Manual Pages  — MAKE

All dependency lines mentioning a particular target must use the same operator.

Targets and sources may contain the shell wildcard values ‘amp;?’, ‘*’, ‘[]’, and ‘{}’. The values ‘amp;?’, ‘*’, and ‘[]’ may only be used as part of the final component of the target or source, and only match existing files. The value ‘{}’ need not necessarily be used to describe existing files. Expansion is in directory order, not alphabetically as done in the shell.

Each line is treated as a separate shell command, unless the end of line is escaped with a backslash ‘\’, in which case that line and the next are combined. If the first characters of the command are any combination of ' @', ' +', or ' -', the command is treated specially.

@	causes the command not to be echoed before it is executed.
+	causes the command to be executed even when -n is given. This is similar to the effect of the .MAKE special source, except that the effect can be limited to a single line of a script.
-	in compatibility mode causes any non-zero exit status of the command line to be ignored.

When make is run in jobs mode with -j max_jobs, the entire script for the target is fed to a single instance of the shell. In compatibility (non-jobs) mode, each command is run in a separate process. If the command contains any shell meta characters (‘#=|^(){};&<>*?[]:$`\\n’), it is passed to the shell; otherwise make attempts direct execution. If a line starts with ' -' and the shell has ErrCtl enabled, failure of the command line is ignored as in compatibility mode. Otherwise ' -' affects the entire job; the script stops at the first command line that fails, but the target is not deemed to have failed.

Makefiles should be written so that the mode of make operation does not change their behavior. For example, any command which uses "cd" or "chdir" without the intention of changing the directory for subsequent commands should be put in parentheses so it executes in a subshell. To force the use of a single shell, escape the line breaks so as to make the whole script one command. For example:

avoid-chdir-side-effects:
        @echo "Building $@ in $$(pwd)"
        @(cd ${.CURDIR} && ${MAKE} $@)
        @echo "Back in $$(pwd)"

ensure-one-shell-regardless-of-mode:
        @echo "Building $@ in $$(pwd)"; \
        (cd ${.CURDIR} && ${MAKE} $@); \
        echo "Back in $$(pwd)"

Since make changes the current working directory to ' .OBJDIR' before executing any targets, each child process starts with that as its current working directory.

Variable assignments have the form 'NAME op value', where:

NAME	is a single-word variable name, consisting, by tradition, of all upper-case letters,
op	is one of the variable assignment operators described below, and
value	is interpreted according to the variable assignment operator.

Whitespace around NAME, op and value is discarded.

References to variables have the form amp;${name[ amp;:modifiers ] amp;} or amp;$(name[ amp;:modifiers ] amp;). If the variable name consists of only a single character and the expression contains no modifiers, the surrounding curly braces or parentheses are not required. This shorter form is not recommended.

If the variable name contains a dollar, the name itself is expanded first. This allows almost arbitrary variable names, however names containing dollar, braces, parentheses or whitespace are really best avoided.

If the result of expanding a nested variable expression contains a dollar sign (‘amp;$’), the result is subject to further expansion.

Variable substitution occurs at four distinct times, depending on where the variable is being used.

1. Variables in dependency lines are expanded as the line is read.
2. Variables in conditionals are expanded individually, but only as far as necessary to determine the result of the conditional.
3. Variables in shell commands are expanded when the shell command is executed.
4. .for loop index variables are expanded on each loop iteration. Note that other variables are not expanded when composing the body of a loop, so the following example code:

   amp;.for i in 1 2 3
   a+=     ${i}
   j=      ${i}
   b+=     ${j}
   amp;.endfor
   
   all:
           @echo ${a}
           @echo ${b}
   

   prints:

   1 2 3
   3 3 3
   

   After the loop is executed:

   a	contains ‘${:U1} ${:U2} ${:U3}’, which expands to ‘1 2 3’.
   j	contains ‘${:U3}’, which expands to ‘3’.
   b	contains ‘${j} ${j} ${j}’, which expands to ‘${:U3} ${:U3} ${:U3}’ and further to ‘3 3 3’.

Local variables can be set on a dependency line, unless .MAKE.TARGET_LOCAL_VARIABLES is set to ‘false’. The rest of the line (which already has had global variables expanded) is the variable value. For example:

COMPILER_WRAPPERS= ccache distcc icecc

${OBJS}: .MAKE.META.CMP_FILTER=${COMPILER_WRAPPERS:S,^,N,}

Only the targets ‘${OBJS}’ are impacted by that filter (in "meta" mode) and simply enabling/disabling any of the compiler wrappers does not render all of those targets out-of-date.

NOTE: target-local variable assignments behave differently in that;

amp;+=	Only appends to a previous local assignment for the same target and variable.
amp;:=	Is redundant with respect to global variables, which have already been expanded.

The seven built-in local variables are:

.ALLSRC
	The list of all sources for this target; also known as ' amp;>'.
.ARCHIVE
	The name of the archive file; also known as ' amp;!'.
.IMPSRC
	In suffix-transformation rules, the name/path of the source from which the target is to be transformed (the "implied" source); also known as ' amp;<'. It is not defined in explicit rules.
.MEMBER
	The name of the archive member; also known as ' %'.
.OODATE
	The list of sources for this target that were deemed out-of-date; also known as ' amp;?'.
.PREFIX
	The name of the target with suffix (if declared in .SUFFIXES) removed; also known as ' *'.
.TARGET
	The name of the target; also known as ' @'. For compatibility with other makes this is an alias for .ARCHIVE in archive member rules.

The shorter forms ( ' amp;>', ' amp;!', ' amp;<', ' amp;%', ' amp;?', ' amp;*', and ' amp;@' ) are permitted for backward compatibility with historical makefiles and legacy POSIX make and are not recommended.

Variants of these variables with the punctuation followed immediately by ‘D’ or ‘F’, e.g.amp; ‘$(@D)’, are legacy forms equivalent to using the ‘:H’ and ‘:T’ modifiers. These forms are accepted for compatibility with AT&T V makefiles and POSIX but are not recommended.

Four of the local variables may be used in sources on dependency lines because they expand to the proper value for each target on the line. These variables are ' .TARGET', ' .PREFIX', ' .ARCHIVE', and ' .MEMBER'.

Each modifier begins with a colon. To escape a colon, precede it with a backslash ‘\’.

A list of indirect modifiers can be specified via a variable, as follows:

modifier_variableamp;= modifier[ amp;:... ]

amp;${ variableamp;:${ modifier_variableamp;} [amp;:... ]amp;}

In this case, the first modifier in the modifier_variable does not start with a colon, since that colon already occurs in the referencing variable. If any of the modifiers in the modifier_variable contains a dollar sign (‘$’), these must be doubled to avoid early expansion.

Some modifiers interpret the expression value as a single string, others treat the expression value as a whitespace-separated list of words. When splitting a string into words, whitespace can be escaped using double quotes, single quotes and backslashes, like in the shell. The quotes and backslashes are retained in the words.

The supported modifiers are:

amp;:E	Replaces each word with its suffix.
amp;:H	Replaces each word with its dirname.
amp;:Mpattern
	Selects only those words that match pattern. The standard shell wildcard characters (‘*’, ‘amp;?’, and ‘amp;[]’) may be used. The wildcard characters may be escaped with a backslash (‘\’). As a consequence of the way values are split into words, matched, and then joined, the construct ‘${VAR:M*}’ removes all leading and trailing whitespace and normalizes the inter-word spacing to a single space.
amp;:Npattern
	This is the opposite of ' amp;:M', selecting all words which do not match pattern.
amp;:O	Orders the words lexicographically.
amp;:On	Orders the words numerically. A number followed by one of ‘k’, ‘M’ or ‘G’ is multiplied by the appropriate factor, which is 1024 for ‘k’, 1048576 for ‘M’, or 1073741824 for ‘G’. Both upper- and lower-case letters are accepted.
amp;:Or	Orders the words in reverse lexicographical order.
amp;:Orn
	Orders the words in reverse numerical order.
amp;:Ox	Shuffles the words. The results are different each time you are referring to the modified variable; use the assignment with expansion ' amp;:=' to prevent such behavior. For example, LIST=                   uno due tre quattro RANDOM_LIST=            ${LIST:Ox} STATIC_RANDOM_LIST:=    ${LIST:Ox} all:         @echo "${RANDOM_LIST}"         @echo "${RANDOM_LIST}"         @echo "${STATIC_RANDOM_LIST}"         @echo "${STATIC_RANDOM_LIST}" may produce output similar to: quattro due tre uno tre due quattro uno due uno quattro tre due uno quattro tre
amp;:Q	Quotes every shell meta-character in the value, so that it can be passed safely to the shell.
amp;:q	Quotes every shell meta-character in the value, and also doubles '$' characters so that it can be passed safely through recursive invocations of make. This is equivalent to ' amp;:S/\amp;$/&&/g:Q'.
amp;:R	Replaces each word with everything but its suffix.
amp;:range[ =count ]
	The value is an integer sequence representing the words of the original value, or the supplied count.
amp;:gmtime[ =timestamp ]
	The value is interpreted as a format string for strftime(3), using gmtime(3), producing the formatted timestamp. If a timestamp value is not provided or is 0, the current time is used.
amp;:hash
	Computes a 32-bit hash of the value and encodes it as 8 hex digits.
amp;:localtime[ =timestamp ]
	The value is interpreted as a format string for strftime(3), using localtime(3), producing the formatted timestamp. If a timestamp value is not provided or is 0, the current time is used.
amp;:mtime[ =timestamp ]
	Call stat(2) with each word as pathname; use ‘st_mtime’ as the new value. If stat(2) fails; use timestamp or current time. If timestamp is set to ‘error’, then stat(2) failure will cause an error.
amp;:tA	Attempts to convert the value to an absolute path using realpath(3). If that fails, the value is unchanged.
amp;:tl	Converts the value to lower-case letters.
amp;:tsc
	When joining the words after a modifier that treats the value as words, the words are normally separated by a space. This modifier changes the separator to the character c. If c is omitted, no separator is used. The common escapes (including octal numeric codes) work as expected.
amp;:tu	Converts the value to upper-case letters.
amp;:tW	Causes subsequent modifiers to treat the value as a single word (possibly containing embedded whitespace). See also ' amp;:[*]'.
amp;:tw	Causes the value to be treated as a list of words. See also ' amp;:[@]'.
amp;:Samp;/ old_stringamp;/ new_stringamp;/ [1gW]
	Modifies the first occurrence of old_string in each word of the value, replacing it with new_string. If a ‘g’ is appended to the last delimiter of the pattern, all occurrences in each word are replaced. If a ‘1’ is appended to the last delimiter of the pattern, only the first occurrence is affected. If a ‘W’ is appended to the last delimiter of the pattern, the value is treated as a single word. If old_string begins with a caret (‘^’), old_string is anchored at the beginning of each word. If old_string ends with a dollar sign (‘amp;$’), it is anchored at the end of each word. Inside new_string, an ampersand (‘&’) is replaced by old_string (without the anchoring ‘^’ or ‘amp;$’). Any character may be used as the delimiter for the parts of the modifier string. The anchoring, ampersand and delimiter characters can be escaped with a backslash (‘\’). Both old_string and new_string may contain nested expressions. To prevent a dollar sign from starting a nested expression, escape it with a backslash.
amp;:Camp;/ patternamp;/ replacementamp;/ [1gW]
	The amp;:C modifier works like the amp;:S modifier except that the old and new strings, instead of being simple strings, are an extended regular expression pattern (see regex(3)) and an ed(1)-style replacement. Normally, the first occurrence of the pattern pattern in each word of the value is substituted with replacement. The ‘1’ modifier causes the substitution to apply to at most one word; the ‘g’ modifier causes the substitution to apply to as many instances of the search pattern pattern as occur in the word or words it is found in; the ‘W’ modifier causes the value to be treated as a single word (possibly containing embedded whitespace). As for the amp;:S modifier, the pattern and replacement are subjected to variable expansion before being parsed as regular expressions.
amp;:T	Replaces each word with its last path component (basename).
amp;:u	Removes adjacent duplicate words (like uniq(1)).
amp;:amp;? true_stringamp;: false_string
	If the variable name (not its value), when parsed as a .if conditional expression, evaluates to true, return as its value the true_string, otherwise return the false_string. Since the variable name is used as the expression, amp;:amp;? must be the first modifier after the variable name itself— which, of course, usually contains variable expansions. A common error is trying to use expressions like     ${NUMBERS:M42:?match:no} which actually tests defined(NUMBERS). To determine if any words match "42", you need to use something like:     ${"${NUMBERS:M42}" != amp;"amp;":?match:no}.
:old_string =new_string
	This is the AT&T V style substitution. It can only be the last modifier specified, as a ‘amp;:’ in either old_string or new_string is treated as a regular character, not as the end of the modifier. If old_string does not contain the pattern matching character ‘%’, and the word ends with old_string or equals it, that suffix is replaced with new_string. Otherwise, the first ‘%’ in old_string matches a possibly empty substring of arbitrary characters, and if the whole pattern is found in the word, the matching part is replaced with new_string, and the first occurrence of ‘%’ in new_string (if any) is replaced with the substring matched by the ‘%’. Both old_string and new_string may contain nested expressions. To prevent a dollar sign from starting a nested expression, escape it with a backslash.
amp;:@ varname@ string@
	This is the loop expansion mechanism from the OSF Development Environment (ODE) make. Unlike amp;.for loops, expansion occurs at the time of reference. For each word in the value, assign the word to the variable named varname and evaluate string. The ODE convention is that varname should start and end with a period, for example:     ${LINKS:@.LINK.@${LN} ${TARGET} ${.LINK.}@} However, a single-letter variable is often more readable:     ${MAKE_PRINT_VAR_ON_ERROR:@v@$v='${$v}'${.newline}@}
amp;:_[ =var ]
	Saves the current variable value in ‘$_’ or the named var for later reference. Example usage: M_cmpv.units = 1 1000 1000000 M_cmpv = S,., ,g:_:range:@i@+ $${_:[-$$i]} amp;\ \* $${M_cmpv.units:[$$i]}@:S,^,expr 0 ,1:sh .if ${VERSION:${M_cmpv}} < ${3.1.12:L:${M_cmpv}} Here ‘$_’ is used to save the result of the ‘:S’ modifier which is later referenced using the index values from ‘:range’.
amp;:Unewval
	If the variable is undefined, newval is the value. If the variable is defined, the existing value is returned. This is another ODE make feature. It is handy for setting per-target CFLAGS for instance:     ${_${.TARGET:T}_CFLAGS:U${DEF_CFLAGS}} If a value is only required if the variable is undefined, use:     ${VAR:D:Unewval}
amp;:Dnewval
	If the variable is defined, newval is the value.
amp;:L	The name of the variable is the value.
amp;:P	The path of the node which has the same name as the variable is the value. If no such node exists or its path is null, the name of the variable is used. In order for this modifier to work, the name (node) must at least have appeared on the right-hand side of a dependency.
amp;:amp;! cmdamp;!
	The output of running cmd is the value.
amp;:sh	The value is run as a command, and the output becomes the new value.
amp;::=str
	The variable is assigned the value str after substitution. This modifier and its variations are useful in obscure situations such as wanting to set a variable at a point where a target's shell commands are being parsed. These assignment modifiers always expand to nothing. The ' amp;::' helps avoid false matches with the AT&T V style ‘amp;:=’ modifier and since substitution always occurs, the ‘amp;::=’ form is vaguely appropriate.
amp;::?=str
	As for amp;::= but only if the variable does not already have a value.
amp;::+=str
	Append str to the variable.
amp;::!=cmd
	Assign the output of cmd to the variable.
amp;:amp;[range amp;]
	Selects one or more words from the value, or performs other operations related to the way in which the value is split into words. An empty value, or a value that consists entirely of white-space, is treated as a single word. For the purposes of the ' amp;:[]' modifier, the words are indexed both forwards using positive integers (where index 1 represents the first word), and backwards using negative integers (where index -1 represents the last word). The range is subjected to variable expansion, and the expanded result is then interpreted as follows:
index	Selects a single word from the value.
start amp;..end
	Selects all words from start to end, inclusive. For example, ' amp;:[2..-1]' selects all words from the second word to the last word. If start is greater than end, the words are output in reverse order. For example, ' amp;:[-1..1]' selects all the words from last to first. If the list is already ordered, this effectively reverses the list, but it is more efficient to use ' amp;:Or' instead of ' amp;:O:[-1..1]'.
amp;*	Causes subsequent modifiers to treat the value as a single word (possibly containing embedded whitespace). Analogous to the effect of amp;$* in Bourne shell.
0	Means the same as ' amp;:[*]'.
amp;@	Causes subsequent modifiers to treat the value as a sequence of words delimited by whitespace. Analogous to the effect of amp;$@ in Bourne shell.
amp;#	Returns the number of words in the value.

For compatibility with other make variants, ' include file ...' (without leading dot) is also accepted.

If the include statement is written as .-include or as .sinclude, errors locating and/or opening include files are ignored.

If the include statement is written as .dinclude, not only are errors locating and/or opening include files ignored, but stale dependencies within the included file are ignored just like in .MAKE.DEPENDFILE.

The operator may be any one of the following:

amp;|amp;|	Logical OR.
amp;&&	Logical AND; of higher precedence than ' amp;|amp;|'.

make only evaluates a conditional as far as is necessary to determine its value. Parentheses can be used to override the operator precedence. The boolean operator ' amp;!' may be used to logically negate an entire conditional. It is of higher precedence than ' amp;&&'.

The value of expression may be any of the following function call expressions:

definedamp;( varnameamp;)	Evaluates to true if the variable varname has been defined.
makeamp;( targetamp;)	Evaluates to true if the target was specified as part of Ns command line or was declared the default target (either implicitly or explicitly, see .MAIN) before the line containing the conditional.
emptyamp;( varname [: modifiers ]amp;)	Evaluates to true if the expansion of the variable, after applying the modifiers, results in an empty string.
existsamp;( pathnameamp;)	Evaluates to true if the given pathname exists. If relative, the pathname is searched for on the system search path (see .PATH).
targetamp;( targetamp;)	Evaluates to true if the target has been defined.
commandsamp;( targetamp;)	Evaluates to true if the target has been defined and has commands associated with it.

Expression may also be an arithmetic or string comparison. Variable expansion is performed on both sides of the comparison. If both sides are numeric and neither is enclosed in quotes, the comparison is done numerically, otherwise lexicographically. A string is interpreted as hexadecimal integer if it is preceded by 0x, otherwise it is a decimal floating-point number; octal numbers are not supported.

All comparisons may use the operators ' amp;==' and ' amp;!='. Numeric comparisons may also use the operators ' amp;<', ' amp;<=', ' amp;>' and ' amp;>='.

If the comparison has neither a comparison operator nor a right side, the expression evaluates to true if it is nonempty and its numeric value (if any) is not zero.

When make is evaluating one of these conditional expressions, and it encounters a (whitespace separated) word it doesn't recognize, either the "make" or "defined" function is applied to it, depending on the form of the conditional. If the form is ' .ifdef', ' .ifndef' or ' .if', the "defined" function is applied. Similarly, if the form is ' .ifmake' or ' .ifnmake', the "make" function is applied.

If the conditional evaluates to true, parsing of the makefile continues as before. If it evaluates to false, the following lines are skipped. In both cases, this continues until the corresponding ' .else' or ' .endif' is found.

amp;.for variable [variable ... ] in expression <make-lines> amp;.endfor

The expression is expanded and then split into words. On each iteration of the loop, one word is taken and assigned to each variable, in order, and these variables are substituted into the make-lines inside the body of the for loop. The number of words must come out even; that is, if there are three iteration variables, the number of words provided must be a multiple of three.

If ' .break' is encountered within a amp;.for loop, it causes early termination of the loop, otherwise a parse error.

amp;.ORDER: b a
b: a

amp;.SHELL: name=ksh path=/bin/ksh hasErrCtl=true \
        check="set -e" ignore="set +e" \
        echo="set -v" quiet="set +v" filter="set +v" \
        echoFlag=v errFlag=e newline="'\n'"

amp;.SUFFIXES: .c .o
amp;.c.o:
        cc -o ${.TARGET} -c ${.IMPSRC}

MAKEOBJDIRPREFIX and MAKEOBJDIR may only be set in the environment or on the command line to make and not as makefile variables; see the description of ' .OBJDIR' for more details.

The way that .for loop variables are substituted changed after NetBSD so that they still appear to be variable expansions. In particular this stops them being treated as syntax, and removes some obscure problems using them in .if statements.

The way that parallel makes are scheduled changed in NetBSD so that .ORDER and .WAIT apply recursively to the dependent nodes. The algorithms used may change again in the future.

• The .WAIT and .ORDER declarations and most functionality pertaining to parallelization. (GNU make supports parallelization but lacks the features needed to control it effectively.)
• Directives, including for loops and conditionals and most of the forms of include files. (GNU make has its own incompatible and less powerful syntax for conditionals.)
• All built-in variables that begin with a dot.
• Most of the special sources and targets that begin with a dot, with the notable exception of .PHONY, .PRECIOUS, and .SUFFIXES.
• Variable modifiers, except for the ‘:old=new’ string substitution, which does not portably support globbing with ‘%’ and historically only works on declared suffixes.
• The $> variable even in its short form; most makes support this functionality but its name varies.

Some features are somewhat more portable, such as assignment with +=, ?=, and !=. The .PATH functionality is based on an older feature VPATH found in GNU make and many versions of SVR4 make; however, historically its behavior is too ill-defined (and too buggy) to rely upon.

The $@ and $< variables are more or less universally portable, as is the $(MAKE) variable. Basic use of suffix rules (for files only in the current directory, not trying to chain transformations together, etc.) is also reasonably portable.

Historically the target/dependency FRC has been used to FoRCe rebuilding (since the target/dependency does not exist ... unless someone creates an FRC file).

There is no way of escaping a space character in a filename.

In jobs mode, when a target fails; make will put an error token into the job token pool. This will cause all other instances of make using that token pool to abort the build and exit with error code 6. Sometimes the attempt to suppress a cascade of unnecessary errors, can result in a seemingly unexplained ‘*** Error code 6’