README.md

gmtt

The GNU make table toolkit

Overview

gmtt is a collection of GNU make functions which among other things implement sorting and searching in tabular data. The intention is to manage build configuration data (like platform dependencies, compilation options, memory ranges, interrupt vectors in embedded system bootloaders and so on) in a way that doesn't force you out to the shell and reaps the benefits of simple relational database functionality. This has the additional benfit that you may get away with easier portability between Unix and Windows.

Tables

make has only one complex data type: list. make lists are just sequences of non-empty strings separated by whitespace characters (space, tab, linefeed). Thus tables are implemented as make lists with a first element denoting the table width (number of colons) as a decimal number: 3 tic tac toe is a table, while alpha beta gamma is not. Of course tables can (and usually will) be written as multi-line defines:

define test-tbl =
3
x1|y1  x2|y1  x3|y1
x1|y2  x2|y2  x3|y2
x1|y3  x2|y3  x3|y3
endef

There are some obvious restrictions on tables: make doesn't recognize string delimiters ("" etc.) so there is no way to have spaces inside one cell:

define illegal-table-there-are-no-strings-in-make = 
2
"This table has 4 rows really!"  x2|y1  
endef

As there is nothing like an empty string as member of a make list, we are also prohibited of using empty cells in a table. You will have to supply a string that denotes "nothing" and that will not get in your way further down the processing line. Here is an example where it is assumed that C-style comments don't harm

define table-with-empty-cells =
3
x1|y1      x2|y1       x3|y1
x1|y2    /*void*/      x3|y2
/*void*/   x2|y3       x3|y3
endef

You can alternatively use the spc-mask function (and its inverse spc-unmask) to concatenate strings with spaces into one table entry. But as always in make, using spaces in the processing of arguments requires much care and endurance to get it right.

Functions on tables

• sort / reverse sort
• select (with a 'where' clause, remotely comparable to SQL)
• map (apply a function on each row)

Arithmetic

All operations are supported for decimal, hexadecimal and octal numbers. The arithemtic range is currently ~64 places (don't exploit the maximum) but this limit can be set rather freely. Numbers are not fixed length (there is only an encapsulated internal representation which the user will never see) so they will be fast if small and slow if long. The bases of all operands can be mixed freely but of course the previously mentioned restrictions apply. The output base is decided by the base of the first operand: $(call add,0x1,1) will produce 0x2 but $(call add,1,0x1) will give 2. Generally the module relies on well-formed numbers:

• 0x1234 (a hexadecimal)
• -000123 (a negative octal)
• 123 (a decimal)

are all processed correctly, while

• 0x1234- (will be taken as negative)
• 12 34 (will produce garbage)

are not.

Functions on numbers

• add
• subtract
• multiply
• divide
• modulus
• log2
• bitwise or
• bitwise and
• bitwise xor
• bitwise not
• round up to a power of 2
• round down to a power of 2

Function List

String functions

$(call explode,stringlist,string)

Insert a blank after every occurrence of the strings from stringlist in string. This function serves mainly to convert a string into a list.

• $(call explode,0 1 2 3 4 5 6 7 8 9,0x1337c0de) --> 0 x1 3 3 7 c0 de

$(call implode,string-with-spaces)

Remove all spaces from the given string. Note that make is mostly unaware of escape characters and therefore takes all spaces verbatim.

• $(call implode,some\ awkward\ Windows\ path) --> some\awkward\Windows\path

$(call n-list,string,number-of-repetitions)

Create a list with exactly number-of-repetitions copies of a string.

• $(call n-list,foo,3) --> foo foo foo

$(call bincnt,binary-literal)

Count the binary-literal up by 1, yielding the following binary literal. Leading zeros are preserved.

• $(call bincnt,010011) -> 010100

$(call symgen)

Generate a different string at each call. The last generated symbol is accessible via $(last-symgen).

• $(call symgen) --> sym0

$(call interval,start,range[,step])

Create a list of integers starting at start and having range elements with an increase (or decrease) of step from one to the next. step is optional and defaults to 1 if not given.

Example:

• $(call interval,5,5) --> 5 6 7 8 9
• $(call interval,2,3,100) --> 2 102 202

$(call lpad,string,final-width,padding-character)

Left-pad an alphanumeric string with the given character up to the given length. If the original string is longer than final-width, nothing is padded. Expample:

• $(call lpad,123,7,0) --> 0000123
• $(call lpad,123,7,-) --> ----123

$(call lstrip,string,prefix)

Remove a prefix from the given string. If the prefix doesn't exist, the string is unchanged.

• $(call lstrip,0xABCD,0x) --> ABCD

$(call str-eq,string1,string2)

Compare two strings on equality. Strings are allowed to have blanks. Return non-empty if string $1 and $2 are identical, empty string otherwise.

• $(call str-eq,yes,no) --> (empty string)
• $(call str-eq,yes ,yes) --> (empty string)
• $(call str-eq,yes ,yes ) --> t

$(call str-ne,string1,string2)

Compare two strings on inequality. The obvious inverse to str-eq. Strings are allowed to have blanks. Return empty if string $1 and $2 are identical, a non-empty string otherwise.

• $(call str-eq,yes,no) --> t
• $(call str-eq,yes ,yes) --> t
• $(call str-eq,yes ,yes ) --> (empty string)

$(call str-le,string1,string2)

Compare two strings lexically for string1 less-or-equal string2. Lexical ordering means that 'aa' < 'aaa' < 'aab' < 'ab'. The empty string always compares smaller than any other string. The strings may contain spaces but leading and trailing spaces are not considered in the comparison and multiple interior spaces are substituted by a single one. Spaces compare smaller than downcase characters but greater than upcase. In short: you should know what you are doing if you have spaces inside your strings. Examples:

• $(call str-le,aaa,ab)) --> t
• $(call str-le, ab aa,aa)) --> (empty string)
• $(call str-le,aa,aa)) --> t
• $(call str-le,aa,a)) --> (empty string)
• $(call str-le,a,)) --> (empty string)
• $(call str-le,,a)) --> t
• $(call str-le,MacGyver John,Mac Gyver John)) --> t
• $(call str-le,macgyver john,mac gyver john)) --> (empty string)

$(call str-ge,string1,string2)

Compare two strings lexically for string1 greater-or-equal string2. Lexical ordering means that 'aa' < 'aaa' < 'aab' < 'ab'. The empty string always compares smaller than any other string. The strings may contain spaces but leading and trailing spaces are not considered in the comparison and multiple interior spaces are substituted by a single one. Spaces compare smaller than downcase characters but greater than upcase. In short: you should know what you are doing if you have spaces inside your strings. Examples:

• $(call str-ge,aaa,ab)) --> (empty string)
• $(call str-ge, ab aa,aa)) --> t
• $(call str-ge,aa,aa)) --> t
• $(call str-ge,aa,a)) --> t
• $(call str-ge,a,)) --> t
• $(call str-ge,,a)) --> (empty string)
• $(call str-ge,MacGyver John,Mac Gyver John)) --> (empty string)
• $(call str-ge,macgyver john,mac gyver john)) --> t

$(call str-match,string1,string2)

Compare two strings on equality under wildcard substitution. The wildcard can appear in both of the strings. Return t if string1 and string2 match where the first % is taken as wildcard, return empty string otherwise.

• $(call str-match,Mickey%Mouse,Mickey Mouse)) --> t
• $(call str-match,Mickey%,MickeyMouse)) --> t
• $(call str-match,Mickey%,)) --> (empty string)
• $(call str-match,Mickey %ouse,Mickey Mouse)) --> t
• $(call str-match,MickeyMouse,MickeyMouse%)) --> t
• $(call str-match,,%)) --> t

$(call glob-match,string,pattern)

Try to match the string with the pattern, applying glob-syntax. Glob-syntax is well known from the shell and https://en.wikipedia.org/wiki/Glob_(programming) All characters match themselves except:

• * - zero or more arbitrary chars
• ? - exactly one arbitrary char
• [] - exactly one character from the set designated inside the brackets:
  • [abc] - explicit, matches one of a, b or c
  • [a-z] - range, matches one of a,b...z. The possibly ranges can be taken from $(all-chars)
  • []abc] - first position is the only way to match a ]
  • [-abc] - first or last position is the only way to match a -
  • [!a-z] - ! inverts the match, i.e. everything but a..z

If no match occurred, the (empty string) is returned. If the string matches, it is returned with all parts corresponding to one of the above wildcards separated by space. As GNUmake treats everything which is different from the empty string as true, this function serves the simple matching test as well as a string dissection by wildcard patterns. Spaces inside your string are automatically preserved. The elements of the string are returned but with all original spaces replaced by an internal character to circumvent the rule that GNUmake treats spaces as list element separators. Use spc-unmask to remove these and restore the spaces. This caveat is necessary to allow access to the elements with a constant index, in case you are dissecting a string. Examples:

• $(call glob-match,Linux 2.6.32-431.el6.i686,Linux 2.6.*.i686) --> Linux§2.6. 32-431.el6 .i686
• $(call glob-match,down/to/unknown/dir/file.txt,down/*/*/*/*.txt) --> down/ to / unknown / dir / file .txt

$(call chop-str,string,group-string[,group-string[,group-string[,group-string]]])

Dissect given string into chunks of characters belonging to the same group-string and return the result as a separator list. A group-string is a tuple of an identifier string and a group-characters string with a space in between (e.g.digit 0123456789). Each character in the given string is tested for membership in one of the group-characters. If adjacent characters belong to the same group-string they stay glued together otherwise a space is inserted. The identifier is attached as a prefix to the character chunks so that subsequent functions can distinguish the groups in the result. If a character in string is not found in any of the group strings, it is dropped but still separates chunks. This function mainly serves to cut formated alphanumeric strings apart. How to handle strings with spaces: if you don't need the spaces after parsing (spaces only separate the string parts) then you don't have to do anything, simply call the function with the string as parameter. If you want to preserve spaces, then you must call chop-str-spc which will leave behind the same list of string parts but with spaces still replaced by an internal character. The space character is handled as if it is part of the first group-string so all sequences of characters from the first group which are separated by space will become one chunk. This way you can still handle the function result as a list after calling. When stepping through this list you can remove the internal character by simply applying $(call spc-unmask,_string-chunk-from-result_). Examples:

• $(call chop-str,Linux 4.13.0-17-generic,A $(-alpha-as-str),1 $(-digit-as-str),. .-+?) --> A¤Linux 1¤4 .¤. 1¤13 .¤. 1¤0 .¤- 1¤17 .¤- A¤generic
• $(call chop-str,Thu Nov 30 18:43:22 CET 2017,alpha $(-alpha-as-str),num $(-digit-as-str),sep :) --> alpha¤Thu alpha¤Nov num¤30 num¤18 sep¤: num¤43 sep¤: num¤22 alpha¤CET num¤2017
• $(call chop-str-spc,Thu Nov 30 18:43:22 CET 2017,alpha $(-alpha-as-str),num $(-digit-as-str),sep :) --> alpha¤Thu§Nov§ num¤30 alpha¤§ num¤18 sep¤: num¤43 sep¤: num¤22 alpha¤§CET§ num¤2017

Predefined character classes for chop-str et al

• $(-upper-as-str) := ABCDEFGHIJKLMNOPQRSTUVWXYZ
• $(-lower-as-str) := abcdefghijklmnopqrstuvwxyz
• $(-digit-as-str) := 0123456789
• $(-xdigit-as-str) := 0123456789abcdefABCDEF
• $(-alpha-as-str) := ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz
• $(-alnum-as-str) := 0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz

$(call spc-mask,string)

Replace all occurrences of space in the given string with the $(-spacereplace) character. This function can be used to convert strings with spaces into valid list elements (which must not contain spaces).

$(call spc-unmask,string)

Replace all occurrences of the character $(-spacereplace) with a true space character. The inverse of the spc-mask function.

$(call drop-prfx,separator-list)

Clear the prefixes from a separator list. A separator list is a make list with a prefix name prepended to each list element and the internal character $(-separator) between them e.g. alpha¤Linux num¤4 dot¤. num¤2.

$(call drop-sufx,separator-list)

Clear the suffixes (usually the data) from a separator list. What remains is a list of prefixes only . A separator list is a make list with a prefix name prepended to each list element and the internal character $(-separator) between them e.g. alpha¤Linux num¤4 dot¤. num¤2.

$(call get-prfx-val,separator-list,prefix-1 [prefix-2..prefix-n][,m-th[,n-th]])

Retrieve the value with the given prefixes from the separator list. Optionally select not the first but the mth occurrence of this prefix value or a range from the m-th to the n-th (inclusive) occurrence.

$(call get-prfx-range,separator-list,opening-prefix,closing-prefix[,n-th])

Retrieve the part of the separator list from opening-prefix to closing-prefix (inclusive). Alternatively return the n-th occurrence of the opening-prefix/closing-prefix combination. This function servers to select e.g. parenthesized parts of a string.

$(call get-sufx-range,separator-list,opening-suffix,closing-suffix[,n-th])

Retrieve the part of the separator list from opening-suffix to closing-suffix (inclusive). Alternatively return the n-th occurrence of the opening-suffix/closing-suffix combination. This function servers to select e.g. parenthesized parts of a string.

List functions

$(call uniq-sufx,list,binary-literal)

Add a ¤ (Character 164) and a unique binary number to all elements of the list. The binary-literal must be present and can be any combination of 0's and 1's.

• $(call uniq-sufx,The quick brown fox,0) --> The¤0 quick¤1 brown¤10 fox¤11
• $(call uniq-sufx,The quick brown fox,111) --> The¤111 quick¤1000 brown¤1001 fox¤1010

$(call sort-all,list)

Sort a list without dropping duplicates. Built-in $(sort) will drop them which is sometimes not what you want. Note: list elements must not contain ¤ (Character 164) as this is character is used internally for processing.

$(call rev-list,list)

Reverse the order of the elements of a list.

• $(call rev-list,The quick brown fox) --> fox brown quick The

$(call list2param,list)

Convert the given list to a string where each list element is separated by a comma. Multiple spaces get reduced to one comma and there is no comma at the start and end of the list.

• $(call list2param,The quick brown fox) --> The,quick,brown,fox

$(call quote-params,[param1,[param2,[param3,[param4,[param5,[param6,[param7,[param8,[param9]]]]]]]]])

Convert all given parameters into a string with the format of the parameter list, i.e. a string with the verbatim parameters separated by commas. Use this function as a macro to channel an unknown number of function parameters into a string.

Examples:

• test = $(info <$(quote-params)>)
• $(call test,1,2,3,4,5,6,7,8,9) -> <1,2,3,4,5,6,7,8,9>
• $(call test,,2,3,4,5,6,7,8,9) -> <,2,3,4,5,6,7,8,9>
• $(call test,1) -> <1>
• $(call test,1,2,3,4,5,6,,,9) -> <1,2,3,4,5,6,,,9>
• $(call test,,,,,,,,,9) -> <,,,,,,,,9>
• $(call test,,,,,,,,,) -> <,,,,,,,,>
• $(call test,) -> <>

$(call exec,quoted-func,params)

Evaluate the quoted-func code in place, using the params as parameter list. quoted-func is any GNUmake 'code' which could also appear on the rhs of a variable definition but in the quoted form: every appearance of '$' is to be quoted with an extra '$' (see examples). params is a list of parameters separated by commas (see list2params function) which is given as the call parameters $1,$2,etc. to the function expression.

$(call up-to,word,list)

Return first part of list up to but excluding the first occurrence of word. If word is not in list, the whole list is returned. Examples:

• $(call up-to,baz,foo bar baz) -> foo bar
• $(call up-to,foo,foo bar baz) -> (empty list)

$(call index-of,word,list)

Return the index of the first occurrence of word if present or the empty list. Indexing starts at 0, contrary to the make-internal behaviour of numbering lists from 1! Note that you can always have an index starting at 1 by prefixing the list at call time with the $(-never-matching) element (or a simple underscore, if you are sure that it is never a real element), i.e. $(call index-of,foo,$(-never-matching) foo bar baz) will give index 1. Examples:

• $(call index-of,foo,foo bar baz) -> 0

$(call up-from,word,list)

Return the portion of list following the first occurrence of word. If word is not in list, the empty string/list is returned. Examples:

• $(call up-from,foo,foo bar baz) -> bar baz
• $(call up-from,baz,foo bar baz) -> (empty list)

$(call down-to,word,list)

Return the last part of list, searching from the end down to but excluding the last occurrence of word. If word is not in list, the whole list is returned. This operation is helpful in dissecting e.g. path expressions (´/´ substituted by ´ ´ or ´/ ´) Examples:

• $(call down-to,baz,foo baz bar baz) -> (empty list)
• $(call down-to,foo,foo foo bar baz) -> bar baz

$(call down-from,word,list)

Return the first part of list, searching from the end down to but excluding the last occurrence of word. If word is not in list, the empty list is returned. This operation is helpful in dissecting e.g. path expressions (´/´ substituted by ´ ´ or ´/ ´) Examples:

• $(call down-from,baz,foo baz bar baz) -> foo baz bar
• $(call down-from,foo,foo bar baz) -> (empty list)

Arithmetic functions

$(call add,num1,num2)

Calculate num1+num2. Both arguments are signed integers. Numeric bases can be mixed, the result is given in the same base as num1. Examples:

• $(call add,-1,0) --> -1
• $(call add,0x12,13) --> 0x1f

$(call sub,num1,num2)

Calculate num1-num2. Both arguments are signed integers. Numeric bases can be mixed, the result is given in the same base as num1. Examples:

• $(call sub,12,-10) --> 22
• $(call sub,012,-10) --> 024

$(call cmp,num1,num2)

Compare num1 with num2 and emit the characters <,> or =. Both arguments are signed integers. Numeric bases can be mixed. Examples:

• $(call cmp,12,-14) --> >
• $(call cmp,012,10) --> =

$(call int-(le|lt|ge|gt|eq),num1,num2)

Compare num1 with num2 and emit the characters <,>,= or the empty string. The postfix alternatives decide which comparison is executed: less-than (lt), less-or-equal (le), greater-than (gt), greater-or-equal (ge) and equal (eq). Both arguments are signed integers. Numeric bases can be mixed. The result can be used in an $(if ) expression. Examples:

• $(call int-ge,12,-14) --> >
• $(call int-le,12,-14) --> (empty string)
• $(if $(call int-eq,012,10),equal,not equal) --> equal

$(call mul,num1,num2)

Multiply integer num1 with integer num2. Both arguments can be signed. Numeric bases can be mixed, the result is given in the same base as num1. Examples:

• $(call mul,-1,0) --> 0
• $(call mul,-0x1,0) --> 0x0
• $(call mul,03,-7) --> -025

$(call div,num1,num2)

Divide integer num1 by integer num2. Both arguments can be signed. Numeric bases can be mixed, the result is given in the same base as num1. Examples:

• $(call div,9876543210,0x13) --> 519818063
• $(call div,0x9876543210,16) --> 0x987654321

Table functions

$(call sort-tbl,table,key-gen)

Sort a table by lines. Key comparison is done by lexical ordering. Lexical ordering means that 'aa' < 'aaa' < 'aab' < 'ab'. The empty string always compares smaller than any other string. The strings may contain spaces but leading and trailing spaces are not considered in the comparison and multiple interior spaces are substituted by a single one. Spaces compare smaller than downcase characters but greater than upcase. In short: you should know what you are doing if you have spaces inside your strings.

The key-gen is a function expression which shall yield the key for the sorting comparison. Note that the key is always evaluated as a string - this means that you usually can't use numerals directly as key but have to left-pad them with 0's to make them the same length.

$(call select,col-nrs,table,where-clause)

Select all rows from a table which fulfill the where-clause and pick the subset of col-nrs from these rows to form an output list. A gmtt table is a list with a leading decimal which denotes the number of columns in this 'table'. See the documentation for gmtt at [https://github.com/markpiffer/gmtt]. The where-clause is a function or a 'exec' expression (i.e. function expression written into the parameter place directly) which receives the elements of each row of the table in order as parameters $1,$2,$3 etc. Note: the function call/exec needs to be $-quoted, that is, every '$' that appears as variable reference must be doubled '$$'. See the examples below. The clause shall return true (non-empty string) or false (empty string) to accept/reject each rows elements into/from the result of the select. The selection is limited to the column numbers given in col-nrs, in their respective order. Do not confuse the col-nrs subset with the parameters given to the 'where' function, the former is just a possibly reordered subset of the latter which is formed after the 'where' function accepted the record. (select mimics a SQL SELECT model, price FROM cars WHERE color="red" and returns effectively a list of subsets from all positively selected rows. If you prepend the result list with its column count, you have a new gmtt table)

Exapmle:

• test-tbl := 4 foo bar baz 11 foo bar baf 22 faa bar baz 33
• $(call select,3 1 2 3,$(test-tbl),$$(call str-match,$$1,%oo)) --> baz foo bar baz baf foo bar baf The same can be achieved, if we use a function as where clause:
• ends-in-oo = $(call str-match,$1,%oo)
• $(call select,3 1 2 3,$(test-tbl),$$(call ends-in-oo,$$1)) --> baz foo bar baz baf foo bar baf

$(call join-tbl,table1,table2[,NIL-value])

Join two tables side by side. The resulting table has as many columns as both input tables combined. The optional NIL value will be used to fill empty places in the resulting table if one input table has less rows than the other. If the NIL value is not given, the shorter of the two tables will be repeated as many times as necessary to fill up the number of rows to that of the longer table. The order of the tables is preserved.

Examples:

• $(call join-tbl,1 one two,2 first 1st second 2nd third 3rd) --> 3 one first 1st two second 2nd one third 3rd
• $(call join-tbl,2 one apple two oranges,1 1st 2nd 3rd,/*NIL*/) --> 3 one apple 1st two oranges 2nd /*NIL*/ /*NIL*/ 3rd

File System Functions

$(call wildcard-rec,list-of-globs)

"wildcard recursive" is an extension to the built in wildcard function. It turns a list of glob expressions into the list of all files (for expressions not ending in /) or directories (expressions ending in /) which are reachable in the file system by these globs. The given paths obey usual glob syntax (*, ? etc.) with a special extension: there is one % or ** allowed in the path expression which starts a recursive search of the file tree at this position of the glob. To explain this: the usual glob syntax * expresses just one directory level down, e.g. in project/*/foo each file "foo" in any of the subdirectories of project is found. There is no way to find a "foo" file deeper down the hierarchy - you would need to know how many levels deep the file is located e.g. project/*/*/*/foo. While this syntax is absolutely allowed, the expression project/**/foo addresses the former and all other existing directory levels in the tree no matter how deep. More complex patterns are also allowed:

• project**/foo will select all first level directories which start with project: also directory project001 will be considered
• project/**foo will select all files in the tree under project which end in foo
• **.h will select all C header files in the current file tree

Example: for the following directory structure:

.(working directory)
|
+---foo
|    |
|    +-
|
+---barfoo
     |   foo.c
     |
     +---bar
     |       bar
     |
     +---foo
     |       bar.foo
     |       foo.txt
     |
     \---foobar

the exemplary globs on the left will create the outputs to the right:

**.c --> ./barfoo/foo.c 
foo*/ --> foo/
bar*/ --> barfoo/
**/foo*/ --> ./barfoo/foo/ ./barfoo/foobar/ ./foo/
**/foo* --> ./barfoo/foo.c ./barfoo/foo/foo.txt
**/*foo* --> ./barfoo/foo.c ./barfoo/foo/bar.foo ./barfoo/foo/foo.txt
**/bar --> ./barfoo/bar/bar
**/bar/ --> ./barfoo/bar/

Miscellaneous Functions

$(call head,list)

Functionally identical to $(firstword ). Return the head (first element) of a list.

$(call tail,list)

Return the second and subsequent elements of a list as a new list.

$(call clr-comments,lines-with-hash-comments)

Return the given string cleared from line comments '#'. The string may contain spaces and newlines, these are returned untouched. Example:

• config.txt:
• first #comment
• # second comment
• third # comment
• cfg-file = $(call clr-comments,$(file < config.txt))

Output:

• first
• third

$(call while,quoted-condition,quoted-code[,quoted-exit-statements])

Execute a while loop of quoted-code as long as condition is true (not the empty string). All code statements need to be given as quoted make code (replace $ with $$). The code in quoted-code and quoted-exit-statements is evaled in the while loop. This means that you can use variable assignments like in ordinary code BUT the assignments will be visible outside of the while loop! The optional quoted-exit-statement is executed always when leaving the while loop, even when the loop body was never executed.

• Example: filter out the -mllvm flags from the CFLAGS variable and put them in an extra variable

CFLAGS := -DFOO -DBAR -mllvm llvmflag1 -mllvm llvmflag2
$(call while, $$(call glob-match,$(space)$$(CFLAGS),*-mllvm *),\
   tmp := $$(call glob-match,$(space)$$(CFLAGS),*-mllvm *)   $(newline)\
   rest := $$(call spc-unmask,$$(word 3,$$(tmp)))            $(newline)\
   llvm_flg := $$(firstword $$(rest))                        $(newline)\
   CFLAGS := $$(firstword $$(tmp)) $$(call tail,$$(rest))    $(newline)\
   $$(info [[[$$(tmp) --- $$(rest) --- $$(CFLAGS) ]]])       $(newline)\
   MLLVM_FLAGS+=$$(llvm_flg),\
MLLVM_FLAGS := $$(strip $$(MLLVM_FLAGS))                     $(newline)\
CFLAGS := $$(call spc-unmask,$$(CFLAGS))\
)

• Condition: as long as glob-match returns a match of -mllvm in CFLAGS. There is a small trick to avoid an empty match on the first *: we sneak a $(space) character in at the beginning of the string so that the match will never be empty.
• Body (notice the quoting of the newlines - this is necessary for $(eval) to correctly interpret the code):
  • Extract the output of glob-match into a temporary variable (spaces are replaced in this output, see glob-match). The output is a list of three elements: all characters (*) up to -mllvm , the string -mllvm itself (notice the space at the end) and all characters (*) following it.
  • Convert back the third element (rest of CFLAGS behind the first -mllvm match) into a string with spaces and put it into rest
  • Pluck the argument to the -mllvm flag (which is the first element of rest) into llvm_flg
  • Modify CFLAGS to contain everything except -mllvm and its argument
  • Print the variables as a debugging aid
  • Append the newly found argument of the -mllvm flag to the output variable
• Exit statements:
  • Pretty-print MLLVM_FLAGS
  • Remove the space replacement characters from CFLAGS which have accumulated during the loop
• Output:

[[[ -DFOO§-DBAR§ -mllvm§ llvmflag1§-mllvm§llvmflag2 --- llvmflag1 -mllvm llvmflag2 --- -DFOO§-DBAR§ -mllvm llvmflag2 ]]]
[[[ -DFOO§-DBAR§§ -mllvm§ llvmflag2 --- llvmflag2 --- -DFOO§-DBAR§§  ]]]
CFLAGS = -DFOO -DBAR
MLLVM_FLAGS = llvmflag1 llvmflag2

$(call verbose,[string0],[string1],[string2],[string3],[string4],[string5],[string6],[string7],[string8],[string9])

Write strings to stdout depending on warning level. The global variable VERBOSITY needs to be defined as a string of numerals from 0..9. This will trigger the output of the according strings for all $(call verbosity ...) invocations in the makefile. Verbosity is additive, not hierarchical, i.e. VERBOSITY=9 will not effect the output of all warnings/info-strings from level 0 to 9 but only for level 9. To output a combination or selection, set e.g. VERBOSITY = 01289. Examples:

• VERBOSITY = 0
• $(call verbose,warninglevel 0,warninglevel 1,warninglevel 2,info 1,info 2,info 3) -> warninglevel 0
• VERBOSITY = 2
• $(call verbose,warninglevel 0,warninglevel 1,warninglevel 2,info 1,info 2,info 3) -> warninglevel 2
• VERBOSITY = 012345
• $(call verbose,warninglevel 0,warninglevel 1,warninglevel 2,info 1,info 2,info 3) -> warninglevel 0 warninglevel 1 warninglevel 2 info 1 info 2 info 3