diff options
Diffstat (limited to 'lib/talloc/script/mksigs.pl')
-rwxr-xr-x | lib/talloc/script/mksigs.pl | 199 |
1 files changed, 0 insertions, 199 deletions
diff --git a/lib/talloc/script/mksigs.pl b/lib/talloc/script/mksigs.pl deleted file mode 100755 index dfe36bc138..0000000000 --- a/lib/talloc/script/mksigs.pl +++ /dev/null @@ -1,199 +0,0 @@ -#!/usr/bin/perl - -# mksigs.pl - extract signatures from C headers -# -# Copyright (C) Michael Adam 2009 -# -# This program 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. -# -# This program 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/>. - -# USAGE: cat $header_files | mksigs.pl > $signature_file -# -# The header files to parse are read from stdin. -# The output is in a form as produced by gcc with the -aux-info switch -# and printed to stdout. - -use strict; -use warnings; - -my $in_comment = 0; -my $in_doxygen = 0; -my $extern_C_block = 0; - -while (my $LINE = <>) { - # find end of started multi-line-comment - if ($in_comment) { - if ($LINE =~ /^.*?\*\/(.*)$/) { - $LINE = $1; - $in_comment = 0; - } else { - # whole line within comment - next; - } - } - - # find end of DOXYGEN section - if ($in_doxygen) { - if ($LINE =~ /^#\s*else(?:\s+.*)?$/) { - $in_doxygen = 0; - } - next; - } - - # strip C++-style comments - $LINE =~ s/^(.*?)\/\/.*$/$1/; - - # strip in-line-comments: - while ($LINE =~ /\/\*.*?\*\//) { - $LINE =~ s/\/\*.*?\*\///; - } - - # find starts of multi-line-comments - if ($LINE =~ /^(.*)\/\*/) { - $in_comment = 1; - $LINE = $1; - } - - # skip empty lines - next if $LINE =~ /^\s*$/; - - # remove leading spaces - $LINE =~ s/^\s*(.*)$/$1/; - - # concatenate lines split with "\" (usually macro defines) - while ($LINE =~ /^(.*?)\s+\\$/) { - my $LINE2 = <>; - $LINE = $1; - $LINE2 =~ s/^\s*(.*)$/$1/; - $LINE .= " " . $LINE2; - } - - # remove DOXYGEN sections - if ($LINE =~ /^#\s*ifdef\s+DOXYGEN(?:\s+.*)?$/) { - $in_doxygen = 1; - next; - } - - - # remove all preprocessor directives - next if ($LINE =~ /^#/); - - if ($LINE =~ /^extern\s+"C"\s+\{/) { - $extern_C_block = 1; - next; - } - - if (($LINE =~ /^[^\{]*\}/) and $extern_C_block) { - $extern_C_block = 0; - next; - } - - $LINE =~ s/^extern\s//; - - # concatenate braces stretched over multiple lines - # (from structs or enums) - my $REST = $LINE; - my $braces = 0; - while (($REST =~ /[\{\}]/) or ($braces)) { - while ($REST =~ /[\{\}]/) { - # collect opening - while ($REST =~ /^[^\{\}]*\{(.*)$/) { - $braces++; - $REST = $1; - } - - # collect closing - while ($REST =~ /^[^\{\}]*\}(.*)$/) { - $braces--; - $REST = $1; - } - } - - # concatenate if not balanced - if ($braces) { - if (my $LINE2 = <>) { - $LINE2 =~ s/^\s*(.*)$/$1/; - chomp($LINE); - $LINE .= " " . $LINE2; - chomp $REST; - $REST .= " " . $LINE2; - } else { - print "ERROR: unbalanced braces ($braces)\n"; - last; - } - } - } - - # concetenate function prototypes that stretch over multiple lines - $REST = $LINE; - my $parenthesis = 0; - while (($REST =~ /[\(\)]/) or ($parenthesis)) { - while ($REST =~ /[\(\)]/) { - # collect opening - while ($REST =~ /^[^\(\)]*\((.*)$/) { - $parenthesis++; - $REST = $1; - } - - # collect closing - while ($REST =~ /^[^\(\)]*\)(.*)$/) { - $parenthesis--; - $REST = $1; - } - } - - # concatenate if not balanced - if ($parenthesis) { - if (my $LINE2 = <>) { - $LINE2 =~ s/^\s*(.*)$/$1/; - chomp($LINE); - $LINE .= " " . $LINE2; - chomp($REST); - $REST .= " " . $LINE2; - } else { - print "ERROR: unbalanced parantheses ($parenthesis)\n"; - last; - } - } - } - - next if ($LINE =~ /^typedef\s/); - next if ($LINE =~ /^enum\s+[^\{\(]+\s+\{/); - next if ($LINE =~ /^struct\s+[^\{\(]+\s+\{.*\}\s*;/); - next if ($LINE =~ /^struct\s+[a-zA-Z0-9_]+\s*;/); - - # remove trailing spaces - $LINE =~ s/(.*?)\s*$/$1/; - - $LINE =~ s/^(.*\))\s+PRINTF_ATTRIBUTE\([^\)]*\)(\s*[;,])/$1$2/; - $LINE =~ s/^(.*\))\s*[a-zA-Z0-9_]+\s*;$/$1;/; - - # remove parameter names - slightly too coarse probably - $LINE =~ s/([\s\(]\*?)[_0-9a-zA-Z]+\s*([,\)])/$1$2/g; - - # remedy (void) from last line - $LINE =~ s/\(\)/(void)/g; - - # normalize spaces - $LINE =~ s/\s*\)\s*/)/g; - $LINE =~ s/\s*\(\s*/ (/g; - $LINE =~ s/\s*,\s*/, /g; - - # normalize unsigned - $LINE =~ s/([\s,\(])unsigned([,\)])/$1unsigned int$2/g; - - # normalize bool - $LINE =~ s/(\b)bool(\b)/_Bool/g; - - print $LINE . "\n"; -} |