spot/tests/sanity/style.test

#! /bin/sh
# -*- coding: utf-8 -*-
# Copyright (C) by the Spot authors, see the AUTHORS file for details.
#
# This file is part of Spot, a model checking library.
#
# Spot is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 3 of the License, or
# (at your option) any later version.
#
# Spot is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
# or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public
# License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.

# Ensure consistent style by catching common improper constructs.

set -e
set +x

diag()
{
  fail=:
  echo "$file:" "$@"
  echo ============================================================
}

rm -f failures.style

GREP=grep

# Get some help from GNU grep.
if (grep --color=auto -n --version)>/dev/null 2>&1; then
  GREP="$GREP --color=auto -n"
  GREP_COLOR='mt=1;31'
  export GREP_COLOR
fi

# Reset the locale, so for instance we don't get bugs because sed is
# expecting utf-8 and we feed it latin-1.  The C locale should be OK,
# because we do not treat extended characters specifically in the
# following style rules.
LC_ALL=C
export LC_ALL

mkdir -p style.dir
tmp=style.dir/incltest.tmp.$$

# We used to loop over more directories before the source tree was
# rearranged.  So there is only one left today, but we keep the loop
# in case we want to add more in the future.
TOP=${srcdir-.}/../..
for dir in "$TOP/spot" "$TOP/bin" "$TOP/tests"; do

  find "$dir" \(    -name "${1-*}.hh" \
                 -o -name "${1-*}.hxx" \
                 -o -name "${1-*}.cc" \
                 -o -name "${1-*}.test" \) \
                 -a -not -path '*.dir/*' \
                 -a -type f -a -print |
  while read file; do
    if $GREP 'GNU Bison' "$file" >/dev/null ||
      $GREP '/robin-hood-hashing' "$file" >/dev/null ||
      $GREP 'generated by flex' "$file" >/dev/null ; then
      continue
    fi

    # Skip the files used by sanity.
    case $file in
	*incltest.cc) continue;;
    esac

    fail=false

    # Check this before stripping comments and strings.
    $GREP -i 'accepting cond' $file && diag 'accepting -> acceptance'
    $GREP -i 'dictionnar[yi]' $file && diag 'dictionnary -> dictionary'
    # "an uninstalled" seems to be the exception so far, but we want
    # "a unique", "a universal", etc.
    $GREP -i 'an uni[^n]' $file && diag 'an uni... -> a uni...'

    $GREP -i 'version 2 of the License' $file &&
      diag 'license text should refer to version 2'
    $GREP -i 'Temple Place' $file &&
      diag 'license text should give a url instead of an address'

    $GREP -q 'coding: utf-8' $file ||
      diag 'missing -*- coding: utf-8 -*-'

    $GREP Copyright $file >/dev/null ||
      diag "missing copyright"

    # If some grep implementation ignores LC_ALL=C, this rule might be
    # a problem on utf-8 characters such as "δ" which really
    # corresponds to multiple bytes, but might be matched as a single
    # character by grep.
    $GREP '<< *"\([^\\]\|\\.\)"' $file &&
      diag "Use << 'c' instead" 'of << "c".'

    # A doxygen comments such as
    #
    # | \brief foo
    # | \ingroup bar
    # |
    # | baz
    #
    # will be output as "foobaz."  But if the first two lines
    # are reversed, it's output correctly.
    perl -ne '/(.*\\brief.*\n.*\\ingroup.*)/ && print("$1\n") && exit(1)' \
      -0777 $file || diag "always put 'ingroup' before 'brief'"

    # Check this before we remove SPOT_API from the input.
    case $file in
	*.cc)
	    $GREP 'SPOT_API' $file &&
		diag 'use SPOT_API only in header files';;
    esac

    # Strip comments and strings.
    #
    # Multi-line comments of the form
    # /* Line 1
    #    Line 2
    #    Line 3 */
    # are replaced by
    # //
    # //
    # //
    # to keep the line numbers correct in the diagnostics.
    #
    # Also get rid of the SPOT_API tags.
    perl -pe 'sub f {my $a = shift; $a =~ s:[^\n]*://:g; return "$a"}
              s,/\*(.*?)\*/,f($1),sge;
              s,//.*?\n,//\n,g;
              s,(?<!include )"(\\.|[^"\\\n])*","",g;
              s,SPOT_API ,,g' -0777 <$file >$tmp

    $GREP '[	 ]$' $tmp &&
      diag 'Trailing whitespace.'

    case $file in
    *.test);;
    *)
      $GREP -E '(>[^>]*|^[^<]*)class[ \t]+[A-Z]' $tmp &&
        diag 'Use lower case class names.'

      $GREP '[	 ]if(' $tmp &&
	diag 'Missing space after "if"'

      $GREP '[	 ]if ([^()]*).*{' $tmp &&
	diag 'Opening { should be on its own line.'

      $GREP '[	 ]if ([^()]*).*;' $tmp &&
	diag 'if body should be on another line.'

      $GREP '[	 ]if ([^()]*([^()]*)[^()]*).*;' $tmp &&
	diag 'if body should be on another line.'

      $GREP -E '[	 ]else.*;(|.*}.*)$' $tmp &&
	diag 'else body should be on another line.'

      $GREP '[	 ]while(' $tmp &&
	diag 'Missing space after "while"'

      $GREP '[	 ]while (.*).*{' $tmp &&
	diag 'Opening { should be on its own line.'

      $GREP '[	 ]while (.*).*[^)];' $tmp &&
	diag 'while body should be on another line.'

      $GREP '[	 ]for(' $tmp &&
	diag 'Missing space after "for"'

      $GREP '[	 ]for (.*).*{' $tmp &&
	diag 'Opening { should be on its own line.'

      $GREP '[	 ]for (.*;.*;.*).*;' $tmp &&
	diag 'for body should be on another line.'

      $GREP '[	 ]switch(' $tmp &&
	diag 'Missing space after "switch"'

      $GREP '[	 ]switch (.*).*{' $tmp &&
	diag 'Opening { should be on its own line.'

      $GREP 'namespace .*{' $tmp &&
	diag 'Opening { should be on its own line.'

      $GREP 'class .*{' $tmp | $GREP -v enum &&
	diag 'Opening { should be on its own line.'

      $GREP '( ' $tmp &&
	diag 'No space after opening (.'

      $GREP ' )' $tmp &&
	diag 'No space before closing ).'

      $GREP '! ' $tmp &&
	diag 'No space after unary operators (!).'

      $GREP ",[^ \"	%'\\\\]" $tmp &&
	diag 'Space after comma.'

      # The 'r' allows operator&&
      # The '.' allows &&...
      # The ')' allows foo(bar&&)
      $GREP '[^	 r]&&[^	 .)]' $tmp &&
	diag 'Space around binary operators.'

      # The 'r' allows operator||
      $GREP '[^	 r]||[^	 ]' $tmp &&
	diag 'Space around binary operators.'

      # The 'r' allows operator==
      $GREP '[^	 r<>][!<>=]=[^	 ]' $tmp &&
	diag 'Space around binary operators.'

      # The 'r' allows operator<<=
      $GREP '[^	 r][<>][<>]=[^	 ]' $tmp &&
	diag 'Space around binary operators.'

      $GREP 'operator[^a-zA-Z0-9_(]*[	 ][^a-zA-Z0-9_(]*(' $tmp &&
        diag 'Write operatorXX(...) without spaces around XX.'

      $GREP 'operator[^(]* (' $tmp &&
        diag 'No space before ('

      $GREP '[	 ]default:[^:].*;' $tmp &&
	diag 'Label should be on their own line.'

      $GREP '[	 ]case.*:[^:].*;' $tmp &&
	diag 'Label should be on their own line.'

      $GREP '[	 ];' $tmp &&
	diag 'No space before semicolon.'

      $GREP -v 'for (.*;;)' $tmp | $GREP ';[^	 ")'"']" &&
	diag 'Must have space or newline after semicolon.'

      # Allow several { or } on the same line only if they are mixed
      # with parentheses, as this often occur with lambdas or
      # initializer lists.  What we want to forbid is cases where
      # multiple scopes are opened/closed on the same line.
      $GREP '^[^():]*}[^()]*}[^()]*$' $tmp &&
	diag 'No two } on the same line.'

      $GREP '^[^()]{[^()]*{[^()]$' $tmp &&
	diag 'No two { on the same line.'

      $GREP 'delete[	 ]*[(][^(]*[)];' $tmp |
      $GREP -v 'operator[	 ]*delete[	 ]*[(][^(]*[)];' &&
	diag 'No useless parentheses after delete.'

      $GREP 'return[	 ]*[(][^(]*[)];' $tmp &&
	diag 'No useless parentheses after return.'

      $GREP 'NULL' $tmp &&
	diag 'Use nullptr instead of NULL.'

      # std::list::size() can be O(n).  Better use empty() whenever
      # possible, even for other containers.
      $GREP -E '(->|[.])size\(\) [=!]= 0|![a-zA-Z0-9_]*(->|[.])size\(\)|(if |while |assert)\([a-zA-Z0-9_]*(->|[.])size\(\)\)' $tmp &&
	diag 'Prefer empty() to check emptiness.'

      $GREP -E 'std::cerr.*<<.*endl' $tmp &&
        diag 'std::cerr has unitbuf set; use \\n instead of endl'

      $GREP -E 'assert\((0|!".*")\)' $tmp &&
        diag 'Prefer SPOT_UNREACHABLE or SPOT_UNIMPLEMENTED.'

      $GREP -E '^[^=*<]*([+][+]|--);' $tmp &&
	diag 'Take good habits: use ++i instead of i++ when you have the choice.'

      $GREP '[^a-zA-Z0-9_](\*[a-zA-Z0-9_]*)\.' $tmp &&
	diag 'Use "x->y", not "(*x).y"'

      # we allow these functions only in ?...:...
      $GREP -E 'bdd_(false|true)[	]*\(' $tmp | $GREP -v '[?:]' &&
	diag 'Use bddfalse and bddtrue instead of bdd_false() and bdd_true()'

      res=`perl -ne '$/ = undef;
	print "$&\n"
	while /if \((.*)(\s*==\s*0)?\)\s*delete(\[\])?\s+\1;(?!\s+else)/g' $tmp`
      if test -n "$res"; then
	 echo "$res"
	 diag 'No "if (x)" required before "delete x;".'
      fi

      # we want catch (const reftype&) or catch (...)
      $GREP 'catch *([^.]' $tmp | $GREP -v 'const.*&' &&
	diag 'Always capture exceptions by const reference.'

      # iterating over edge_vector() is suspicious, because the
      # first edge should not be used.
      $GREP 'for (.*:.*edge_vector()' $tmp &&
	diag 'Did you mean to iterate over edges()?'

      case $file in
	*.hh | *.hxx)
	  if $GREP -E '(<<|>>)' $tmp >/dev/null; then
	    :
	  else
	    $GREP '#.*include.*<iostream>' $tmp &&
	      diag 'Avoid <iostream> in headers, better use <iosfwd>.'
	  fi
          $GREP '^[ ]*	' $tmp &&
              diag 'Use spaces instead of tabs.'
	  # Headers from spot/priv/ are not installed, so may only be
	  # included from *.cc files or from other spot/priv/ headers
	  # (in the latter case they do not have to specify the priv/
	  # directory, so they will not match this regex).
	  case $file in
	    */priv/*|*/bin/*|*/parsedecl.hh);;
	    *)
	      $GREP '#.*include.*priv/' $tmp &&
		diag 'Do not include private headers in public headers.'
              $GREP '^[^#]*[^#_]assert[	 ]*(.*)' $tmp &&
                diag 'Use SPOT_ASSERT() instead of assert() in public headers.'
	      ;;
	  esac
          case $file in
              */bin/*);;
              *.hh)
                  $GREP '#.*include.*[<"]config.h[">]' $tmp &&
                      diag 'Do not include config.h in header files.'
          esac
	  $GREP '[<"]assert.h[">]' $tmp &&
              diag 'include <cassert> in C++ files'
	  ;;
	*.cc)
	  if $GREP 'namespace$' $tmp >/dev/null; then
	    :
	  else
	    # We only check classes, but the rule should apply to functions too
	    $GREP '^[	 ]*class[	 ]' $tmp &&
	      diag 'Private definitions must be in anonymous namespace.'
	  fi
          $GREP '	' $tmp &&
	      diag 'Use spaces instead of tabs.'
          case $file in
              */bin/*) ;;
              *)
                  if $GREP -q '#.*include.*"config.h"' $tmp; then
                      :
                  else
                      diag "*.cc files should include "'"config.h"'
                  fi;;
          esac
          $GREP '[<"]assert.h[">]' $tmp &&
              diag 'include <cassert> in C++ files'
          ;;
      esac
      case $file in
	*/bin/*|*.cc);;
	*)
	  $GREP -v '#.*include.*priv/' $tmp | $GREP '#.*include.*"' &&
	      diag 'Use <header> instead of "header" for public headers.'
	  $GREP '#.*include.*<spot/priv/.*>' $tmp &&
	      diag 'Use "spot/priv/..." instead of <spot/priv/...>".'
	  ;;
      esac
      ;;
    esac


    $fail && echo "$file" >>failures.style

  done || : # Make sure sh does not abort when read exits with false.
done


# Rules for Makefiles.
for dir in "${INCDIR-..}" "${INCDIR-..}/../bin" "${INCDIR-..}/../tests"; do

  find "$dir" -name "Makefile.am" -a -type f -a -print |
  while read file; do
    fail=false

    # Strip comments.
    sed 's,#.*,,' < $file > $tmp

    $GREP '[	 ]$' $tmp &&
      diag 'Trailing whitespace.'

    $GREP '\.libs/' $tmp &&
	diag "Don't reference files in .libs/, use Libtool instead."

    $fail && echo "$file" >>failures.style
  done || : # Make sure sh does not abort when read exits with false.
done

# Rules for Python tests
for dir in "${INCDIR-..}/../tests"; do

  find "$dir" -name "*.py" -a -type f -a -print |
  while read file; do
    fail=false

    # Strip comments.
    sed 's,[	 ]*#.*,,' < $file > $tmp

    $GREP '[	 ]$' $tmp &&
      diag 'Trailing whitespace.'

    $GREP -E '([	 ]|^)assert[ (]' $tmp &&
	diag "replace assert keywords by unittest assertion tests"

    $fail && echo "$file" >>failures.style
  done || : # Make sure sh does not abort when read exits with false.
done


if test -f failures.style; then
   echo "The following files contain style errors:"
   cat failures.style
   rm failures.style
   exit 1;
fi

exit 0