Instead, it uses the dictionary stored in the $ENV construction variable as the external environment for executing commands.

The most important ramification of this behavior is that the PATH environment variable, which controls where the operating system will look for commands and utilities, is not the same as in the external environment from which you called SCons. This means that SCons will not, by default, necessarily find all of the tools that you can execute from the command line.

The default value of the PATH environment variable on a POSIX system is /usr/local/bin:/bin:/usr/bin. The default value of the PATH environment variable on a Windows system comes from the Windows registry value for the command interpreter. If you want to execute any commands--compilers, linkers, etc.--that are not in these default locations, you need to set the PATH value in the $ENV dictionary in your construction environment.

The simplest way to do this is to initialize explicitly the value when you create the construction environment; this is one way to do that:

path = ['/usr/local/bin', '/bin', '/usr/bin']
env = Environment(ENV = {'PATH' : path})
    

Assign a dictionary to the $ENV construction variable in this way completely resets the external environment so that the only variable that will be set when external commands are executed will be the PATH value. If you want to use the rest of the values in $ENV and only set the value of PATH, the most straightforward way is probably:

env['ENV']['PATH'] = ['/usr/local/bin', '/bin', '/usr/bin']
    

Note that SCons does allow you to define the directories in the PATH in a string, separated by the pathname-separator character for your system (':' on POSIX systems, ';' on Windows):

env['ENV']['PATH'] = '/usr/local/bin:/bin:/usr/bin'
    

But doing so makes your SConscript file less portable, (although in this case that may not be a huge concern since the directories you list are likley system-specific, anyway).

This chapter describes the MergeFlags, ParseFlags, and ParseConfig methods of a construction environment.

SCons construction environments have a MergeFlags method that merges a dictionary of values into the construction environment. MergeFlags treats each value in the dictionary as a list of options such as one might pass to a command (such as a compiler or linker). MergeFlags will not duplicate an option if it already exists in the construction environment variable.

MergeFlags tries to be intelligent about merging options. When merging options to any variable whose name ends in PATH, MergeFlags keeps the leftmost occurrence of the option, because in typical lists of directory paths, the first occurrence "wins." When merging options to any other variable name, MergeFlags keeps the rightmost occurrence of the option, because in a list of typical command-line options, the last occurrence "wins."

env = Environment()
env.Append(CCFLAGS = '-option -O3 -O1')
flags = { 'CCFLAGS' : '-whatever -O3' }
env.MergeFlags(flags)
print env['CCFLAGS']
   
% scons -Q
['-option', '-O1', '-whatever', '-O3']
scons: `.' is up to date.

Note that the default value for $CCFLAGS is an internal SCons object which automatically converts the options we specified as a string into a list.

env = Environment()
env.Append(CPPPATH = ['/include', '/usr/local/include', '/usr/include'])
flags = { 'CPPPATH' : ['/usr/opt/include', '/usr/local/include'] }
env.MergeFlags(flags)
print env['CPPPATH']
   
% scons -Q
['/include', '/usr/local/include', '/usr/include', '/usr/opt/include']
scons: `.' is up to date.

Note that the default value for $CPPPATH is a normal Python list, so we must specify its values as a list in the dictionary we pass to the MergeFlags function.

If MergeFlags is passed anything other than a dictionary, it calls the ParseFlags method to convert it into a dictionary.

env = Environment()
env.Append(CCFLAGS = '-option -O3 -O1')
env.Append(CPPPATH = ['/include', '/usr/local/include', '/usr/include'])
env.MergeFlags('-whatever -I/usr/opt/include -O3 -I/usr/local/include')
print env['CCFLAGS']
print env['CPPPATH']
   
% scons -Q
['-option', '-O1', '-whatever', '-O3']
['/include', '/usr/local/include', '/usr/include', '/usr/opt/include']
scons: `.' is up to date.

In the combined example above, ParseFlags has sorted the options into their corresponding variables and returned a dictionary for MergeFlags to apply to the construction variables in the specified construction environment.

SCons has a bewildering array of construction variables for different types of options when building programs. Sometimes you may not know exactly which variable should be used for a particular option.

SCons construction environments have a ParseFlags method that takes a set of typical command-line options and distrbutes them into the appropriate construction variables. Historically, it was created to support the ParseConfig method, so it focuses on options used by the GNU Compiler Collection (GCC) for the C and C++ toolchains.

ParseFlags returns a dictionary containing the options distributed into their respective construction variables. Normally, this dictionary would be passed to MergeFlags to merge the options into a construction environment, but the dictionary can be edited if desired to provide additional functionality. (Note that if the flags are not going to be edited, calling MergeFlags with the options directly will avoid an additional step.)

env = Environment()
d = env.ParseFlags("-I/opt/include -L/opt/lib -lfoo")
for k,v in sorted(d.items()):
    if v:
        print k, v
env.MergeFlags(d)
env.Program('f1.c')
   
% scons -Q
CPPPATH ['/opt/include']
LIBPATH ['/opt/lib']
LIBS ['foo']
cc -o f1.o -c -I/opt/include f1.c
cc -o f1 f1.o -L/opt/lib -lfoo

Note that if the options are limited to generic types like those above, they will be correctly translated for other platform types:

C:\>scons -Q
CPPPATH ['/opt/include']
LIBPATH ['/opt/lib']
LIBS ['foo']
cl /Fof1.obj /c f1.c /nologo /I\opt\include
link /nologo /OUT:f1.exe /LIBPATH:\opt\lib foo.lib f1.obj
embedManifestExeCheck(target, source, env)

Since the assumption is that the flags are used for the GCC toolchain, unrecognized flags are placed in $CCFLAGS so they will be used for both C and C++ compiles:

env = Environment()
d = env.ParseFlags("-whatever")
for k,v in sorted(d.items()):
    if v:
        print k, v
env.MergeFlags(d)
env.Program('f1.c')
   
% scons -Q
CCFLAGS -whatever
cc -o f1.o -c -whatever f1.c
cc -o f1 f1.o

ParseFlags will also accept a (recursive) list of strings as input; the list is flattened before the strings are processed:

env = Environment()
d = env.ParseFlags(["-I/opt/include", ["-L/opt/lib", "-lfoo"]])
for k,v in sorted(d.items()):
    if v:
        print k, v
env.MergeFlags(d)
env.Program('f1.c')
   
% scons -Q
CPPPATH ['/opt/include']
LIBPATH ['/opt/lib']
LIBS ['foo']
cc -o f1.o -c -I/opt/include f1.c
cc -o f1 f1.o -L/opt/lib -lfoo

If a string begins with a "!" (an exclamation mark, often called a bang), the string is passed to the shell for execution. The output of the command is then parsed:

env = Environment()
d = env.ParseFlags(["!echo -I/opt/include", "!echo -L/opt/lib", "-lfoo"])
for k,v in sorted(d.items()):
    if v:
        print k, v
env.MergeFlags(d)
env.Program('f1.c')
   
% scons -Q
CPPPATH ['/opt/include']
LIBPATH ['/opt/lib']
LIBS ['foo']
cc -o f1.o -c -I/opt/include f1.c
cc -o f1 f1.o -L/opt/lib -lfoo

ParseFlags is regularly updated for new options; consult the man page for details about those currently recognized.

Configuring the right options to build programs to work with libraries--especially shared libraries--that are available on POSIX systems can be very complicated. To help this situation, various utilies with names that end in config return the command-line options for the GNU Compiler Collection (GCC) that are needed to use these libraries; for example, the command-line options to use a library named lib would be found by calling a utility named lib-config.

A more recent convention is that these options are available from the generic pkg-config program, which has common framework, error handling, and the like, so that all the package creator has to do is provide the set of strings for his particular package.

SCons construction environments have a ParseConfig method that executes a *config utility (either pkg-config or a more specific utility) and configures the appropriate construction variables in the environment based on the command-line options returned by the specified command.

env = Environment()
env['CPPPATH'] = ['/lib/compat']
env.ParseConfig("pkg-config x11 --cflags --libs")
print env['CPPPATH']
   

SCons will execute the specified command string, parse the resultant flags, and add the flags to the appropriate environment variables.

% scons -Q
['/lib/compat', '/usr/X11/include']
scons: `.' is up to date.
 

In the example above, SCons has added the include directory to CPPPATH. (Depending upon what other flags are emitted by the pkg-config command, other variables may have been extended as well.)

Note that the options are merged with existing options using the MergeFlags method, so that each option only occurs once in the construction variable:

env = Environment()
env.ParseConfig("pkg-config x11 --cflags --libs")
env.ParseConfig("pkg-config x11 --cflags --libs")
print env['CPPPATH']
   
% scons -Q
['/usr/X11/include']
scons: `.' is up to date.
 

A key aspect of creating a usable build configuration is providing good output from the build so its users can readily understand what the build is doing and get information about how to control the build. SCons provides several ways of controlling output from the build configuration to help make the build more useful and understandable.

It's often very useful to be able to give users some help that describes the specific targets, build options, etc., that can be used for your build. SCons provides the Help function to allow you to specify this help text:

Help("""
Type: 'scons program' to build the production program,
      'scons debug' to build the debug version.
""")
       

(Note the above use of the Python triple-quote syntax, which comes in very handy for specifying multi-line strings like help text.)

When the SConstruct or SConscript files contain such a call to the Help function, the specified help text will be displayed in response to the SCons -h option:

% scons -h
scons: Reading SConscript files ...
scons: done reading SConscript files.

Type: 'scons program' to build the production program,
      'scons debug' to build the debug version.

Use scons -H for help about command-line options.

The SConscript files may contain multiple calls to the Help function, in which case the specified text(s) will be concatenated when displayed. This allows you to split up the help text across multiple SConscript files. In this situation, the order in which the SConscript files are called will determine the order in which the Help functions are called, which will determine the order in which the various bits of text will get concatenated.

Another use would be to make the help text conditional on some variable. For example, suppose you only want to display a line about building a Windows-only version of a program when actually run on Windows. The following SConstruct file:

env = Environment()

Help("\nType: 'scons program' to build the production program.\n")

if env['PLATFORM'] == 'win32':
    Help("\nType: 'scons windebug' to build the Windows debug version.\n")
       

Will display the complete help text on Windows:

C:\>scons -h
scons: Reading SConscript files ...
scons: done reading SConscript files.

Type: 'scons program' to build the production program.

Type: 'scons windebug' to build the Windows debug version.

Use scons -H for help about command-line options.

But only show the relevant option on a Linux or UNIX system:

% scons -h
scons: Reading SConscript files ...
scons: done reading SConscript files.

Type: 'scons program' to build the production program.

Use scons -H for help about command-line options.

If there is no Help text in the SConstruct or SConscript files, SCons will revert to displaying its standard list that describes the SCons command-line options. This list is also always displayed whenever the -H option is used.

Sometimes the commands executed to compile object files or link programs (or build other targets) can get very long, long enough to make it difficult for users to distinguish error messages or other important build output from the commands themselves. All of the default $*COM variables that specify the command lines used to build various types of target files have a corresponding $*COMSTR variable that can be set to an alternative string that will be displayed when the target is built.

For example, suppose you want to have SCons display a "Compiling" message whenever it's compiling an object file, and a "Linking" when it's linking an executable. You could write a SConstruct file that looks like:

env = Environment(CCCOMSTR = "Compiling $TARGET",
                  LINKCOMSTR = "Linking $TARGET")
env.Program('foo.c')
       

Which would then yield the output:

% scons -Q
Compiling foo.o
Linking foo
    

SCons performs complete variable substitution on $*COMSTR variables, so they have access to all of the standard variables like $TARGET $SOURCES, etc., as well as any construction variables that happen to be configured in the construction environment used to build a specific target.

Of course, sometimes it's still important to be able to see the exact command that SCons will execute to build a target. For example, you may simply need to verify that SCons is configured to supply the right options to the compiler, or a developer may want to cut-and-paste a compile command to add a few options for a custom test.

One common way to give users control over whether or not SCons should print the actual command line or a short, configured summary is to add support for a VERBOSE command-line variable to your SConstruct file. A simple configuration for this might look like:

env = Environment()
if ARGUMENTS.get('VERBOSE') != "1':
    env['CCCOMSTR'] = "Compiling $TARGET"
    env['LINKCOMSTR'] = "Linking $TARGET"
env.Program('foo.c')
       

By only setting the appropriate $*COMSTR variables if the user specifies VERBOSE=1 on the command line, the user has control over how SCons displays these particular command lines:

% scons -Q
Compiling foo.o
Linking foo
% scons -Q -c
Removed foo.o
Removed foo
% scons -Q VERBOSE=1
cc -o foo.o -c foo.c
cc -o foo foo.o
    

Another aspect of providing good build output is to give the user feedback about what SCons is doing even when nothing is being built at the moment. This can be especially true for large builds when most of the targets are already up-to-date. Because SCons can take a long time making absolutely sure that every target is, in fact, up-to-date with respect to a lot of dependency files, it can be easy for users to mistakenly conclude that SCons is hung or that there is some other problem with the build.

One way to deal with this perception is to configure SCons to print something to let the user know what it's "thinking about." The Progress function allows you to specify a string that will be printed for every file that SCons is "considering" while it is traversing the dependency graph to decide what targets are or are not up-to-date.

Progress('Evaluating $TARGET\n')
Program('f1.c')
Program('f2.c')
      

Note that the Progress function does not arrange for a newline to be printed automatically at the end of the string (as does the Python print statement), and we must specify the \n that we want printed at the end of the configured string. This configuration, then, will have SCons print that it is Evaluating each file that it encounters in turn as it traverses the dependency graph:

% scons -Q
Evaluating SConstruct
Evaluating f1.c
Evaluating f1.o
cc -o f1.o -c f1.c
Evaluating f1
cc -o f1 f1.o
Evaluating f2.c
Evaluating f2.o
cc -o f2.o -c f2.c
Evaluating f2
cc -o f2 f2.o
Evaluating .

Of course, normally you don't want to add all of these additional lines to your build output, as that can make it difficult for the user to find errors or other important messages. A more useful way to display this progress might be to have the file names printed directly to the user's screen, not to the same standard output stream where build output is printed, and to use a carriage return character (\r) so that each file name gets re-printed on the same line. Such a configuration would look like:

Progress('$TARGET\r',
         file=open('/dev/tty', 'w'),
         overwrite=True)
Program('f1.c')
Program('f2.c')
    

Note that we also specified the overwrite=True argument to the Progress function, which causes SCons to "wipe out" the previous string with space characters before printing the next Progress string. Without the overwrite=True argument, a shorter file name would not overwrite all of the charactes in a longer file name that precedes it, making it difficult to tell what the actual file name is on the output. Also note that we opened up the /dev/tty file for direct access (on POSIX) to the user's screen. On Windows, the equivalent would be to open the con: file name.

Also, it's important to know that although you can use $TARGET to substitute the name of the node in the string, the Progress function does not perform general variable substitution (because there's not necessarily a construction environment involved in evaluating a node like a source file, for example).

You can also specify a list of strings to the Progress function, in which case SCons will display each string in turn. This can be used to implement a "spinner" by having SCons cycle through a sequence of strings:

Progress(['-\r', '\\\r', '|\r', '/\r'], interval=5)
Program('f1.c')
Program('f2.c')
    

Note that here we have also used the interval= keyword argument to have SCons only print a new "spinner" string once every five evaluated nodes. Using an interval= count, even with strings that use $TARGET like our examples above, can be a good way to lessen the work that SCons expends printing Progress strings, while still giving the user feedback that indicates SCons is still working on evaluating the build.

Lastly, you can have direct control over how to print each evaluated node by passing a Python function (or other Python callable) to the Progress function. Your function will be called for each evaluated node, allowing you to implement more sophisticated logic like adding a counter:

screen = open('/dev/tty', 'w')
count = 0
def progress_function(node)
    count += 1
    screen.write('Node %4d: %s\r' % (count, node))

Progress(progress_function)
      

Of course, if you choose, you could completely ignore the node argument to the function, and just print a count, or anything else you wish.

(Note that there's an obvious follow-on question here: how would you find the total number of nodes that will be evaluated so you can tell the user how close the build is to finishing? Unfortunately, in the general case, there isn't a good way to do that, short of having SCons evaluate its dependency graph twice, first to count the total and the second time to actually build the targets. This would be necessary because you can't know in advance which target(s) the user actually requested to be built. The entire build may consist of thousands of Nodes, for example, but maybe the user specifically requested that only a single object file be built.)

SCons, like most build tools, returns zero status to the shell on success and nonzero status on failure. Sometimes it's useful to give more information about the build status at the end of the run, for instance to print an informative message, send an email, or page the poor slob who broke the build.

SCons provides a GetBuildFailures method that you can use in a python atexit function to get a list of objects describing the actions that failed while attempting to build targets. There can be more than one if you're using -j. Here's a simple example:

import atexit

def print_build_failures():
    from SCons.Script import GetBuildFailures
    for bf in GetBuildFailures():
        print "%s failed: %s" % (bf.node, bf.errstr)
atexit.register(print_build_failures)
      

The atexit.register call registers print_build_failures as an atexit callback, to be called before SCons exits. When that function is called, it calls GetBuildFailures to fetch the list of failed objects. See the man page for the detailed contents of the returned objects; some of the more useful attributes are .node, .errstr, .filename, and .command. The filename is not necessarily the same file as the node; the node is the target that was being built when the error occurred, while the filenameis the file or dir that actually caused the error. Note: only call GetBuildFailures at the end of the build; calling it at any other time is undefined.

Here is a more complete example showing how to turn each element of GetBuildFailures into a string:

# Make the build fail if we pass fail=1 on the command line
if ARGUMENTS.get('fail', 0):
    Command('target', 'source', ['/bin/false'])

def bf_to_str(bf):
    """Convert an element of GetBuildFailures() to a string
    in a useful way."""
    import SCons.Errors
    if bf is None: # unknown targets product None in list
        return '(unknown tgt)'
    elif isinstance(bf, SCons.Errors.StopError):
        return str(bf)
    elif bf.node:
        return str(bf.node) + ': ' + bf.errstr
    elif bf.filename:
        return bf.filename + ': ' + bf.errstr
    return 'unknown failure: ' + bf.errstr
import atexit

def build_status():
    """Convert the build status to a 2-tuple, (status, msg)."""
    from SCons.Script import GetBuildFailures
    bf = GetBuildFailures()
    if bf:
        # bf is normally a list of build failures; if an element is None,
        # it's because of a target that scons doesn't know anything about.
        status = 'failed'
        failures_message = "\n".join(["Failed building %s" % bf_to_str(x)
                           for x in bf if x is not None])
    else:
        # if bf is None, the build completed successfully.
        status = 'ok'
        failures_message = ''
    return (status, failures_message)

def display_build_status():
    """Display the build status.  Called by atexit.
    Here you could do all kinds of complicated things."""
    status, failures_message = build_status()
    if status == 'failed':
       print "FAILED!!!!"  # could display alert, ring bell, etc.
    elif status == 'ok':
       print "Build succeeded."
    print failures_message

atexit.register(display_build_status)
      

When this runs, you'll see the appropriate output:

% scons -Q
scons: `.' is up to date.
Build succeeded.
% scons -Q fail=1
scons: *** [target] Source `source' not found, needed by target `target'.
FAILED!!!!
Failed building target: Source `source' not found, needed by target `target'.

SCons provides a number of ways for the writer of the SConscript files to give the users who will run SCons a great deal of control over the build execution. The arguments that the user can specify on the command line are broken down into three types:

Options

Command-line options always begin with one or two - (hyphen) characters. SCons provides ways for you to examine and set options values from within your SConscript files, as well as the ability to define your own custom options. See Section 1, “Command-Line Options”, below.

Variables

Any command-line argument containing an = (equal sign) is considered a variable setting with the form variable=value. SCons provides direct access to all of the command-line variable settings, the ability to apply command-line variable settings to construction environments, and functions for configuring specific types of variables (Boolean values, path names, etc.) with automatic validation of the user's specified values. See Section 2, “Command-Line variable=value Build Variables”, below.

Targets

Any command-line argument that is not an option or a variable setting (does not begin with a hyphen and does not contain an equal sign) is considered a target that the user (presumably) wants SCons to build. A list of Node objects representing the target or targets to build. SCons provides access to the list of specified targets, as well as ways to set the default list of targets from within the SConscript files. See Section 3, “Command-Line Targets”, below.

SCons has many command-line options that control its behavior. A SCons command-line option always begins with one or two - (hyphen) characters.

You can also set the values of SCons command-line options from within the SConscript files by using the SetOption function. The strings that you use to set the values of SCons command-line options are available in the Section 1.4, “Strings for Getting or Setting Values of SCons Command-Line Options” section, below.

One use of the SetOption function is to specify a value for the -j or --jobs option, so that users get the improved performance of a parallel build without having to specify the option by hand. A complicating factor is that a good value for the -j option is somewhat system-dependent. One rough guideline is that the more processors your system has, the higher you want to set the -j value, in order to take advantage of the number of CPUs.

For example, suppose the administrators of your development systems have standardized on setting a NUM_CPU environment variable to the number of processors on each system. A little bit of Python code to access the environment variable and the SetOption function provide the right level of flexibility:

import os
num_cpu = int(os.environ.get('NUM_CPU', 2))
SetOption('num_jobs', num_cpu)
print "running with -j", GetOption('num_jobs')
        

The above snippet of code sets the value of the --jobs option to the value specified in the $NUM_CPU environment variable. (This is one of the exception cases where the string is spelled differently from the from command-line option. The string for fetching or setting the --jobs value is num_jobs for historical reasons.) The code in this example prints the num_jobs value for illustrative purposes. It uses a default value of 2 to provide some minimal parallelism even on single-processor systems:

% scons -Q
running with -j 2
scons: `.' is up to date.

But if the $NUM_CPU environment variable is set, then we use that for the default number of jobs:

% export NUM_CPU="4"
% scons -Q
running with -j 4
scons: `.' is up to date.

But any explicit -j or --jobs value the user specifies an the command line is used first, regardless of whether or not the $NUM_CPU environment variable is set:

% scons -Q -j 7
running with -j 7
scons: `.' is up to date.
% export NUM_CPU="4"
% scons -Q -j 3
running with -j 3
scons: `.' is up to date.

SCons also allows you to define your own command-line options with the AddOption function. The AddOption function takes the same arguments as the optparse.add_option function from the standard Python library. [3] Once you have added a custom command-line option with the AddOption function, the value of the option (if any) is immediately available using the standard GetOption function. (The value can also be set using SetOption, although that's not very useful in