Cmake language reference
CMake input files are written in the "CMake Language" in source files named CMakeLists.txt or ending in a .cmake file name extension.
CMake Language source files in a project are organized into:
Directories (CMakeLists.txt),
Scripts (<script>.cmake), and
Modules (<module>.cmake).
When CMake processes a project source tree, the entry point is a source file called CMakeLists.txt in the top-level source directory. This file may contain the entire build specification or use the add_subdirectory() command to add subdirectories to the build. Each subdirectory added by the command must also contain a CMakeLists.txt file as the entry point to that directory. For each source directory whose CMakeLists.txt file is processed CMake generates a corresponding directory in the build tree to act as the default working and output directory.
An individual <script>.cmake source file may be processed in script mode by using the cmake(1) command-line tool with the -P option. Script mode simply runs the commands in the given CMake Language source file and does not generate a build system. It does not allow CMake commands that define build targets or actions.
CMake Language code in either Directories or Scripts may use the include() command to load a <module>.cmake source file in the scope of the including context. See the cmake-modules(7) manual page for documentation of modules included with the CMake distribution. Project source trees may also provide their own modules and specify their location(s) in the CMAKE_MODULE_PATH variable.
A CMake Language source file must be written in 7-bit ASCII text to be portable across all supported platforms. Newlines may be encoded as either \n or \r\n but will be converted to \n as input files are read.
Note that the implementation is 8-bit clean so source files may be encoded as UTF-8 on platforms with system APIs supporting this encoding. Furthermore, CMake 3.0 and above allow a leading UTF-8 Byte-Order Mark in source files.
A CMake Language source file consists of zero or more Command Invocations separated by newlines and optionally spaces and Comments:
file ::= file_element*
file_element ::= command_invocation line_ending |
(bracket_comment|space)* line_ending
line_ending ::= line_comment? newline
space ::= <match '[ \t]+'>
newline ::= <match '\n'>
Note that any source file line not inside Command Arguments or a Bracket Comment can end in a Line Comment.
A command invocation is a name followed by paren-enclosed arguments separated by whitespace:
command_invocation ::= space* identifier space* '(' arguments ')'
identifier ::= <match '[A-Za-z_][A-Za-z0-9_]*'>
arguments ::= argument? separated_arguments*
separated_arguments ::= separation+ argument? |
separation* '(' arguments ')'
separation ::= space | line_ending
For example:
add_executable(hello world.c)
Command names are case-insensitive. Nested unquoted parentheses in the arguments must balance. Each ( or ) is given to the command invocation as a literal Unquoted Argument. This may be used in calls to the if() command to enclose conditions. For example:
if(FALSE AND (FALSE OR TRUE)) # evaluates to FALSE
NOTE: CMake versions prior to 3.0 require command name identifiers to be at least 2 characters.
CMake versions prior to 2.8.12 silently accept an Unquoted Argument or a Quoted Argument immediately following a Quoted Argument and not separated by any whitespace. For compatibility, CMake 2.8.12 and higher accept such code but produce a warning.
There are three types of arguments within Command Invocations:
argument ::= bracket_argument | quoted_argument | unquoted_argument
A bracket argument, inspired by Lua long bracket syntax, encloses content between opening and closing "brackets" of the same length:
bracket_argument ::= bracket_open bracket_content bracket_close
bracket_open ::= '[' '='{len} '['
bracket_content ::= <any text not containing a bracket_close
of the same {len} as the bracket_open>
bracket_close ::= ']' '='{len} ']'
An opening bracket of length len >= 0 is written [ followed by len = followed by [ and the corresponding closing bracket is written ] followed by len = followed by ]. Brackets do not nest. A unique length may always be chosen for the opening and closing brackets to contain closing brackets of other lengths.
Bracket argument content consists of all text between the opening and closing brackets, except that one newline immediately following the opening bracket, if any, is ignored. No evaluation of the enclosed content, such as Escape Sequences or Variable References, is performed. A bracket argument is always given to the command invocation as exactly one argument.
For example:
message([=[
This is the first line in a bracket argument with bracket length 1.
No \-escape sequences or ${variable} references are evaluated.
This is always one argument even though it contains a ; character.
The text does not end on a closing bracket of length 0 like ]].
It does end in a closing bracket of length 1.
]=])
NOTE: CMake versions prior to 3.0 do not support bracket arguments. They interpret the opening bracket as the start of an Unquoted Argument.
A quoted argument encloses content between opening and closing double-quote characters:
quoted_argument ::= '"' quoted_element* '"'
quoted_element ::= <any character except '\' or '"'> |
escape_sequence |
quoted_continuation
quoted_continuation ::= '\' newline
Quoted argument content consists of all text between opening and closing quotes. Both Escape Sequences and Variable References are evaluated. A quoted argument is always given to the command invocation as exactly one argument.
For example:
message("This is a quoted argument containing multiple lines.
This is always one argument even though it contains a ; character.
Both \\-escape sequences and ${variable} references are evaluated.
The text does not end on an escaped double-quote like \".
It does end in an unescaped double quote.
")
The final \ on any line ending in an odd number of backslashes is treated as a line continuation and ignored along with the immediately following newline character. For example:
message("\
This is the first line of a quoted argument. \
In fact it is the only line but since it is long \
the source code uses line continuation.\
")
NOTE: CMake versions prior to 3.0 do not support continuation with \. They report errors in quoted arguments containing lines ending in an odd number of \ characters.
An unquoted argument is not enclosed by any quoting syntax. It may not contain any whitespace, (, ), #, ", or \ except when escaped by a backslash:
unquoted_argument ::= unquoted_element+ | unquoted_legacy
unquoted_element ::= <any character except whitespace or one of '()#"\'> |
escape_sequence
unquoted_legacy ::= <see note in text>
Unquoted argument content consists of all text in a contiguous block of allowed or escaped characters. Both Escape Sequences and Variable References are evaluated. The resulting value is divided in the same way Lists divide into elements. Each non-empty element is given to the command invocation as an argument. Therefore an unquoted argument may be given to a command invocation as zero or more arguments.
For example:
foreach(arg
NoSpace
Escaped\ Space
This;Divides;Into;Five;Arguments
Escaped\;Semicolon
)
message("${arg}")
endforeach()
NOTE: To support legacy CMake code, unquoted arguments may also contain double-quoted strings ("...", possibly enclosing horizontal whitespace), and make-style variable references ($(MAKEVAR)). Unescaped double-quotes must balance, may not appear at the beginning of an unquoted argument, and are treated as part of the content. For example, the unquoted arguments -Da="b c", -Da=$(v), and a" "b"c"d are each interpreted literally.
The above "unquoted_legacy" production represents such arguments. We do not recommend using legacy unquoted arguments in new code. Instead use a Quoted Argument or a Bracket Argument to represent the content.
An escape sequence is a \ followed by one character:
escape_sequence ::= escape_identity | escape_encoded | escape_semicolon
escape_identity ::= '\(' | '\)' | '\#' | '\"' | '\ ' |
'\\' | '\$' | '\@' | '\^'
escape_encoded ::= '\t' | '\r' | '\n'
escape_semicolon ::= '\;'
A \ followed by one of ()#" \#@^ simply encodes the literal character without interpreting it as syntax. A \t, \r, or \n encodes a tab, carriage return, or newline character, respectively. A \; encodes itself but may be used in an Unquoted Argument to encode the ; without dividing the argument value on it.
A variable reference has the form ${variable_name} and is evaluated inside a Quoted Argument or an Unquoted Argument. A variable reference is replaced by the value of the variable, or by the empty string if the variable is not set. Variable references can nest and are evaluated from the inside out, e.g. ${outer_${inner_variable}_variable}.
The Variables section documents the scope of variable names and how their values are set.
An environment variable reference has the form $ENV{VAR} and is evaluated in the same contexts as a normal variable reference.
A comment starts with a # character that is not inside a Bracket Argument, Quoted Argument, or escaped with \ as part of an Unquoted Argument. There are two types of comments: a Bracket Comment and a Line Comment.
A # immediately followed by a Bracket Argument forms a bracket comment consisting of the entire bracket enclosure:
bracket_comment ::= '#' bracket_argument
For example:
#[[This is a bracket comment.
It runs until the close bracket.]]
message("First Argument\n" #[[Bracket Comment]] "Second Argument")
NOTE: CMake versions prior to 3.0 do not support bracket comments. They interpret the opening # as the start of a Line Comment.
A # not immediately followed by a Bracket Argument forms a line comment that runs until the end of the line:
line_comment ::= '#' <any text not starting in a bracket_argument
and not containing a newline>
For example:
# This is a line comment.
message("First Argument\n" # This is a line comment :)
"Second Argument") # This is a line comment.
The if()/elseif()/else()/endif() commands delimit code blocks to be executed conditionally.
The foreach()/endforeach() and while()/endwhile() commands delimit code blocks to be executed in a loop. The break() command may be used inside such blocks to terminate the loop early.
The macro()/endmacro(), and function()/endfunction() commands delimit code blocks to be recorded for later invocation as commands.
Variables are the basic unit of storage in the CMake Language. Their values are always of string type, though some commands may interpret the strings as values of other types. The set() and unset() commands explicitly set or unset a variable, but other commands have semantics that modify variables as well. Variable names are case-sensitive and may consist of almost any text, but we recommend sticking to names consisting only of alphanumeric characters plus _ and -.
Variables have dynamic scope. Each variable "set" or "unset" creates a binding in the current scope:
Function Scope
Command Definitions created by the function() command create commands that, when invoked, process the recorded commands in a new variable binding scope. A variable "set" or "unset" binds in this scope and is visible for the current function and any nested calls, but not after the function returns.
Directory Scope
Each of the Directories in a source tree has its own variable bindings. Before processing the CMakeLists.txt file for a directory, CMake copies all variable bindings currently defined in the parent directory, if any, to initialize the new directory scope. CMake Scripts, when processed with cmake -P, bind variables in one "directory" scope.
A variable "set" or "unset" not inside a function call binds to the current directory scope.
Persistent Cache
CMake stores a separate set of "cache" variables, or "cache entries", whose values persist across multiple runs within a project build tree. Cache entries have an isolated binding scope modified only by explicit request, such as by the CACHE option of the set() and unset() commands.
When evaluating Variable References, CMake first searches the function call stack, if any, for a binding and then falls back to the binding in the current directory scope, if any. If a "set" binding is found, its value is used. If an "unset" binding is found, or no binding is found, CMake then searches for a cache entry. If a cache entry is found, its value is used. Otherwise, the variable reference evaluates to an empty string.
The cmake-variables(7) manual documents many variables that are provided by CMake or have meaning to CMake when set by project code.
Although all values in CMake are stored as strings, a string may be treated as a list in certain contexts, such as during evaluation of an Unquoted Argument. In such contexts, a string is divided into list elements by splitting on ; characters not following an unequal number of [ and ] characters and not immediately preceded by a \. The sequence \; does not divide a value but is replaced by ; in the resulting element.
A list of elements is represented as a string by concatenating the elements separated by ;. For example, the set() command stores multiple values into the destination variable as a list:
set(srcs a.c b.c c.c) # sets "srcs" to "a.c;b.c;c.c"
Lists are meant for simple use cases such as a list of source files and should not be used for complex data processing tasks. Most commands that construct lists do not escape ; characters in list elements, thus flattening nested lists:
set(x a "b;c") # sets "x" to "a;b;c", not "a;b\;c"
2000-2014 Kitware, Inc.