Instantly share code, notes, and snippets.

What would you like to do?
# Hello, and welcome to makefile basics.
# You will learn why `make` is so great, and why, despite its "weird" syntax,
# it is actually a highly expressive, efficient, and powerful way to build
# programs.
# Once you're done here, go to
# to learn SOOOO much more.
# To do stuff with make, you type `make` in a directory that has a file called
# "Makefile". You can also type `make -f <makefile>` to use a different
# filename.
# A Makefile is a collection of rules. Each rule is a recipe to do a specific
# thing, sort of like a grunt task or an npm package.json script.
# A rule looks like this:
# <target>: <prerequisites...>
# <commands>
# The "target" is required. The prerequisites are optional, and the commands
# are also optional, but you have to have one or the other.
# Type "make" and see what happens:
@# todo: have this actually run some kind of tutorial wizard?
@echo "Please read the 'Makefile' file to go through this tutorial"
# By default, the first target is run if you don't specify one. So, in this
# dir, typing "make" is the same as typing "make tutorial"
# By default, make prints out the command before it runs it, so you can see
# what it's doing. This is a departure from the "success should be silent"
# UNIX dogma, but without that default, it'd be very difficult to see what
# build logs etc are actually doing.
# To suppress the output, we've added @ signs before each line, above.
# Each line of the command list is run as a separate invocation of the shell.
# So, if you set a variable, it won't be available in the next line! To see
# this in action, try running `make var-lost`
export foo=bar
echo "foo=[$$foo]"
# Notice that we have to use a double-$ in the command line. That is because
# each line of a makefile is parsed first using the makefile syntax, and THEN
# the result is passed to the shell.
# Let's try running both of the commands in the *same* shell invocation, by
# escaping the \n character. Run `make var-kept` and note the difference.
export foo=bar; \
echo "foo=[$$foo]"
# Now let's try making something that depends on something else. In this case,
# we're going to create a file called "result.txt" which depends on
# "source.txt".
result.txt: source.txt
@echo "building result.txt from source.txt"
cp source.txt result.txt
# When we type `make result.txt`, we get an error!
# $ make result.txt
# make: *** No rule to make target `source.txt', needed by `result.txt'. Stop.
# The problem here is that we've told make to create result.txt from
# source.txt, but we haven't told it how to get source.txt, and the file is
# not in our tree right now.
# Un-comment the next ruleset to fix the problem.
# @echo "building source.txt"
# echo "this is the source" > source.txt
# Run `make result.txt` and you'll see it first creates source.txt, and then
# copies it to result.txt. Try running `make result.txt` again, and you'll see
# that nothing happens! That's because the dependency, source.txt, hasn't
# changed, so there's no need to re-build result.txt.
# Run `touch source.txt`, or edit the file, and you'll see that
# `make result.txt` re-builds the file.
# Let's say that we were working on a project with 100 .c files, and each of
# those .c files we wanted to turn into a corresponding .o file, and then link
# all the .o files into a binary. (This is effectively the same if you have
# 100 .styl files to turn into .css files, and then link together into a big
# single concatenated main.min.css file.)
# It would be SUPER TEDIOUS to create a rule for each one of those. Luckily,
# make makes this easy for us. We can create one generic rule that handles
# any files matching a specific pattern, and declare that we're going to
# transform it into the corresponding file of a different pattern.
# Within the ruleset, we can use some special syntax to refer to the input
# file and the output file. Here are the special variables:
# $@ The file that is being made right now by this rule (aka the "target")
# You can remember this because it's like the "$@" list in a
# shell script. @ is like a letter "a" for "arguments.
# When you type "make foo", then "foo" is the argument.
# $< The input file (that is, the first prerequisite in the list)
# You can remember this becasue the < is like a file input
# pipe in bash. `head <foo.txt` is using the contents of
# foo.txt as the input. Also the < points INto the $
# $^ This is the list of ALL input files, not just the first one.
# You can remember it because it's like $<, but turned up a notch.
# If a file shows up more than once in the input list for some reason,
# it's still only going to show one time in $^.
# $? All the input files that are newer than the target
# It's like a question. "Wait, why are you doing this? What
# files changed to make this necessary?"
# $$ A literal $ character inside of the rules section
# More dollar signs equals more cash money equals dollar sign.
# $* The "stem" part that matched in the rule definition's % bit
# You can remember this because in make rules, % is like * on
# the shell, so $* is telling you what matched the pattern.
# You can also use the special syntax $(@D) and $(@F) to refer to
# just the dir and file portions of $@, respectively. $(<D) and
# $(<F) work the same way on the $< variable. You can do the D/F
# trick on any variable that looks like a filename.
# There are a few other special variables, and we can define our own
# as well. Most of the other special variables, you'll never use, so
# don't worry about them.
# So, our rule for result.txt could've been written like this
# instead:
result-using-var.txt: source.txt
@echo "buildling result-using-var.txt using the $$< and $$@ vars"
cp $< $@
# Let's say that we had 100 source files, that we want to convert
# into 100 result files. Rather than list them out one by one in the
# makefile, we can use a bit of shell scripting to generate them, and
# save them in a variable.
# Note that make uses := for assignment instead of =
# I don't know why that is. The sooner you accept that this isn't
# bash/sh, the better.
# Also, usually you'd use `$(wildcard src/*.txt)` instead, since
# probably the files would already exist in your project. Since this
# is a tutorial, though we're going to generate them using make.
# This will execute the shell program to generate a list of files.
srcfiles := $(shell echo src/{00..99}.txt)
# How do we make a text file in the src dir?
# We define the filename using a "stem" with the % as a placeholder.
# What this means is "any file named src/*.txt", and it puts whatever
# matches the "%" bit into the $* variable.
@# First things first, create the dir if it doesn't exist.
@# Prepend with @ because srsly who cares about dir creation
@[ -d src ] || mkdir src
@# then, we just echo some data into the file
@# The $* expands to the "stem" bit matched by %
@# So, we get a bunch of files with numeric names, containing their number
echo $* > $@
# Try running `make src/00.txt` and `make src/01.txt` now.
# To not have to run make for each file, we define a "phony" target that
# depends on all of the srcfiles, and has no other rules. It's good
# practice to define your phony rules in a .PHONY declaration in the file.
# (See the .PHONY entry at the very bottom of this file.)
# Running `make source` will make ALL of the files in the src/ dir. Before
# it can make any of them, it'll first make the src/ dir itself. Then
# it'll copy the "stem" value (that is, the number in the filename matched
# by the %) into the file, like the rule says above.
# Try typing "make source" to make all this happen.
source: $(srcfiles)
# So, to make a dest file, let's copy a source file into its destination.
# Also, it has to create the destination folder first.
# The destination of any dest/*.txt file is the src/*.txt file with
# the matching stem. You could just as easily say that %.css depends
# on %.styl
dest/%.txt: src/%.txt
@[ -d dest ] || mkdir dest
cp $< $@
# So, this is great and all, but we don't want to type `make dest/#.txt`
# 100 times!
# Let's create a "phony" target that depends on all the destination files.
# We can use the built-in pattern substitution "patsubst" so we don't have
# to re-build the list. This patsubst function uses the same "stem"
# concept explained above.
destfiles := $(patsubst src/%.txt,dest/%.txt,$(srcfiles))
destination: $(destfiles)
# Since "destination" isn't an actual filename, we define that as a .PHONY
# as well (see below). This way, Make won't bother itself checking to see
# if the file named "destination" exists if we have something that depends
# on it later.
# Let's say that all of these dest files should be gathered up into a
# proper compiled program. Since this is a tutorial, we'll use the
# venerable feline compiler called "cat", which is included in every
# posix system because cats are wonderful and a core part of UNIX.
kitty: $(destfiles)
@# Remember, $< is the input file, but $^ is ALL the input files.
@# Cat them into the kitty.
cat $^ > kitty
# Note what's happening here:
# kitty -> (all of the dest files)
# Then, each destfile depends on a corresponding srcfile
# If you `make kitty` again, it'll say "kitty is up to date"
# Let's update just ONE of the source files, and see what happens
# Run this: touch src/25.txt; make kitty
# Note that it is smart enough to re-build JUST the single destfile that
# corresponds to the 25.txt file, and then concats them all to kitty. It
# *doesn't* re-generate EVERY source file, and then EVERY dest file,
# every time
# It's good practice to have a `test` target, because people will come to
# your project, and if there's a Makefile, then they'll expect `make test`
# to do something.
# We can't test the kitty unless it exists, so we have to depend on that.
test: kitty
@echo "miao" && echo "tests all pass!"
# Last but not least, `make clean` should always remove all of the stuff
# that your makefile created, so that we can remove bad stuff if anything
# gets corrupted or otherwise screwed up.
rm -rf *.txt src dest kitty
# What happens if there's an error!? Let's say you're building stuff, and
# one of the commands fails. Make will abort and refuse to proceed if any
# of the commands exits with a non-zero error code.
# To demonstrate this, we'll use the `false` program, which just exits with
# a code of 1 and does nothing else.
$(MAKE) kitty # The special var $(MAKE) means "the make currently in use"
false # <-- this will fail
echo "should not get here"
.PHONY: source destination clean test badkitty

This comment has been minimized.

ceejbot commented Jan 15, 2015

You cover what I think is the important magic: rules, pattern substitution, special variables. With those building blocks, you can write just about anything. I tripped over the shell command to find files, but then I read backwards & saw the wildcard explanation, which I liked.


This comment has been minimized.

rmg commented Jan 15, 2015

Some accidental (or potentially confusing to the target audience) indentation on lines 82-88.

For line 191, what about $^?


This comment has been minimized.


isaacs commented Jan 15, 2015

@rmg: Thanks! Added an explanation of $^, and used it to build the kitty.


This comment has been minimized.

ericelliott commented Jan 15, 2015

Cool! We've been discussing build tools for JavaScript programs for Cloverfield. The first example was coded with make by @jesstelford.



This comment has been minimized.

ericelliott commented Jan 15, 2015

Does this work on Windows? That's a concern that @vivainio has brought up.

I have built lots of things with Make on Windows, but I used cygwin for years.


This comment has been minimized.

jellea commented Jan 15, 2015

Great! You should totally add this to!


This comment has been minimized.

DarthGandalf commented Jan 15, 2015

# Note that make uses := for assignment instead of =
# I don't know why that is. The sooner you accept that this isn't
# bash/sh, the better. 

Make has several different assignments: =, :=, +=, ?=, !=


This comment has been minimized.

adamdicarlo commented Jan 16, 2015

This is really awesome! I don't follow the explanation of .PHONY though. The section in the GNU make docs made sense, but coming back here I still don't follow this explanation. Isn't .PHONY just a way to label targets as not producing a file with the target name? What's that got to do with not having to run make for each file?


This comment has been minimized.

adamdicarlo commented Jan 16, 2015

Oh! I get it. It's the difference between phony and .PHONY that tripped me up. I think that changing

It's good practice to define your phony rules in a .PHONY declaration in the file.

to something like

It's good practice to tell make which of your rules are phony; make will recognize that any dependencies of a special .PHONY target are phonies themselves

would have made it clearer, at least to me 😃


This comment has been minimized.

elrasguno commented Jan 24, 2015

Hi, so,

  1. Thanks for doing this. Imma make some stuff now.
  2. You spelled "meow" wrong, and I don't understand why nobody has said anything yet.
  3. So here's what I can't figure out yet, and you're being so helpful, so maybe you can help

I have a directory structure that looks like this:

  • aaa
  • bbb

Final Update: Answering my own question if anyone is interested

And a recently updated (again) Makefile that looks like this:

source_files := $(wildcard */*.md)
# source_files: aaa/ aaa/ bbb/
dest_files := $(
# dest_files: aaa/README.html aaa/MEOW.html bbb/MIAO.html

all: docs

    md2html $< > $@

docs: $(dest_files)
    rm -rf $(dest_files)

And that will generate .html files for .md files in their respective directories.

Thanks again for making this gist!


This comment has been minimized.

nik-kor commented Feb 14, 2015

Great tutorial! Thank you!

I don't why(maybe different version of make, environment, etc), but this command

srcfiles := $(shell echo src/{00..99}.txt)

didn't produce the needed sequence, so

->make source
echo {00..99} > src/{00..99}.txt

So I just rewrited it to

srcfiles := $(shell for i in `seq 0 100`; do echo "src/$$i.txt"; done)```

This comment has been minimized.

joepvd commented Apr 3, 2015

Excellent! In the make target result-using-var.txt, the target special variable needs to be escaped to be displayed, like this:

result-using-var.txt: source.txt
    @echo "buildling result-using-var.txt using the $$< and \$$@ vars"
    cp $< $@

This comment has been minimized.

tehmoth commented Apr 5, 2015

buildling result-using-var.txt using the $< and vars
Using $< in a non-suffix rule context is a GNUmake idiom (line 145 of Makefile)



This comment has been minimized.

nucleartide commented Jul 20, 2015

@nik-kor I got confused by that too. If you're on Ubuntu, it seems like you have to explicitly be using Bash in order for brace expansions to work. See this StackOverflow answer.


This comment has been minimized.

giuseppeg commented Oct 24, 2015

nit: you have a typo at L111 s/becasue/because
<3 for the gist


This comment has been minimized.

BuildMonkey-zz commented Feb 1, 2016

Thanks for this. Are you planning on covering more concepts like conditional branching and functions?


This comment has been minimized.

patil16nit commented Mar 25, 2017

Great article for beginners.
You need a minor change at L113 s/INto/into.


This comment has been minimized.

JiaqiL commented Jan 4, 2018

Great! As @nucleartide said, should consider add SHELL=/bin/bash at the first line.


This comment has been minimized.

nhasbun commented May 14, 2018

Pretty useful guide.
Better than many I found online and everything packed on a single gist.


This comment has been minimized.

hedenface commented May 25, 2018

This is awesome!


This comment has been minimized.

liuyibao commented Aug 30, 2018

Great! Thanks


This comment has been minimized.

codcodog commented Nov 21, 2018

Thanks for this. It's really great.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment