How do I write the 'cd' command in a makefile?

For example, I have something like this in my makefile:

all:
     cd some_directory

But when I typed make I saw only 'cd some_directory', like in the echo command.

It is actually executing the command, changing the directory to some_directory, however, this is performed in a sub-process shell, and affects neither make nor the shell you're working from.

If you're looking to perform more tasks within some_directory, you need to add a semi-colon and append the other commands as well. Note that you cannot use new lines as they are interpreted by make as the end of the rule, so any new lines you use for clarity need to be escaped by a backslash.

For example:

all:
        cd some_dir; echo "I'm in some_dir"; \
          gcc -Wall -o myTest myTest.c

Note also that the semicolon is necessary between every command even though you add a backslash and a newline. This is due to the fact that the entire string is parsed as a single line by the shell. As noted in the comments, you should use '&&' to join commands, which means they only get executed if the preceding command was successful.

all:
        cd some_dir && echo "I'm in some_dir" && \
          gcc -Wall -o myTest myTest.c

This is especially crucial when doing destructive work, such as clean-up, as you'll otherwise destroy the wrong stuff, should the cd fail for whatever reason.

A common usage, though, is to call make in the subdirectory, which you might want to look into. There's a command-line option for this, so you don't have to call cd yourself, so your rule would look like this

all:
        $(MAKE) -C some_dir all

which will change into some_dir and execute the Makefile in that directory, with the target "all". As a best practice, use $(MAKE) instead of calling make directly, as it'll take care to call the right make instance (if you, for example, use a special make version for your build environment), as well as provide slightly different behavior when running using certain switches, such as -t.

For the record, make always echos the command it executes (unless explicitly suppressed), even if it has no output, which is what you're seeing.

Starting from GNU make 3.82 (July 2010), you can use the .ONESHELL special target to run all recipes in a single instantiation of the shell (bold emphasis mine):

> * New special target: .ONESHELL instructs make to invoke a single instance of the shell and provide it with the entire recipe, regardless of how many lines it contains.

.ONESHELL: # Applies to every targets in the file!

all:
    cd ~/some_dir
    pwd # Prints ~/some_dir if cd succeeded

another_rule:
    cd ~/some_dir
    pwd # Prints ~/some_dir if cd succeeded

Note that this will be equivalent to manually running

$(SHELL) $(.SHELLFLAGS) "cd ~/some_dir; pwd"
# Which gets replaced to this, most of the time:
/bin/sh -c "cd ~/some_dir; pwd"

Commands are not linked with && so if you want to stop at the first one that fails, you should also add the -e flag to your .SHELLFLAGS:

.SHELLFLAGS += -e

Also the -o pipefail flag might be of interest:

> If set, the return value of a pipeline is the value of the last (rightmost) command to exit with a non-zero status, or zero if all commands in the pipeline exit successfully. This option is disabled by default.

Here's a cute trick to deal with directories and make. Instead of using multiline strings, or "cd ;" on each command, define a simple chdir function as so:

CHDIR_SHELL := $(SHELL)
define chdir
   $(eval _D=$(firstword $(1) $(@D)))
   $(info $(MAKE): cd $(_D)) $(eval SHELL = cd $(_D); $(CHDIR_SHELL))
endef

Then all you have to do is call it in your rule as so:

all:
          $(call chdir,some_dir)
          echo "I'm now always in some_dir"
          gcc -Wall -o myTest myTest.c

You can even do the following:

some_dir/myTest:
          $(call chdir)
          echo "I'm now always in some_dir"
          gcc -Wall -o myTest myTest.c

What do you want it to do once it gets there? Each command is executed in a subshell, so the subshell changes directory, but the end result is that the next command is still in the current directory.

With GNU make, you can do something like:

BIN=/bin
foo:
    $(shell cd $(BIN); ls)

To change dir

foo: 
    $(MAKE) -C mydir

multi:
    $(MAKE) -C / -C my-custom-dir   ## Equivalent to /my-custom-dir

Here is the pattern I've used:

.PHONY: test_py_utils
PY_UTILS_DIR = py_utils
test_py_utils:
	cd $(PY_UTILS_DIR) && black .
	cd $(PY_UTILS_DIR) && isort .
	cd $(PY_UTILS_DIR) && mypy .
	cd $(PY_UTILS_DIR) && pytest -sl .
	cd $(PY_UTILS_DIR) && flake8 .

My motivations for this pattern are:

• The above solution is simple and readable (albeit verbose)

• I read the classic paper "Recursive Make Considered Harmful", which discouraged me from using $(MAKE) -C some_dir all

• I didn't want to use just one line of code (punctuated by semicolons or &&) because it is less readable, and I fear that I will make a typo when editing the make recipe.

• I didn't want to use the .ONESHELL special target because:

  • that is a global option that affects all recipes in the makefile
  • using .ONESHELL causes all lines of the recipe to be executed even if one of the earlier lines has failed with a nonzero exit status. Workarounds like calling set -e are possible, but such workarounds would have to be implemented for every recipe in the makefile.

Like this:

target:
    $(shell cd ....); \
    # ... commands execution in this directory
    # ... no need to go back (using "cd -" or so)
    # ... next target will be automatically in prev dir

Good luck!