How to show 'preprocessed' code ignoring includes with GCC

Question

I'd like to know if it's possible to output 'preprocessed' code wit gcc but 'ignoring' (not expanding) includes:

ES I got this main:

#include <stdio.h>
#define prn(s) printf("this is a macro for printing a string: %s\n", s);

int int(){
char str[5] = "test"; 
prn(str);
return 0;
}

I run gcc -E main -o out.c

I got:

/*
all stdio stuff
*/

int int(){
char str[5] = "test";
printf("this is a macro for printing a string: %s\n", str);
return 0;
}

I'd like to output only:

#include <stdio.h>
int int(){
char str[5] = "test";
printf("this is a macro for printing a string: %s\n", str);
return 0;
}

or, at least, just

int int(){
char str[5] = "test";
printf("this is a macro for printing a string: %s\n", str);
return 0;
}

PS: would be great if possible to expand "local" "" includes and not to expand "global" <> includes

Answer 1

TL;DR

Assign file name to fname and run following commands in shell. Throughout this ansfer fname is assumed to be sh variable containing the source file to be processed.

fname=file_to_process.c ;
grep -G '^#include' <./"$fname" ;
grep -Gv '^#include[ ]*<' <./"$fname" | gcc -x c - -E -o - $(grep -G '^#include[ ]*<' <./"$fname" | xargs -I {} -- expr "{}" : '#include[ ]*<[ ]*\(.*\)[ ]*>' | xargs -I {} printf '-imacros %s ' "{}" ) | grep -Ev '^([ ]*|#.*)$'

All except gcc here is pure POSIX sh, no bashisms, or nonportable options. First grep is there to output #include directives.

GCC's -imacros

From gcc documentation:

-imacros file: Exactly like ‘-include’, except that any output produced by scanning file is thrown away. Macros it defines remain defined. This allows you to acquire all the macros from a header without also processing its declarations

So, what is -include anyway?

-include file: Process file as if #include "file" appeared as the first line of the primary source file. However, the first directory searched for file is the preprocessor’s working directory instead of the directory containing the main source file. If not found there, it is searched for in the remainder of the #include "..." search chain as normal.

Simply speaking, because you cannot use <> or "" in -include directive, it will always behave as if #include <file> were in source code.

First approach

ANSI C guarantees assert to be macro, so it is perfect for simple test: printf 'int main(){\nassert(1);\nreturn 0;}\n' | gcc -x c -E - -imacros assert.h. Options -x c and - tells gcc to read source file from stdin and that the language used is C. Output doesn't contain any declarations from assert.h, but there is still mess, that can be cleaned up with grep:

printf 'int main(){\nassert(1);\nreturn 0;}\n' | gcc -x c -E - -imacros assert.h | grep -Ev '^([ ]*|#.*)$'

Note: in general, gcc won't expand tokens that intended to be macros, but the definition is missing. Nevertheless assert happens to expand entirely: __extension__ is compiler option, __assert_fail is function, and __PRETTY_FUNCTION__ is string literal.

Automatisation

Previous approach works, but it can be tedious;

1. each #include needs to be deleted from file manually, and

2. it has to be added to gcc call as -imacros's argument.

First part is easy to script: pipe grep -Gv '^#include[ ]*<' <./"$fname" to gcc.

Second part takes some exercising (at least without awk):

2.1 Drop -v negative matching from previous grep command: grep -G '^#include[ ]*<' <./"$fname"

2.2 Pipe previous to expr inside xarg to extract header name from each include directive: xargs -I {} -- expr "{}" : '#include[ ]*<[ ]*\(.*\)[ ]*>'

2.3 Pipe again to xarg, and printf with -imacros prefix: xargs -I {} printf '-imacros %s ' "{}" 2.4 Enclose all in command substitution "$()" and place inside gcc.

Done. This is how you end up with the lengthy command from the beginning of my answer.

Solving subtle problems

This solution still has flaws; if local header files themselves contains global ones, these global will be expanded. One way to solve this problem is to use grep+sed to transfer all global includes from local files and collect them in each *.c file.

printf '' > std ;
for header in *.h ; do
    grep -G '^#include[ ]*<' <./$header >> std ;
    sed -i '/#include[ ]*</d' $header ;
done;
for source in *.c ; do
    cat std > tmp;
    cat $source >> tmp;
    mv -f tmp $source ;
done

Now the processing script can be called on any *.c file inside pwd without worry, that anything from global includes would leak into. The final problem is duplication. Local headers including themselves local includes might be duplicated, but this could occur only, when headers aren't guarded, and in general every header should be always guarded.

---

Final version and example

To show these scripts in action, here is small demo:

File h1.h:

#ifndef H1H
#define H1H
#include <stdio.h>
#include <limits.h>
#define H1 printf("H1:%i\n", h1_int)
int h1_int=INT_MAX;
#endif

File h2.h:

#ifndef H2H
#define H2H
#include <stdio.h>
#include "h1.h"
#define H2 printf("H2:%i\n", h2_int)
int h2_int;
#endif

File main.c:

#include <assert.h>
#include "h1.h"
#include "h2.h"
int main(){
  assert(1);
  H1;
  H2;
}

Final version of the script preproc.sh:

fname="$1"

printf '' > std ;
for source in *.[ch] ; do
    grep -G '^#include[ ]*<' <./$source >> std ;
    sed -i '/#include[ ]*</d' $source ;
    sort -u std > std2;
    mv -f std2 std;
done;
for source in *.c ; do
    cat std > tmp;
    cat $source >> tmp;
    mv -f tmp $source ;
done

grep -G '^#include[ ]*<' <./"$fname" ;

grep -Gv '^#include[ ]*<' <./"$fname" | gcc -x c - -E -o - $(grep -G '^#include[ ]*<' <./"$fname" | xargs -I {} -- expr "{}" : '#include[ ]*<[ ]*\(.*\)[ ]*>' | xargs -I {} printf '-imacros %s ' "{}" ) | grep -Ev '^([ ]*|#.*)$'

Output of the call ./preproc.sh main.c:

#include <assert.h>
#include <limits.h>
#include <stdio.h>
int h1_int=0x7fffffff;
int h2_int;
int main(){
 ((void) sizeof ((
 1
 ) ? 1 : 0), __extension__ ({ if (
 1
 ) ; else __assert_fail (
 "1"
 , "<stdin>", 4, __extension__ __PRETTY_FUNCTION__); }))
          ;
  printf("H1:%i\n", h1_int);
  printf("H2:%i\n", h2_int);
}

This should always compile. If you really want to print every #include "file", then delete < from grep pattern '^#include[ ]*<' in 16-th line of preproc.sh`, but be warned, that content of headers will then be duplicated, and code might fail, if headers contain initialisation of variables. This is purposefully the case in my example to address the problem.

Summary

There are plenty of good answers here so why yet another? Because this seems to be unique solution with following properties:

1. Local includes are expanded
2. Global included are discarded
3. Macros defined either in local or global includes are expanded

Approach is general enough to be usable not only with toy examples, but actually in small and medium projects that reside in a single directory.

Answer 2

Many previous answers went in the direction of using the tracing # directives.

It's actually a one-liner in classical Unix (with awk):

gcc -E file.c | awk '/# [1-9][0-9]* "file.c"/ {skip=0; next} /# [1-9][0-9]* ".*"/ {skip=1} (skip<1) {print}'

Answer 3

supposing the file is named c.c :

gcc -E c.c | tail -n +`gcc -E c.c | grep -n -e "#*\"c.c\""  | tail -1 | awk -F: '{print $1}'`

It seems # <number> "c.c" marks the lines after each #include

Of course you can also save gcc -E c.c in a file to not do it two times

The advantage is to not modify the source nor to remove the #include before to do the gcc -E, that just removes all the lines from the top up to the last produced by an #include ... if I am right

Answer 4

You can use -dI to show the #include directives and post-process the preprocessor output.

Assuming the name of your your file is foo.c

SOURCEFILE=foo.c
gcc -E -dI "$SOURCEFILE" | awk '
    /^# [0-9]* "/ { if ($3 == "\"'"$SOURCEFILE"'\"") show=1; else show=0; }
    { if(show) print; }'

or to suppress all # line_number "file" lines for $SOURCEFILE:

SOURCEFILE=foo.c
gcc -E -dI "$SOURCEFILE" | awk '
    /^# [0-9]* "/ { ignore = 1; if ($3 == "\"'"$SOURCEFILE"'\"") show=1; else show=0; }
    { if(ignore) ignore=0; else if(show) print; }'

Note: The AWK scripts do not work for file names that include whitespace. To handle file names with spaces you could modify the AWK script to compare $0 instead of $3.

Answer 5

When cpp expands includes it adds # directives (linemarkers) to trace back errors to the original files.

You can add a post processing step (it can be trivially written in any scripting language, or even in C if you feel like it) to parse just the linemarkers and filter out the lines coming from files outside of your project directory; even better, one of the flags (3) marks system header files (stuff coming from paths provided through -isystem, either implicitly by the compiler driver or explicitly), so that's something you could exploit as well.

For example in Python 3:

#!/usr/bin/env python3
import sys

skip = False
for l in sys.stdin:
    if not skip:
        sys.stdout.write(l)
    if l.startswith("# "):
        toks = l.strip().split(" ")
        linenum, filename = toks[1:3]
        flags = toks[3:]
        skip = "3" in flags

Using gcc -E foo.c | ./filter.py I get

# 1 "foo.c"
# 1 "<built-in>"
# 1 "<command-line>"
# 31 "<command-line>"
# 1 "/usr/include/stdc-predef.h" 1 3 4
# 1 "foo.c"
# 1 "/usr/include/stdio.h" 1 3 4



# 4 "foo.c"
int int(){
char str[5] = "test";
printf("this is a macro for printing a string: %s\n", str);;
return 0;
}

Answer 6

Protect the #includes from getting expanded, run the preprocessor textually, remove the # 1 "<stdint>" etc. junk the textual preprocessor generates and reexpose the protected #includes.

This shell function does it:

expand_cpp(){
     sed 's|^\([ \t]*#[ \t]*include\)|magic_fjdsa9f8j932j9\1|' "$@" \
     | cpp | sed 's|^magic_fjdsa9f8j932j9||; /^# [0-9]/d'
}

as long as you keep the include word together instead of doing crazy stuff like

#i\
ncl\
u??/
de <iostream>

(above you can see 2 backslash continuation lines + 1 trigraph (??/ == \ ) backslash continuation line).

If you wish, you can protect #ifs #ifdefs #ifndefs #endifs and #elses the same way.

Applied to your example

example.c:

#include <stdio.h>
#define prn(s) printf("this is a macro for printing a string: %s\n", s);

int int(){
char str[5] = "test";
prn(str);
return 0;
}

like as with expand_cpp < example.c or expand_cpp example.c, it generates:

#include <stdio.h>


int int(){
char str[5] = "test";
printf("this is a macro for printing a string: %s\n", str);;
return 0;
}

Answer 7

I agree with Matteo Italia's comment that if you just prevent the #include directives from being expanded, then the resulting code won't represent what the compiler actually sees, and therefore it will be of limited use in troubleshooting.

Here's an idea to get around that. Add a variable declaration before and after your includes. Any variable that is reasonably unique will do.

int begin_includes_tag;
#include <stdio.h>
... other includes
int end_includes_tag;

Then you can do:

> gcc -E main -o out.c | sed '/begin_includes_tag/,/end_includes_tag/d'

The sed command will delete everything between those variable declarations.