A perl module to parse through C headers
authorRichard Levitte <levitte@openssl.org>
Wed, 3 Oct 2018 15:43:48 +0000 (17:43 +0200)
committerRichard Levitte <levitte@openssl.org>
Thu, 4 Oct 2018 10:45:40 +0000 (12:45 +0200)
OpenSSL::ParseC is a module that parses through a C header file and
returns a list with information on what it found.  Currently, the
information it returns covers function and variable declarations,
macro definitions, struct declarations/definitions and typedef
definitions.

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/7191)

util/perl/OpenSSL/ParseC.pm [new file with mode: 0644]

diff --git a/util/perl/OpenSSL/ParseC.pm b/util/perl/OpenSSL/ParseC.pm
new file mode 100644 (file)
index 0000000..ba2427c
--- /dev/null
@@ -0,0 +1,1129 @@
+#! /usr/bin/env perl
+# Copyright 2018 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License").  You may not use
+# this file except in compliance with the License.  You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+package OpenSSL::ParseC;
+
+use strict;
+use warnings;
+
+use Exporter;
+use vars qw($VERSION @ISA @EXPORT @EXPORT_OK %EXPORT_TAGS);
+$VERSION = "0.9";
+@ISA = qw(Exporter);
+@EXPORT = qw(parse);
+
+# Global handler data
+my @preprocessor_conds;         # A list of simple preprocessor conditions,
+                                # each item being a list of macros defined
+                                # or not defined.
+
+# Handler helpers
+sub all_conds {
+    return map { ( @$_ ) } @preprocessor_conds;
+}
+
+# A list of handlers that will look at a "complete" string and try to
+# figure out what to make of it.
+# Each handler is a hash with the following keys:
+#
+# regexp                a regexp to compare the "complete" string with.
+# checker               a function that does a more complex comparison.
+#                       Use this instead of regexp if that isn't enough.
+# massager              massages the "complete" string into an array with
+#                       the following elements:
+#
+#                       [0]     String that needs further processing (this
+#                               applies to typedefs of structs), or empty.
+#                       [1]     The name of what was found.
+#                       [2]     A character that denotes what type of thing
+#                               this is: 'F' for function, 'S' for struct,
+#                               'T' for typedef, 'M' for macro, 'V' for
+#                               variable.
+#                       [3]     Return type (only for type 'F' and 'V')
+#                       [4]     Value (for type 'M') or signature (for type 'F',
+#                               'V', 'T' or 'S')
+#                       [5...]  The list of preprocessor conditions this is
+#                               found in, as in checks for macro definitions
+#                               (stored as the macro's name) or the absence
+#                               of definition (stored as the macro's name
+#                               prefixed with a '!'
+#
+#                       If the massager returns an empty list, it means the
+#                       "complete" string has side effects but should otherwise
+#                       be ignored.
+#                       If the massager is undefined, the "complete" string
+#                       should be ignored.
+my @opensslcpphandlers = (
+    ##################################################################
+    # OpenSSL CPP specials
+    #
+    # These are used to convert certain pre-precessor expressions into
+    # others that @cpphandlers have a better chance to understand.
+
+    { regexp   => qr/#if OPENSSL_API_COMPAT(\S+)(0x[0-9a-fA-F]{8})L$/,
+      massager => sub {
+          my $op = $1;
+          my $v = hex($2);
+          if ($op ne '<' && $op ne '>=') {
+              die "Error: unacceptable operator $op: $_[0]\n";
+          }
+          my ($one, $major, $minor) =
+              ( ($v >> 28) & 0xf,
+                ($v >> 20) & 0xff,
+                ($v >> 12) & 0xff );
+          my $t = "DEPRECATEDIN_${one}_${major}_${minor}";
+          my $cond = $op eq '<' ? 'ifndef' : 'ifdef';
+          return (<<"EOF");
+#$cond $t
+EOF
+      }
+   }
+);
+my @cpphandlers = (
+    ##################################################################
+    # CPP stuff
+
+    { regexp   => qr/#ifdef ?(.*)/,
+      massager => sub {
+          my %opts;
+          if (ref($_[$#_]) eq "HASH") {
+              %opts = %{$_[$#_]};
+              pop @_;
+          }
+          push @preprocessor_conds, [ $1 ];
+          print STDERR "DEBUG[",$opts{debug_type},"]: preprocessor level: ", scalar(@preprocessor_conds), "\n"
+              if $opts{debug};
+          return ();
+      },
+    },
+    { regexp   => qr/#ifndef ?(.*)/,
+      massager => sub {
+          my %opts;
+          if (ref($_[$#_]) eq "HASH") {
+              %opts = %{$_[$#_]};
+              pop @_;
+          }
+          push @preprocessor_conds, [ '!'.$1 ];
+          print STDERR "DEBUG[",$opts{debug_type},"]: preprocessor level: ", scalar(@preprocessor_conds), "\n"
+              if $opts{debug};
+          return ();
+      },
+    },
+    { regexp   => qr/#if (0|1)/,
+      massager => sub {
+          my %opts;
+          if (ref($_[$#_]) eq "HASH") {
+              %opts = %{$_[$#_]};
+              pop @_;
+          }
+          if ($1 eq "1") {
+              push @preprocessor_conds, [ "TRUE" ];
+          } else {
+              push @preprocessor_conds, [ "!TRUE" ];
+          }
+          print STDERR "DEBUG[",$opts{debug_type},"]: preprocessor level: ", scalar(@preprocessor_conds), "\n"
+              if $opts{debug};
+          return ();
+      },
+    },
+    { regexp   => qr/#if ?(.*)/,
+      massager => sub {
+          my %opts;
+          if (ref($_[$#_]) eq "HASH") {
+              %opts = %{$_[$#_]};
+              pop @_;
+          }
+          my @results = ();
+          my $conds = $1;
+          if ($conds =~ m|^defined<<<\(([^\)]*)\)>>>(.*)$|) {
+              push @results, $1; # Handle the simple case
+              my $rest = $2;
+              my $re = qr/^(?:\|\|defined<<<\([^\)]*\)>>>)*$/;
+              print STDERR "DEBUG[",$opts{debug_type},"]: Matching '$rest' with '$re'\n"
+                  if $opts{debug};
+              if ($rest =~ m/$re/) {
+                  my @rest = split /\|\|/, $rest;
+                  shift @rest;
+                  foreach (@rest) {
+                      m|^defined<<<\(([^\)]*)\)>>>$|;
+                      die "Something wrong...$opts{PLACE}" if $1 eq "";
+                      push @results, $1;
+                  }
+              } else {
+                  $conds =~ s/<<<|>>>//g;
+                  warn "Warning: complicated #if expression(1): $conds$opts{PLACE}"
+                      if $opts{warnings};
+              }
+          } elsif ($conds =~ m|^!defined<<<\(([^\)]*)\)>>>(.*)$|) {
+              push @results, '!'.$1; # Handle the simple case
+              my $rest = $2;
+              my $re = qr/^(?:\&\&!defined<<<\([^\)]*\)>>>)*$/;
+              print STDERR "DEBUG[",$opts{debug_type},"]: Matching '$rest' with '$re'\n"
+                  if $opts{debug};
+              if ($rest =~ m/$re/) {
+                  my @rest = split /\&\&/, $rest;
+                  shift @rest;
+                  foreach (@rest) {
+                      m|^!defined<<<\(([^\)]*)\)>>>$|;
+                      die "Something wrong...$opts{PLACE}" if $1 eq "";
+                      push @results, '!'.$1;
+                  }
+              } else {
+                  $conds =~ s/<<<|>>>//g;
+                  warn "Warning: complicated #if expression(2): $conds$opts{PLACE}"
+                      if $opts{warnings};
+              }
+          } else {
+              $conds =~ s/<<<|>>>//g;
+              warn "Warning: complicated #if expression(3): $conds$opts{PLACE}"
+                  if $opts{warnings};
+          }
+          print STDERR "DEBUG[",$opts{debug_type},"]: Added preprocessor conds: '", join("', '", @results), "'\n"
+              if $opts{debug};
+          push @preprocessor_conds, [ @results ];
+          print STDERR "DEBUG[",$opts{debug_type},"]: preprocessor level: ", scalar(@preprocessor_conds), "\n"
+              if $opts{debug};
+          return ();
+      },
+    },
+    { regexp   => qr/#elif (.*)/,
+      massager => sub {
+          my %opts;
+          if (ref($_[$#_]) eq "HASH") {
+              %opts = %{$_[$#_]};
+              pop @_;
+          }
+          die "An #elif without corresponding condition$opts{PLACE}"
+              if !@preprocessor_conds;
+          pop @preprocessor_conds;
+          print STDERR "DEBUG[",$opts{debug_type},"]: preprocessor level: ", scalar(@preprocessor_conds), "\n"
+              if $opts{debug};
+          return (<<"EOF");
+#if $1
+EOF
+      },
+    },
+    { regexp   => qr/#else/,
+      massager => sub {
+          my %opts;
+          if (ref($_[$#_]) eq "HASH") {
+              %opts = %{$_[$#_]};
+              pop @_;
+          }
+          die "An #else without corresponding condition$opts{PLACE}"
+              if !@preprocessor_conds;
+          # Invert all conditions on the last level
+          my $stuff = pop @preprocessor_conds;
+          push @preprocessor_conds, [
+              map { m|^!(.*)$| ? $1 : '!'.$_ } @$stuff
+          ];
+          print STDERR "DEBUG[",$opts{debug_type},"]: preprocessor level: ", scalar(@preprocessor_conds), "\n"
+              if $opts{debug};
+          return ();
+      },
+    },
+    { regexp   => qr/#endif ?/,
+      massager => sub {
+          my %opts;
+          if (ref($_[$#_]) eq "HASH") {
+              %opts = %{$_[$#_]};
+              pop @_;
+          }
+          die "An #endif without corresponding condition$opts{PLACE}"
+              if !@preprocessor_conds;
+          pop @preprocessor_conds;
+          print STDERR "DEBUG[",$opts{debug_type},"]: preprocessor level: ", scalar(@preprocessor_conds), "\n"
+              if $opts{debug};
+          return ();
+      },
+    },
+    { regexp   => qr/#define ([[:alpha:]_]\w*)(<<<\(.*?\)>>>)?( (.*))?/,
+      massager => sub {
+          my $name = $1;
+          my $params = $2;
+          my $spaceval = $3||"";
+          my $val = $4||"";
+          return ("",
+                  $1, 'M', "", $params ? "$name$params$spaceval" : $val,
+                  all_conds()); }
+    },
+    { regexp   => qr/#.*/,
+      massager => sub { return (); }
+    },
+    );
+
+my @opensslchandlers = (
+    ##################################################################
+    # OpenSSL C specials
+    #
+    # They are really preprocessor stuff, but they look like C stuff
+    # to this parser.  All of these do replacements, anything else is
+    # an error.
+
+    #####
+    # Global variable stuff
+    { regexp   => qr/OPENSSL_DECLARE_GLOBAL<<<\((.*),(.*)\)>>>;/,
+      massager => sub { return (<<"EOF");
+#ifndef OPENSSL_EXPORT_VAR_AS_FUNCTION
+OPENSSL_EXPORT $1 _shadow_$2;
+#else
+$1 *_shadow_$2(void);
+#endif
+EOF
+      },
+    },
+
+    #####
+    # Deprecated stuff, by OpenSSL release.
+
+    # We trick the parser by pretending that the declaration is wrapped in a
+    # check if the DEPRECATEDIN macro is defined or not.  Callers of parse()
+    # will have to decide what to do with it.
+    { regexp   => qr/(DEPRECATEDIN_\d+_\d+_\d+)<<<\((.*)\)>>>/,
+      massager => sub { return (<<"EOF");
+#ifndef $1
+$2;
+#endif
+EOF
+      },
+    },
+
+    #####
+    # LHASH stuff
+
+    # LHASH_OF(foo) is used as a type, but the chandlers won't take it
+    # gracefully, so we expand it here.
+    { regexp   => qr/(.*)\bLHASH_OF<<<\((.*?)\)>>>(.*)/,
+      massager => sub { return ("$1struct lhash_st_$2$3"); }
+    },
+    { regexp   => qr/DEFINE_LHASH_OF<<<\((.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+static ossl_inline LHASH_OF($1) * lh_$1_new(unsigned long (*hfn)(const $1 *),
+                                            int (*cfn)(const $1 *, const $1 *));
+static ossl_inline void lh_$1_free(LHASH_OF($1) *lh);
+static ossl_inline $1 *lh_$1_insert(LHASH_OF($1) *lh, $1 *d);
+static ossl_inline $1 *lh_$1_delete(LHASH_OF($1) *lh, const $1 *d);
+static ossl_inline $1 *lh_$1_retrieve(LHASH_OF($1) *lh, const $1 *d);
+static ossl_inline int lh_$1_error(LHASH_OF($1) *lh);
+static ossl_inline unsigned long lh_$1_num_items(LHASH_OF($1) *lh);
+static ossl_inline void lh_$1_node_stats_bio(const LHASH_OF($1) *lh, BIO *out);
+static ossl_inline void lh_$1_node_usage_stats_bio(const LHASH_OF($1) *lh,
+                                                   BIO *out);
+static ossl_inline void lh_$1_stats_bio(const LHASH_OF($1) *lh, BIO *out);
+static ossl_inline unsigned long lh_$1_get_down_load(LHASH_OF($1) *lh);
+static ossl_inline void lh_$1_set_down_load(LHASH_OF($1) *lh, unsigned long dl);
+static ossl_inline void lh_$1_doall(LHASH_OF($1) *lh, void (*doall)($1 *));
+LHASH_OF($1)
+EOF
+      }
+     },
+
+    #####
+    # STACK stuff
+
+    # STACK_OF(foo) is used as a type, but the chandlers won't take it
+    # gracefully, so we expand it here.
+    { regexp   => qr/(.*)\bSTACK_OF<<<\((.*?)\)>>>(.*)/,
+      massager => sub { return ("$1struct stack_st_$2$3"); }
+    },
+#    { regexp   => qr/(.*)\bSTACK_OF\((.*?)\)(.*)/,
+#      massager => sub {
+#          my $before = $1;
+#          my $stack_of = "struct stack_st_$2";
+#          my $after = $3;
+#          if ($after =~ m|^\w|) { $after = " ".$after; }
+#          return ("$before$stack_of$after");
+#      }
+#    },
+    { regexp   => qr/SKM_DEFINE_STACK_OF<<<\((.*),(.*),(.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+STACK_OF($1);
+typedef int (*sk_$1_compfunc)(const $3 * const *a, const $3 *const *b);
+typedef void (*sk_$1_freefunc)($3 *a);
+typedef $3 * (*sk_$1_copyfunc)(const $3 *a);
+static ossl_inline int sk_$1_num(const STACK_OF($1) *sk);
+static ossl_inline $2 *sk_$1_value(const STACK_OF($1) *sk, int idx);
+static ossl_inline STACK_OF($1) *sk_$1_new(sk_$1_compfunc compare);
+static ossl_inline STACK_OF($1) *sk_$1_new_null(void);
+static ossl_inline STACK_OF($1) *sk_$1_new_reserve(sk_$1_compfunc compare,
+                                                   int n);
+static ossl_inline int sk_$1_reserve(STACK_OF($1) *sk, int n);
+static ossl_inline void sk_$1_free(STACK_OF($1) *sk);
+static ossl_inline void sk_$1_zero(STACK_OF($1) *sk);
+static ossl_inline $2 *sk_$1_delete(STACK_OF($1) *sk, int i);
+static ossl_inline $2 *sk_$1_delete_ptr(STACK_OF($1) *sk, $2 *ptr);
+static ossl_inline int sk_$1_push(STACK_OF($1) *sk, $2 *ptr);
+static ossl_inline int sk_$1_unshift(STACK_OF($1) *sk, $2 *ptr);
+static ossl_inline $2 *sk_$1_pop(STACK_OF($1) *sk);
+static ossl_inline $2 *sk_$1_shift(STACK_OF($1) *sk);
+static ossl_inline void sk_$1_pop_free(STACK_OF($1) *sk,
+                                       sk_$1_freefunc freefunc);
+static ossl_inline int sk_$1_insert(STACK_OF($1) *sk, $2 *ptr, int idx);
+static ossl_inline $2 *sk_$1_set(STACK_OF($1) *sk, int idx, $2 *ptr);
+static ossl_inline int sk_$1_find(STACK_OF($1) *sk, $2 *ptr);
+static ossl_inline int sk_$1_find_ex(STACK_OF($1) *sk, $2 *ptr);
+static ossl_inline void sk_$1_sort(STACK_OF($1) *sk);
+static ossl_inline int sk_$1_is_sorted(const STACK_OF($1) *sk);
+static ossl_inline STACK_OF($1) * sk_$1_dup(const STACK_OF($1) *sk);
+static ossl_inline STACK_OF($1) *sk_$1_deep_copy(const STACK_OF($1) *sk,
+                                                 sk_$1_copyfunc copyfunc,
+                                                 sk_$1_freefunc freefunc);
+static ossl_inline sk_$1_compfunc sk_$1_set_cmp_func(STACK_OF($1) *sk,
+                                                     sk_$1_compfunc compare);
+EOF
+      }
+    },
+    { regexp   => qr/DEFINE_SPECIAL_STACK_OF<<<\((.*),(.*)\)>>>/,
+      massager => sub { return ("SKM_DEFINE_STACK_OF($1,$2,$2)"); },
+    },
+    { regexp   => qr/DEFINE_STACK_OF<<<\((.*)\)>>>/,
+      massager => sub { return ("SKM_DEFINE_STACK_OF($1,$1,$1)"); },
+    },
+    { regexp   => qr/DEFINE_SPECIAL_STACK_OF_CONST<<<\((.*),(.*)\)>>>/,
+      massager => sub { return ("SKM_DEFINE_STACK_OF($1,const $2,$2)"); },
+    },
+    { regexp   => qr/DEFINE_STACK_OF_CONST<<<\((.*)\)>>>/,
+      massager => sub { return ("SKM_DEFINE_STACK_OF($1,const $1,$1)"); },
+    },
+    { regexp   => qr/PREDECLARE_STACK_OF<<<\((.*)\)>>>/,
+      massager => sub { return ("STACK_OF($1);"); }
+    },
+    { regexp   => qr/DECLARE_STACK_OF<<<\((.*)\)>>>/,
+      massager => sub { return ("STACK_OF($1);"); }
+    },
+    { regexp   => qr/DECLARE_SPECIAL_STACK_OF<<<\((.*?),(.*?)\)>>>/,
+      massager => sub { return ("STACK_OF($1);"); }
+     },
+
+    #####
+    # ASN1 stuff
+
+    { regexp   => qr/TYPEDEF_D2I_OF<<<\((.*)\)>>>/,
+      massager => sub {
+          return ("typedef $1 *d2i_of_$1($1 **,const unsigned char **,long)");
+      },
+    },
+    { regexp   => qr/TYPEDEF_I2D_OF<<<\((.*)\)>>>/,
+      massager => sub {
+          return ("typedef $1 *i2d_of_$1($1 *,unsigned char **)");
+      },
+    },
+    { regexp   => qr/TYPEDEF_D2I2D_OF<<<\((.*)\)>>>/,
+      massager => sub {
+          return ("TYPEDEF_D2I_OF($1); TYPEDEF_I2D_OF($1)");
+      },
+    },
+    { regexp   => qr/DECLARE_ASN1_ITEM<<<\((.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+#ifndef OPENSSL_EXPORT_VAR_AS_FUNCTION
+OPENSSL_EXTERN const ASN1_ITEM *$1_it;
+#else
+const ASN1_ITEM *$1_it(void);
+#endif
+EOF
+      },
+    },
+    { regexp   => qr/DECLARE_ASN1_ENCODE_FUNCTIONS<<<\((.*),(.*),(.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+int d2i_$3(void);
+int i2d_$3(void);
+DECLARE_ASN1_ITEM($2)
+EOF
+      },
+    },
+    { regexp   => qr/DECLARE_ASN1_ENCODE_FUNCTIONS_const<<<\((.*),(.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+int d2i_$2(void);
+int i2d_$2(void);
+DECLARE_ASN1_ITEM($2)
+EOF
+      },
+    },
+    { regexp   => qr/DECLARE_ASN1_ALLOC_FUNCTIONS<<<\((.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+int $1_free(void);
+int $1_new(void);
+EOF
+      },
+    },
+    { regexp   => qr/DECLARE_ASN1_FUNCTIONS_name<<<\((.*),(.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+int d2i_$2(void);
+int i2d_$2(void);
+int $2_free(void);
+int $2_new(void);
+DECLARE_ASN1_ITEM($2)
+EOF
+      },
+    },
+    { regexp   => qr/DECLARE_ASN1_FUNCTIONS_fname<<<\((.*),(.*),(.*)\)>>>/,
+      massager => sub { return (<<"EOF");
+int d2i_$3(void);
+int i2d_$3(void);
+int $3_free(void);
+int $3_new(void);
+DECLARE_ASN1_ITEM($2)
+EOF
+      }
+    },
+    { regexp   => qr/DECLARE_ASN1_FUNCTIONS(?:_const)?<<<\((.*)\)>>>/,
+      massager => sub { return (<<"EOF");
+int d2i_$1(void);
+int i2d_$1(void);
+int $1_free(void);
+int $1_new(void);
+DECLARE_ASN1_ITEM($1)
+EOF
+      }
+    },
+    { regexp   => qr/DECLARE_ASN1_NDEF_FUNCTION<<<\((.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+int i2d_$1_NDEF(void);
+EOF
+      }
+    },
+    { regexp   => qr/DECLARE_ASN1_PRINT_FUNCTION<<<\((.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+int $1_print_ctx(void);
+EOF
+      }
+    },
+    { regexp   => qr/DECLARE_ASN1_PRINT_FUNCTION_name<<<\((.*),(.*)\)>>>/,
+      massager => sub {
+          return (<<"EOF");
+int $2_print_ctx(void);
+EOF
+      }
+    },
+    { regexp   => qr/DECLARE_ASN1_SET_OF<<<\((.*)\)>>>/,
+      massager => sub { return (); }
+    },
+    { regexp   => qr/DECLARE_PKCS12_SET_OF<<<\((.*)\)>>>/,
+      massager => sub { return (); }
+    },
+    { regexp   => qr/DECLARE_PEM(?|_rw|_rw_cb|_rw_const)<<<\((.*?),.*\)>>>/,
+      massager => sub { return (<<"EOF");
+#ifndef OPENSSL_NO_STDIO
+int PEM_read_$1(void);
+int PEM_write_$1(void);
+#endif
+int PEM_read_bio_$1(void);
+int PEM_write_bio_$1(void);
+EOF
+      },
+    },
+
+    #####
+    # PEM stuff
+    { regexp   => qr/DECLARE_PEM(?|_write|_write_cb|_write_const)<<<\((.*?),.*\)>>>/,
+      massager => sub { return (<<"EOF");
+#ifndef OPENSSL_NO_STDIO
+int PEM_write_$1(void);
+#endif
+int PEM_write_bio_$1(void);
+EOF
+      },
+    },
+    { regexp   => qr/DECLARE_PEM(?|_read|_read_cb)<<<\((.*?),.*\)>>>/,
+      massager => sub { return (<<"EOF");
+#ifndef OPENSSL_NO_STDIO
+int PEM_read_$1(void);
+#endif
+int PEM_read_bio_$1(void);
+EOF
+      },
+    },
+
+    # Spurious stuff found in the OpenSSL headers
+    # Usually, these are just macros that expand to, well, something
+    { regexp   => qr/__NDK_FPABI__/,
+      massager => sub { return (); }
+    },
+    );
+
+my $anoncnt = 0;
+
+my @chandlers = (
+    ##################################################################
+    # C stuff
+
+    # extern "C" of individual items
+    # Note that the main parse function has a special hack for 'extern "C" {'
+    # which can't be done in handlers
+    # We simply ignore it.
+    { regexp   => qr/extern "C" (.*;)/,
+      massager => sub { return ($1); },
+    },
+    # union, struct and enum definitions
+    # Because this one might appear a little everywhere within type
+    # definitions, we take it out and replace it with just
+    # 'union|struct|enum name' while registering it.
+    # This makes use of the parser trick to surround the outer braces
+    # with <<< and >>>
+    { regexp   => qr/(.*)                       # Anything before       ($1)
+                     \b                         # word to non-word boundary
+                     (union|struct|enum)        # The word used         ($2)
+                     (?:\s([[:alpha:]_]\w*))?   # Struct or enum name   ($3)
+                     <<<(\{.*?\})>>>            # Struct or enum definition ($4)
+                     (.*)                       # Anything after        ($5)
+                     ;
+                    /x,
+      massager => sub {
+          my $before = $1;
+          my $word = $2;
+          my $name = $3
+              || sprintf("__anon%03d", ++$anoncnt); # Anonymous struct
+          my $definition = $4;
+          my $after = $5;
+          my $type = $word eq "struct" ? 'S' : 'E';
+          if ($before ne "" || $after ne ";") {
+              if ($after =~ m|^\w|) { $after = " ".$after; }
+              return ("$before$word $name$after;",
+                      "$word $name", $type, "", "$word$definition", all_conds());
+          }
+          # If there was no before nor after, make the return much simple
+          return ("", "$word $name", $type, "", "$word$definition", all_conds());
+      }
+    },
+    # Named struct and enum forward declarations
+    # We really just ignore them, but we need to parse them or the variable
+    # declaration handler further down will think it's a variable declaration.
+    { regexp   => qr/^(union|struct|enum) ([[:alpha:]_]\w*);/,
+      massager => sub { return (); }
+    },
+    # Function returning function pointer declaration
+    { regexp   => qr/(?:(typedef)\s?)?          # Possible typedef      ($1)
+                     ((?:\w|\*|\s)*?)           # Return type           ($2)
+                     \s?                        # Possible space
+                     <<<\(\*
+                     ([[:alpha:]_]\w*)          # Function name         ($3)
+                     (\(.*\))                   # Parameters            ($4)
+                     \)>>>
+                     <<<(\(.*\))>>>             # F.p. parameters       ($5)
+                     ;
+                    /x,
+      massager => sub {
+          return ("", $3, 'F', "", "$2(*$4)$5", all_conds())
+              if defined $1;
+          return ("", $3, 'F', "$2(*)$5", "$2(*$4)$5", all_conds()); }
+    },
+    # Function pointer declaration, or typedef thereof
+    { regexp   => qr/(?:(typedef)\s?)?          # Possible typedef      ($1)
+                     ((?:\w|\*|\s)*?)           # Return type           ($2)
+                     <<<\(\*([[:alpha:]_]\w*)\)>>> # T.d. or var name   ($3)
+                     <<<(\(.*\))>>>             # F.p. parameters       ($4)
+                     ;
+                    /x,
+      massager => sub {
+          return ("", $3, 'T', "", "$2(*)$4", all_conds())
+              if defined $1;
+          return ("", $3, 'V', "$2(*)$4", "$2(*)$4", all_conds());
+      },
+    },
+    # Function declaration, or typedef thereof
+    { regexp   => qr/(?:(typedef)\s?)?          # Possible typedef      ($1)
+                     ((?:\w|\*|\s)*?)           # Return type           ($2)
+                     \s?                        # Possible space
+                     ([[:alpha:]_]\w*)          # Function name         ($3)
+                     <<<(\(.*\))>>>             # Parameters            ($4)
+                     ;
+                    /x,
+      massager => sub {
+          return ("", $3, 'T', "", "$2$4", all_conds())
+              if defined $1;
+          return ("", $3, 'F', $2, "$2$4", all_conds());
+      },
+    },
+    # Variable declaration, including arrays, or typedef thereof
+    { regexp   => qr/(?:(typedef)\s?)?          # Possible typedef      ($1)
+                     ((?:\w|\*|\s)*?)           # Type                  ($2)
+                     \s?                        # Possible space
+                     ([[:alpha:]_]\w*)          # Variable name         ($3)
+                     ((?:<<<\[[^\]]*\]>>>)*)    # Possible array declaration ($4)
+                     ;
+                    /x,
+      massager => sub {
+          return ("", $3, 'T', "", $2.($4||""), all_conds())
+              if defined $1;
+          return ("", $3, 'V', $2.($4||""), $2.($4||""), all_conds());
+      },
+    },
+);
+
+# End handlers are almost the same as handlers, except they are run through
+# ONCE when the input has been parsed through.  These are used to check for
+# remaining stuff, such as an unfinished #ifdef and stuff like that that the
+# main parser can't check on its own.
+my @endhandlers = (
+    { massager => sub {
+        my %opts = %{$_[0]};
+
+        die "Unfinished preprocessor conditions levels: ",scalar(@preprocessor_conds),($opts{filename} ? " in file ".$opts{filename}: ""),$opts{PLACE}
+            if @preprocessor_conds;
+      }
+    }
+    );
+
+# takes a list of strings that can each contain one or several lines of code
+# also takes a hash of options as last argument.
+#
+# returns a list of hashes with information:
+#
+#       name            name of the thing
+#       type            type, see the massage handler function
+#       returntype      return type of functions and variables
+#       value           value for macros, signature for functions, variables
+#                       and structs
+#       conds           preprocessor conditions (array ref)
+
+sub parse {
+    my %opts;
+    if (ref($_[$#_]) eq "HASH") {
+        %opts = %{$_[$#_]};
+        pop @_;
+    }
+    my %state = (
+        in_extern_C => 0,       # An exception to parenthesis processing.
+        cpp_parens => [],       # A list of ending parens and braces found in
+                                # preprocessor directives
+        c_parens => [],         # A list of ending parens and braces found in
+                                # C statements
+        in_string => "",        # empty string when outside a string, otherwise
+                                # "'" or '"' depending on the starting quote.
+        in_comment => "",       # empty string when outside a comment, otherwise
+                                # "/*" or "//" depending on the type of comment
+                                # found.  The latter will never be multiline
+                                # NOTE: in_string and in_comment will never be
+                                # true (in perl semantics) at the same time.
+        current_line => 0,
+        );
+    my @result = ();
+    my $normalized_line = "";   # $input_line, but normalized.  In essence, this
+                                # means that ALL whitespace is removed unless
+                                # it absolutely has to be present, and in that
+                                # case, there's only one space.
+                                # The cases where a space needs to stay present
+                                # are:
+                                # 1. between words
+                                # 2. between words and number
+                                # 3. after the first word of a preprocessor
+                                #    directive.
+                                # 4. for the #define directive, between the macro
+                                #    name/args and its value, so we end up with:
+                                #       #define FOO val
+                                #       #define BAR(x) something(x)
+    my $collected_stmt = "";    # Where we're building up a C line until it's a
+                                # complete definition/declaration, as determined
+                                # by any handler being capable of matching it.
+
+    # We use $_ shamelessly when looking through @lines.
+    # In case we find a \ at the end, we keep filling it up with more lines.
+    $_ = undef;
+
+    foreach my $line (@_) {
+        # split tries to be smart when a string ends with the thing we split on
+        $line .= "\n" unless $line =~ m|\R$|;
+        $line .= "#";
+
+        # We use ¬¶undef¬¶ as a marker for a new line from the file.
+        # Since we convert one line to several and unshift that into @lines,
+        # that's the only safe way we have to track the original lines
+        my @lines = map { ( undef, $_ ) } split $/, $line;
+
+        # Remember that extra # we added above?  Now we remove it
+        pop @lines;
+        pop @lines;             # Don't forget the undef
+
+        while (@lines) {
+            if (!defined($lines[0])) {
+                shift @lines;
+                $state{current_line}++;
+                if (!defined($_)) {
+                    $opts{PLACE} = " at ".$opts{filename}." line ".$state{current_line}."\n";
+                    $opts{PLACE2} = $opts{filename}.":".$state{current_line};
+                }
+                next;
+            }
+
+            $_ = "" unless defined $_;
+            $_ .= shift @lines;
+
+            if (m|\\$|) {
+                $_ = $`;
+                next;
+            }
+
+            if ($opts{debug}) {
+                print STDERR "DEBUG:----------------------------\n";
+                print STDERR "DEBUG: \$_      = '$_'\n";
+            }
+
+            ##########################################################
+            # Now that we have a full line, let's process through it
+            while(1) {
+                unless ($state{in_comment}) {
+                    # Begin with checking if the current $normalized_line
+                    # contains a preprocessor directive
+                    # This is only done if we're not inside a comment and
+                    # if it's a preprocessor directive and it's finished.
+                    if ($normalized_line =~ m|^#| && $_ eq "") {
+                        print STDERR "DEBUG[OPENSSL CPP]: \$normalized_line = '$normalized_line'\n"
+                            if $opts{debug};
+                        $opts{debug_type} = "OPENSSL CPP";
+                        my @r = ( _run_handlers($normalized_line,
+                                                @opensslcpphandlers,
+                                                \%opts) );
+                        if (shift @r) {
+                            # Checking if there are lines to inject.
+                            if (@r) {
+                                @r = split $/, (pop @r).$_;
+                                print STDERR "DEBUG[OPENSSL CPP]: injecting '", join("', '", @r),"'\n"
+                                    if $opts{debug} && @r;
+                                @lines = ( @r, @lines );
+
+                                $_ = "";
+                            }
+                        } else {
+                            print STDERR "DEBUG[CPP]: \$normalized_line = '$normalized_line'\n"
+                                if $opts{debug};
+                            $opts{debug_type} = "CPP";
+                            my @r = ( _run_handlers($normalized_line,
+                                                    @cpphandlers,
+                                                    \%opts) );
+                            if (shift @r) {
+                                if (ref($r[0]) eq "HASH") {
+                                    push @result, shift @r;
+                                }
+
+                                # Now, check if there are lines to inject.
+                                # Really, this should never happen, it IS a
+                                # preprocessor directive after all...
+                                if (@r) {
+                                    @r = split $/, pop @r;
+                                    print STDERR "DEBUG[CPP]: injecting '", join("', '", @r),"'\n"
+                                    if $opts{debug} && @r;
+                                    @lines = ( @r, @lines );
+                                    $_ = "";
+                                }
+                            }
+                        }
+
+                        # Note: we simply ignore all directives that no
+                        # handler matches
+                        $normalized_line = "";
+                    }
+
+                    # If the two strings end and start with a character that
+                    # shouldn't get concatenated, add a space
+                    my $space =
+                        ($collected_stmt =~ m/(?:"|')$/
+                         || ($collected_stmt =~ m/(?:\w|\d)$/
+                             && $normalized_line =~ m/^(?:\w|\d)/)) ? " " : "";
+
+                    # Now, unless we're building up a preprocessor directive or
+                    # are in the middle of a string, or the parens et al aren't
+                    # balanced up yet, let's try and see if there's a OpenSSL
+                    # or C handler that can make sense of what we have so far.
+                    if ( $normalized_line !~ m|^#|
+                         && ($collected_stmt ne "" || $normalized_line ne "")
+                         && ! @{$state{c_parens}}
+                         && ! $state{in_string} ) {
+                        if ($opts{debug}) {
+                            print STDERR "DEBUG[OPENSSL C]: \$collected_stmt  = '$collected_stmt'\n";
+                            print STDERR "DEBUG[OPENSSL C]: \$normalized_line = '$normalized_line'\n";
+                        }
+                        $opts{debug_type} = "OPENSSL C";
+                        my @r = ( _run_handlers($collected_stmt
+                                                    .$space
+                                                    .$normalized_line,
+                                                @opensslchandlers,
+                                                \%opts) );
+                        if (shift @r) {
+                            # Checking if there are lines to inject.
+                            if (@r) {
+                                @r = split $/, (pop @r).$_;
+                                print STDERR "DEBUG[OPENSSL]: injecting '", join("', '", @r),"'\n"
+                                    if $opts{debug} && @r;
+                                @lines = ( @r, @lines );
+
+                                $_ = "";
+                            }
+                            $normalized_line = "";
+                            $collected_stmt = "";
+                        } else {
+                            if ($opts{debug}) {
+                                print STDERR "DEBUG[C]: \$collected_stmt  = '$collected_stmt'\n";
+                                print STDERR "DEBUG[C]: \$normalized_line = '$normalized_line'\n";
+                            }
+                            $opts{debug_type} = "C";
+                            my @r = ( _run_handlers($collected_stmt
+                                                        .$space
+                                                        .$normalized_line,
+                                                    @chandlers,
+                                                    \%opts) );
+                            if (shift @r) {
+                                if (ref($r[0]) eq "HASH") {
+                                    push @result, shift @r;
+                                }
+
+                                # Checking if there are lines to inject.
+                                if (@r) {
+                                    @r = split $/, (pop @r).$_;
+                                    print STDERR "DEBUG[C]: injecting '", join("', '", @r),"'\n"
+                                        if $opts{debug} && @r;
+                                    @lines = ( @r, @lines );
+
+                                    $_ = "";
+                                }
+                                $normalized_line = "";
+                                $collected_stmt = "";
+                            }
+                        }
+                    }
+                    if ($_ eq "") {
+                        $collected_stmt .= $space.$normalized_line;
+                        $normalized_line = "";
+                    }
+                }
+
+                if ($_ eq "") {
+                    $_ = undef;
+                    last;
+                }
+
+                # Take care of inside string first.
+                if ($state{in_string}) {
+                    if (m/ (?:^|(?<!\\))        # Make sure it's not escaped
+                           $state{in_string}    # Look for matching quote
+                         /x) {
+                        $normalized_line .= $`.$&;
+                        $state{in_string} = "";
+                        $_ = $';
+                        next;
+                    } else {
+                        die "Unfinished string without continuation found$opts{PLACE}\n";
+                    }
+                }
+                # ... or inside comments, whichever happens to apply
+                elsif ($state{in_comment}) {
+
+                    # This should never happen
+                    die "Something went seriously wrong, multiline //???$opts{PLACE}\n"
+                        if ($state{in_comment} eq "//");
+
+                    # A note: comments are simply discarded.
+
+                    if (m/ (?:^|(?<!\\))        # Make sure it's not escaped
+                           \*\/                 # Look for C comment end
+                         /x) {
+                        $state{in_comment} = "";
+                        $_ = $';
+                        print STDERR "DEBUG: Found end of comment, followed by '$_'\n"
+                            if $opts{debug};
+                        next;
+                    } else {
+                        $_ = "";
+                        next;
+                    }
+                }
+
+                # At this point, it's safe to remove leading whites, but
+                # we need to be careful with some preprocessor lines
+                if (m|^\s+|) {
+                    my $rest = $';
+                    my $space = "";
+                    $space = " "
+                        if ($normalized_line =~ m/^
+                                                  \#define\s\w(?:\w|\d)*(?:<<<\([^\)]*\)>>>)?
+                                                  | \#[a-z]+
+                                                  $/x);
+                    print STDERR "DEBUG: Processing leading spaces: \$normalized_line = '$normalized_line', \$space = '$space', \$rest = '$rest'\n"
+                        if $opts{debug};
+                    $_ = $space.$rest;
+                }
+
+                my $parens =
+                    $normalized_line =~ m|^#| ? 'cpp_parens' : 'c_parens';
+                (my $paren_singular = $parens) =~ s|s$||;
+
+                # Now check for specific tokens, and if they are parens,
+                # check them against $state{$parens}.  Note that we surround
+                # the outermost parens with extra "<<<" and ">>>".  Those
+                # are for the benefit of handlers who to need to detect
+                # them, and they will be removed from the final output.
+                if (m|^[\{\[\(]|) {
+                    my $body = $&;
+                    $_ = $';
+                    if (!@{$state{$parens}}) {
+                        if ("$normalized_line$body" =~ m|^extern "C"\{$|) {
+                            $state{in_extern_C} = 1;
+                            print STDERR "DEBUG: found start of 'extern \"C\"' ($normalized_line$body)\n"
+                                if $opts{debug};
+                            $normalized_line = "";
+                        } else {
+                            $normalized_line .= "<<<".$body;
+                        }
+                    } else {
+                        $normalized_line .= $body;
+                    }
+
+                    if ($normalized_line ne "") {
+                        print STDERR "DEBUG: found $paren_singular start '$body'\n"
+                            if $opts{debug};
+                        $body =~ tr|\{\[\(|\}\]\)|;
+                        print STDERR "DEBUG: pushing $paren_singular end '$body'\n"
+                            if $opts{debug};
+                        push @{$state{$parens}}, $body;
+                    }
+                } elsif (m|^[\}\]\)]|) {
+                    $_ = $';
+
+                    if (!@{$state{$parens}}
+                        && $& eq '}' && $state{in_extern_C}) {
+                        print STDERR "DEBUG: found end of 'extern \"C\"'\n"
+                            if $opts{debug};
+                        $state{in_extern_C} = 0;
+                    } else {
+                        print STDERR "DEBUG: Trying to match '$&' against '"
+                            ,join("', '", @{$state{$parens}})
+                            ,"'\n"
+                            if $opts{debug};
+                        die "Unmatched parentheses$opts{PLACE}\n"
+                            unless (@{$state{$parens}}
+                                    && pop @{$state{$parens}} eq $&);
+                        if (!@{$state{$parens}}) {
+                            $normalized_line .= $&.">>>";
+                        } else {
+                            $normalized_line .= $&;
+                        }
+                    }
+                } elsif (m|^["']|) { # string start
+                    my $body = $&;
+                    $_ = $';
+
+                    # We want to separate strings from \w and \d with one space.
+                    $normalized_line .= " " if $normalized_line =~ m/(\w|\d)$/;
+                    $normalized_line .= $body;
+                    $state{in_string} = $body;
+                } elsif (m|^\/\*|) { # C style comment
+                    print STDERR "DEBUG: found start of C style comment\n"
+                        if $opts{debug};
+                    $state{in_comment} = $&;
+                    $_ = $';
+                } elsif (m|^\/\/|) { # C++ style comment
+                    print STDERR "DEBUG: found C++ style comment\n"
+                        if $opts{debug};
+                    $_ = "";    # (just discard it entirely)
+                } elsif (m/^ (?| (?: 0[xX][[:xdigit:]]+ | 0[bB][01]+ | [0-9]+ )
+                                 (?i: U | L | UL | LL | ULL )?
+                               | [0-9]+\.[0-9]+(?:[eE][\-\+]\d+)? (?i: F | L)?
+                               ) /x) {
+                    print STDERR "DEBUG: Processing numbers: \$normalized_line = '$normalized_line', \$& = '$&', \$' = '$''\n"
+                        if $opts{debug};
+                    $normalized_line .= $&;
+                    $_ = $';
+                } elsif (m/^[[:alpha:]_]\w*/) {
+                    my $body = $&;
+                    my $rest = $';
+                    my $space = "";
+
+                    # Now, only add a space if it's needed to separate
+                    # two \w characters, and we also surround strings with
+                    # a space.  In this case, that's if $normalized_line ends
+                    # with a \w, \d, " or '.
+                    $space = " "
+                        if ($normalized_line =~ m/("|')$/
+                            || ($normalized_line =~ m/(\w|\d)$/
+                                && $body =~ m/^(\w|\d)/));
+
+                    print STDERR "DEBUG: Processing words: \$normalized_line = '$normalized_line', \$space = '$space', \$body = '$body', \$rest = '$rest'\n"
+                        if $opts{debug};
+                    $normalized_line .= $space.$body;
+                    $_ = $rest;
+                } elsif (m|^(?:\\)?.|) { # Catch-all
+                    $normalized_line .= $&;
+                    $_ = $';
+                }
+            }
+        }
+    }
+    foreach my $handler (@endhandlers) {
+        if ($handler->{massager}) {
+            $handler->{massager}->(\%opts);
+        }
+    }
+    return @result;
+}
+
+# arg1:    line to check
+# arg2...: handlers to check
+# return undef when no handler matched
+sub _run_handlers {
+    my %opts;
+    if (ref($_[$#_]) eq "HASH") {
+        %opts = %{$_[$#_]};
+        pop @_;
+    }
+    my $line = shift;
+    my @handlers = @_;
+
+    foreach my $handler (@handlers) {
+        if ($handler->{regexp}
+            && $line =~ m|^$handler->{regexp}$|) {
+            if ($handler->{massager}) {
+                if ($opts{debug}) {
+                    print STDERR "DEBUG[",$opts{debug_type},"]: Trying to handle '$line'\n";
+                    print STDERR "DEBUG[",$opts{debug_type},"]: (matches /\^",$handler->{regexp},"\$/)\n";
+                }
+                my $saved_line = $line;
+                my @massaged =
+                    map { s/(<<<|>>>)//g; $_ }
+                    $handler->{massager}->($saved_line, \%opts);
+                print STDERR "DEBUG[",$opts{debug_type},"]: Got back '"
+                    , join("', '", @massaged), "'\n"
+                    if $opts{debug};
+
+                # Because we may get back new lines to be
+                # injected before whatever else that follows,
+                # and the injected stuff might include
+                # preprocessor lines, we need to inject them
+                # in @lines and set $_ to the empty string to
+                # break out from the inner loops
+                my $injected_lines = shift @massaged || "";
+
+                if (@massaged) {
+                    return (1,
+                            {
+                                name    => shift @massaged,
+                                type    => shift @massaged,
+                                returntype => shift @massaged,
+                                value   => shift @massaged,
+                                conds   => [ @massaged ]
+                            },
+                            $injected_lines
+                        );
+                } else {
+                    print STDERR "DEBUG[",$opts{debug_type},"]:   (ignore, possible side effects)\n"
+                        if $opts{debug} && $injected_lines eq "";
+                    return (1, $injected_lines);
+                }
+            }
+            return (1);
+        }
+    }
+    return (0);
+}