[RFC] duplicate #include check for build system

[Herbert Poetzl <herbert@13thfloor.at>]

Hi Sam! Folks!

recently had the idea to utilize cpp or sparse to
do some automated #include checking, and I came up
with the following proof of concept:

I just replaced the sparse binary by the following
script (basically hijacking the make C=1 system)

it would allow kernel developers to easily identify
duplicate includes, which in turn, might reduce
dependancies and thus build time ...


----------------
#!/bin/bash

while [ $# -gt 1 ]; do
  case $1 in
        -D*) DEF="$DEF $1";;
        -W*) ;;
          *) OPT="$OPT $1";;
  esac
  shift
done

# ( $CPP $DEF -H -dI $OPT $1 1>&2 ) 2>&1 | grep "^[#.]"
$CPP $DEF -H  $OPT $1 2>&1 >/dev/null | gawk -vFILE="$1" '

BEGIN	{ I[0]=FILE; }

/^[.]+/ { D=length($1);

	  for (i=0; i<D; i++)
    	    C[i,$2]++; 
	    
	  I[D]=$2;

	  for (i=0; i<D; i++)
	    M[i,$2]=I[i];

	  if (C[D-1,$2]>1) {
	    printf "··· %s in %s ",$2,I[D-1];

	    for (i=D; M[i,$2]; i++) 
	      printf "%c%s", (i==D)?"[":"·", M[i,$2];

	    printf (i>D) ? "]\n" :
	    	((X[D,$2]==I[D-1]) ? "(dup)\n" : "\n");
	  }

          X[D,$2]=I[D-1];
        }
' 1>&2

true
----------------

of course, most of it would not be required if
there was support in the kernel build system,
and, if there is any preference for perl over
bash/gawk it could be easily rewritten ...

please let me know what you think of this and if
you could imagine adding something similar to the
build system 

TIA,
Herbert