CFLAGS, CCFLAGS, CXXFLAGS - what exactly do these variables control?

GccNaming ConventionsMakefile

Gcc Problem Overview


I am using GNU make to compile my C++ code, and i would like to understand how to make my compilations customizable.

I read in different places that CFLAGS, CCFLAGS and CXXFLAGS are used for this purpose. So how should i use them? If i have additional command-line arguments to the compiler, should i append them to CFLAGS or prepend them? Is there a common practice?

Why the three different variables? I suppose the C compiler should get CFLAGS and CCFLAGS, while the C++ compiler should get CFLAGS and CXXFLAGS - did i get it right?

Is the human user supposed to set these variables at all? Do any automatic tools (automake, autoconf, etc) set them? The linux system that i am supposed to use doesn't define any of these variables - is this typical?

Currently my Makefile looks like this, and i feel it's a bit dirty:

ifdef code_coverage
    GCOV_FLAG := -fprofile-arcs -ftest-coverage
else
    GCOV_FLAG :=
endif

WFLAGS := -Wall

INC_FLAGS := -Istuff -Imore_stuff -Ietc

CCFLAGSINT := -O3 $(WFLAGS) $(INC_FLAGS) $(CCFLAGS)

... (somewhere in the makefile, the command-line for compilation looks like this)
    $(CC) $(CCFLAGSINT) -c $< -o $@

... (somewhere in the makefile, the command-line for linking looks like this)
    $(CC) $(GCOV_FLAG) $(CCFLAGSINT) $(OBJLIST) $(LDFLAGS) -o $@

I am pretty sure there are no bugs here; the Makefile works very well. But is there anything that goes against conventions (like CCFLAGSINT - should i just overwrite CCFLAGS instead? Or CXXFLAGS? FUD!)

Sorry for so many questions; you will obviously not answer them all but i hope the answers will help me understand the general idea behind these settings.

Gcc Solutions


Solution 1 - Gcc

As you noticed, these are Makefile {macros or variables}, not compiler options. They implement a set of conventions. (Macros is an old name for them, still used by some. GNU make doc calls them variables.)

The only reason that the names matter is the default make rules, visible via make -p, which use some of them.

If you write all your own rules, you get to pick all your own macro names.

In a vanilla gnu make, there's no such thing as CCFLAGS. There are CFLAGS, CPPFLAGS, and CXXFLAGS. CFLAGS for the C compiler, CXXFLAGS for C++, and CPPFLAGS for both.

Why is CPPFLAGS in both? Conventionally, it's the home of preprocessor flags (-D, -U) and both c and c++ use them. Now, the assumption that everyone wants the same define environment for c and c++ is perhaps questionable, but traditional.


P.S. As noted by James Moore, some projects use CPPFLAGS for flags to the C++ compiler, not flags to the C preprocessor. The Android NDK, for one huge example.

Solution 2 - Gcc

According to the GNU make manual:

> CFLAGS: Extra flags to give to the C compiler.
> CXXFLAGS: Extra flags to give to the C++ compiler.
> CPPFLAGS: Extra flags to give to the C preprocessor and programs that use it (the C and Fortran compilers).

src: https://www.gnu.org/software/make/manual/make.html#index-CFLAGS
note: PP stands for PreProcessor (and not Plus Plus), i.e. > CPP: Program for running the C preprocessor, with results to standard output; default ‘$(CC) -E’.

These variables are used by the implicit rules of make

> Compiling C programs
n.o is made automatically from n.c with a recipe of the form
‘$(CC) $(CPPFLAGS) $(CFLAGS) -c’.

> Compiling C++ programs
n.o is made automatically from n.cc, n.cpp, or n.C with a recipe of the form
‘$(CXX) $(CPPFLAGS) $(CXXFLAGS) -c’.
We encourage you to use the suffix ‘.cc’ for C++ source files instead of ‘.C’.

src: https://www.gnu.org/software/make/manual/make.html#Catalogue-of-Rules

Solution 3 - Gcc

Minimal example

And just to make what Mizux said as a minimal example:

main_c.c

#include <stdio.h>

int main(void) {
    puts("hello");
}

main_cpp.cpp

#include <iostream>

int main(void) {
    std::cout << "hello" << std::endl;
}

Then, without any Makefile:

make CFLAGS='-g -O3' \
     CXXFLAGS='-ggdb3 -O0' \
     CPPFLAGS='-DX=1 -DY=2' \
     CCFLAGS='--asdf' \
     main_c \
     main_cpp

runs:

cc -g -O3 -DX=1 -DY=2   main_c.c   -o main_c
g++ -ggdb3 -O0 -DX=1 -DY=2   main_cpp.cpp   -o main_cpp

So we understand that:

  • make had implicit rules to make main_c and main_cpp from main_c.c and main_cpp.cpp

  • CFLAGS and CPPFLAGS were used as part of the implicit rule for .c compilation

  • CXXFLAGS and CPPFLAGS were used as part of the implicit rule for .cpp compilation

  • CCFLAGS is not used.

    BTW, the SCons build system for example uses CCFLAGS for flags that are common to C and C++, which is a convention I sometimes follow on my custom make rules.

Those variables are only used in make's implicit rules automatically: if compilation had used our own explicit rules, then we would have to explicitly use those variables as in:

main_c: main_c.c
	$(CC) $(CFLAGS) $(CPPFLAGS) -o $@ $<

main_cpp: main_c.c
	$(CXX) $(CXXFLAGS) $(CPPFLAGS) -o $@ $<

to achieve a similar affect to the implicit rules.

We could also name those variables however we want: but since Make already treats them magically in the implicit rules, those make good name choices.

Tested in Ubuntu 16.04, GNU Make 4.1.

Related: https://stackoverflow.com/questions/2754966/cflags-vs-cppflags

Solution 4 - Gcc

This is an addition to all the existing answers, which mostly describe how these make variables affect the compilation process. In this answer, I describe a suggested way to use these variables, and any surrounding considerations. It is partly inspired by an answer to a similar question.

> How should I use them? Should I append or prepend?

Compilation flags are of two types:

  1. Those which are required to make the compilation work (example: -I, which tells the compiler where to find h-files).
  2. Those which customize the compilation (example: -O, optimization level).

Only type 2 (customization) should go into CCFLAGS, and if the user specifies them, they should override whatever is defined in the Makefile. Use ?= syntax for that in your Makefile:

CCFLAGS ?= -O3

Type 1 settings should go into a new made-up variable, which should aggregate them all, and also include the standard variables, like CCFLAGS:

INC_FLAGS := -Istuff -Imore_stuff -Ietc

CCFLAGS_INTERNAL := $(INC_FLAGS) $(CCFLAGS)

> Is there a common practice?

No, it doesn't seem so. Stuff I described above is one option, which looks good, but it's not standard.

> Is the human user supposed to set these variables at all? The linux system that I use doesn't define any of these variables - is this typical?

Yes, these are for specifying options using the command line. The environment variables usually don't exist.

> Is there anything that goes against conventions? (like CCFLAGSINT - should I just overwrite CCFLAGS instead? Or CXXFLAGS? FUD!)

No, the idea of this Makefile is correct: append CCFLAGS and other flags to the compiler's command-line; don't update CCFLAGS.

Attributions

All content for this solution is sourced from the original question on Stackoverflow.

The content on this page is licensed under the Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license.

Content TypeOriginal AuthorOriginal Content on Stackoverflow
QuestionanatolygView Question on Stackoverflow
Solution 1 - GccbmarguliesView Answer on Stackoverflow
Solution 2 - GccMizuxView Answer on Stackoverflow
Solution 3 - GccCiro Santilli Путлер Капут 六四事View Answer on Stackoverflow
Solution 4 - GccanatolygView Answer on Stackoverflow