type: Make an assert()ion independent of the last entry of an enum.
[cparser] / preprocessor.c
index 4c5f1d4..5d4213b 100644 (file)
@@ -1,3 +1,7 @@
+/*
+ * This file is part of cparser.
+ * Copyright (C) 2012 Matthias Braun <matze@braunis.de>
+ */
 #include <config.h>
 
 #include <assert.h>
@@ -38,7 +42,7 @@ typedef struct whitespace_info_t {
 
 struct pp_definition_t {
        symbol_t          *symbol;
-       source_position_t  source_position;
+       position_t         pos;
        pp_definition_t   *parent_expansion;
        size_t             expand_pos;
        whitespace_info_t  expand_info;
@@ -57,7 +61,7 @@ struct pp_definition_t {
 
 typedef struct pp_conditional_t pp_conditional_t;
 struct pp_conditional_t {
-       source_position_t  source_position;
+       position_t         pos;
        bool               condition;
        bool               in_else;
        /** conditional in skip mode (then+else gets skipped) */
@@ -73,7 +77,7 @@ struct pp_input_t {
        utf32               buf[1024+MAX_PUTBACK];
        const utf32        *bufend;
        const utf32        *bufpos;
-       source_position_t   position;
+       position_t          pos;
        pp_input_t         *parent;
        unsigned            output_line;
        searchpath_entry_t *path;
@@ -82,6 +86,7 @@ struct pp_input_t {
 struct searchpath_entry_t {
        const char         *path;
        searchpath_entry_t *next;
+       bool                is_system_path;
 };
 
 static pp_input_t      input;
@@ -101,7 +106,7 @@ static FILE             *out;
 static struct obstack    pp_obstack;
 static struct obstack    config_obstack;
 static const char       *printed_input_name = NULL;
-static source_position_t expansion_pos;
+static position_t        expansion_pos;
 static pp_definition_t  *current_expansion  = NULL;
 static pp_definition_t  *current_call       = NULL;
 static pp_definition_t  *current_argument   = NULL;
@@ -110,14 +115,23 @@ static unsigned          argument_brace_count;
 static strset_t          stringset;
 static token_kind_t      last_token;
 
-static searchpath_entry_t *searchpath;
+struct searchpath_t {
+       searchpath_entry_t  *first;
+       searchpath_entry_t **anchor;
+       bool                 is_system_path;
+};
+
+searchpath_t bracket_searchpath = { NULL, &bracket_searchpath.first, false };
+searchpath_t quote_searchpath   = { NULL, &quote_searchpath.first,   false };
+searchpath_t system_searchpath  = { NULL, &system_searchpath.first,  true  };
+searchpath_t after_searchpath   = { NULL, &after_searchpath.first,   true  };
 
 static whitespace_info_t next_info; /* valid if had_whitespace is true */
 static whitespace_info_t info;
 
 static inline void next_char(void);
 static void next_input_token(void);
-static void print_line_directive(const source_position_t *pos, const char *add);
+static void print_line_directive(const position_t *pos, const char *add);
 
 static symbol_t *symbol_colongreater;
 static symbol_t *symbol_lesscolon;
@@ -126,6 +140,11 @@ static symbol_t *symbol_percentcolon;
 static symbol_t *symbol_percentcolonpercentcolon;
 static symbol_t *symbol_percentgreater;
 
+static symbol_t *symbol_L;
+static symbol_t *symbol_U;
+static symbol_t *symbol_u;
+static symbol_t *symbol_u8;
+
 static void init_symbols(void)
 {
        symbol_colongreater             = symbol_table_insert(":>");
@@ -134,25 +153,31 @@ static void init_symbols(void)
        symbol_percentcolon             = symbol_table_insert("%:");
        symbol_percentcolonpercentcolon = symbol_table_insert("%:%:");
        symbol_percentgreater           = symbol_table_insert("%>");
+
+       symbol_L  = symbol_table_insert("L");
+       symbol_U  = symbol_table_insert("U");
+       symbol_u  = symbol_table_insert("u");
+       symbol_u8 = symbol_table_insert("u8");
 }
 
-void switch_pp_input(FILE *const file, char const *const filename, searchpath_entry_t *const path)
+void switch_pp_input(FILE *const file, char const *const filename, searchpath_entry_t *const path, bool const is_system_header)
 {
-       input.file                = file;
-       input.input               = input_from_stream(file, NULL);
-       input.bufend              = NULL;
-       input.bufpos              = NULL;
-       input.output_line         = 0;
-       input.position.input_name = filename;
-       input.position.lineno     = 1;
-       input.path                = path;
+       input.file                 = file;
+       input.input                = input_from_stream(file, NULL);
+       input.bufend               = NULL;
+       input.bufpos               = NULL;
+       input.output_line          = 0;
+       input.pos.input_name       = filename;
+       input.pos.lineno           = 1;
+       input.pos.is_system_header = is_system_header;
+       input.path                 = path;
 
        /* indicate that we're at a new input */
-       print_line_directive(&input.position, input_stack != NULL ? "1" : NULL);
+       print_line_directive(&input.pos, input_stack != NULL ? "1" : NULL);
 
        /* place a virtual '\n' so we realize we're at line begin */
-       input.position.lineno = 0;
-       input.c               = '\n';
+       input.pos.lineno = 0;
+       input.c          = '\n';
 }
 
 FILE *close_pp_input(void)
@@ -214,7 +239,7 @@ static void pop_restore_input(void)
  */
 static void parse_error(const char *msg)
 {
-       errorf(&pp_token.base.source_position,  "%s", msg);
+       errorf(&pp_token.base.pos,  "%s", msg);
 }
 
 static inline void next_real_char(void)
@@ -230,7 +255,7 @@ static inline void next_real_char(void)
                input.bufend = input.bufpos + n;
        }
        input.c = *input.bufpos++;
-       ++input.position.colno;
+       ++input.pos.colno;
 }
 
 /**
@@ -242,7 +267,7 @@ static inline void put_back(utf32 const pc)
 {
        assert(input.bufpos > input.buf);
        *(--input.bufpos - input.buf + input.buf) = (char) pc;
-       --input.position.colno;
+       --input.pos.colno;
 }
 
 #define NEWLINE \
@@ -252,8 +277,8 @@ static inline void put_back(utf32 const pc)
        case '\n': \
                        next_char(); \
                } \
-               ++input.position.lineno; \
-               input.position.colno = 1; \
+               ++input.pos.lineno; \
+               input.pos.colno = 1; \
                goto newline; \
                newline // Let it look like an ordinary case label.
 
@@ -434,21 +459,96 @@ static utf32 parse_universal_char(unsigned const n_digits)
                                obstack_1grow(&symbol_obstack, input.c);
                        next_char();
                } else {
-                       errorf(&input.position,
+                       errorf(&input.pos,
                               "short universal character name, expected %u more digits",
                                   k);
                        break;
                }
        }
        if (!is_universal_char_valid(v)) {
-               errorf(&input.position,
+               errorf(&input.pos,
                       "\\%c%0*X is not a valid universal character name",
                       n_digits == 4 ? 'u' : 'U', (int)n_digits, v);
        }
        return v;
 }
 
-static bool is_universal_char_valid_identifier(utf32 const v)
+static bool is_universal_char_valid_identifier_c99(utf32 const v)
+{
+       static const utf32 single_chars[] = {
+               0x00AA, 0x00BA, 0x0386, 0x038C, 0x03DA, 0x03DC, 0x03DE, 0x03E0,
+               0x1F59, 0x1F5B, 0x1F5D, 0x05BF, 0x09B2, 0x0A02, 0x0A5E, 0x0A74,
+               0x0A8D, 0x0AD0, 0x0AE0, 0x0B9C, 0x0CDE, 0x0E84, 0x0E8A, 0x0E8D,
+               0x0EA5, 0x0EA7, 0x0EC6, 0x0F00, 0x0F35, 0x0F37, 0x0F39, 0x0F97,
+               0x0FB9, 0x00B5, 0x00B7, 0x02BB, 0x037A, 0x0559, 0x093D, 0x0B3D,
+               0x1FBE, 0x2102, 0x2107, 0x2115, 0x2124, 0x2126, 0x2128
+       };
+
+       static const utf32 ranges[][2] = {
+               {0x00C0, 0x00D6}, {0x00D8, 0x00F6}, {0x00F8, 0x01F5}, {0x01FA, 0x0217},
+               {0x0250, 0x02A8}, {0x1E00, 0x1E9B}, {0x1EA0, 0x1EF9}, {0x0388, 0x038A},
+               {0x038E, 0x03A1}, {0x03A3, 0x03CE}, {0x03D0, 0x03D6}, {0x03E2, 0x03F3},
+               {0x1F00, 0x1F15}, {0x1F18, 0x1F1D}, {0x1F20, 0x1F45}, {0x1F48, 0x1F4D},
+               {0x1F50, 0x1F57}, {0x1F5F, 0x1F7D}, {0x1F80, 0x1FB4}, {0x1FB6, 0x1FBC},
+               {0x1FC2, 0x1FC4}, {0x1FC6, 0x1FCC}, {0x1FD0, 0x1FD3}, {0x1FD6, 0x1FDB},
+               {0x1FE0, 0x1FEC}, {0x1FF2, 0x1FF4}, {0x1FF6, 0x1FFC}, {0x0401, 0x040C},
+               {0x040E, 0x044F}, {0x0451, 0x045C}, {0x045E, 0x0481}, {0x0490, 0x04C4},
+               {0x04C7, 0x04C8}, {0x04CB, 0x04CC}, {0x04D0, 0x04EB}, {0x04EE, 0x04F5},
+               {0x04F8, 0x04F9}, {0x0531, 0x0556}, {0x0561, 0x0587}, {0x05B0, 0x05B9},
+               {0x05BB, 0x05BD}, {0x05C1, 0x05C2}, {0x05D0, 0x05EA}, {0x05F0, 0x05F2},
+               {0x0621, 0x063A}, {0x0640, 0x0652}, {0x0670, 0x06B7}, {0x06BA, 0x06BE},
+               {0x06C0, 0x06CE}, {0x06D0, 0x06DC}, {0x06E5, 0x06E8}, {0x06EA, 0x06ED},
+               {0x0901, 0x0903}, {0x0905, 0x0939}, {0x093E, 0x094D}, {0x0950, 0x0952},
+               {0x0958, 0x0963}, {0x0981, 0x0983}, {0x0985, 0x098C}, {0x098F, 0x0990},
+               {0x0993, 0x09A8}, {0x09AA, 0x09B0}, {0x09B6, 0x09B9}, {0x09BE, 0x09C4},
+               {0x09C7, 0x09C8}, {0x09CB, 0x09CD}, {0x09DC, 0x09DD}, {0x09DF, 0x09E3},
+               {0x09F0, 0x09F1}, {0x0A05, 0x0A0A}, {0x0A0F, 0x0A10}, {0x0A13, 0x0A28},
+               {0x0A2A, 0x0A30}, {0x0A32, 0x0A33}, {0x0A35, 0x0A36}, {0x0A38, 0x0A39},
+               {0x0A3E, 0x0A42}, {0x0A47, 0x0A48}, {0x0A4B, 0x0A4D}, {0x0A59, 0x0A5C},
+               {0x0A81, 0x0A83}, {0x0A85, 0x0A8B}, {0x0A8F, 0x0A91}, {0x0A93, 0x0AA8},
+               {0x0AAA, 0x0AB0}, {0x0AB2, 0x0AB3}, {0x0AB5, 0x0AB9}, {0x0ABD, 0x0AC5},
+               {0x0AC7, 0x0AC9}, {0x0ACB, 0x0ACD}, {0x0B01, 0x0B03}, {0x0B05, 0x0B0C},
+               {0x0B0F, 0x0B10}, {0x0B13, 0x0B28}, {0x0B2A, 0x0B30}, {0x0B32, 0x0B33},
+               {0x0B36, 0x0B39}, {0x0B3E, 0x0B43}, {0x0B47, 0x0B48}, {0x0B4B, 0x0B4D},
+               {0x0B5C, 0x0B5D}, {0x0B5F, 0x0B61}, {0x0B82, 0x0B83}, {0x0B85, 0x0B8A},
+               {0x0B8E, 0x0B90}, {0x0B92, 0x0B95}, {0x0B99, 0x0B9A}, {0x0B9E, 0x0B9F},
+               {0x0BA3, 0x0BA4}, {0x0BA8, 0x0BAA}, {0x0BAE, 0x0BB5}, {0x0BB7, 0x0BB9},
+               {0x0BBE, 0x0BC2}, {0x0BC6, 0x0BC8}, {0x0BCA, 0x0BCD}, {0x0C01, 0x0C03},
+               {0x0C05, 0x0C0C}, {0x0C0E, 0x0C10}, {0x0C12, 0x0C28}, {0x0C2A, 0x0C33},
+               {0x0C35, 0x0C39}, {0x0C3E, 0x0C44}, {0x0C46, 0x0C48}, {0x0C4A, 0x0C4D},
+               {0x0C60, 0x0C61}, {0x0C82, 0x0C83}, {0x0C85, 0x0C8C}, {0x0C8E, 0x0C90},
+               {0x0C92, 0x0CA8}, {0x0CAA, 0x0CB3}, {0x0CB5, 0x0CB9}, {0x0CBE, 0x0CC4},
+               {0x0CC6, 0x0CC8}, {0x0CCA, 0x0CCD}, {0x0CE0, 0x0CE1}, {0x0D02, 0x0D03},
+               {0x0D05, 0x0D0C}, {0x0D0E, 0x0D10}, {0x0D12, 0x0D28}, {0x0D2A, 0x0D39},
+               {0x0D3E, 0x0D43}, {0x0D46, 0x0D48}, {0x0D4A, 0x0D4D}, {0x0D60, 0x0D61},
+               {0x0E01, 0x0E3A}, {0x0E40, 0x0E5B}, {0x0E81, 0x0E82}, {0x0E87, 0x0E88},
+               {0x0E94, 0x0E97}, {0x0E99, 0x0E9F}, {0x0EA1, 0x0EA3}, {0x0EAA, 0x0EAB},
+               {0x0EAD, 0x0EAE}, {0x0EB0, 0x0EB9}, {0x0EBB, 0x0EBD}, {0x0EC0, 0x0EC4},
+               {0x0EC8, 0x0ECD}, {0x0EDC, 0x0EDD}, {0x0F18, 0x0F19}, {0x0F3E, 0x0F47},
+               {0x0F49, 0x0F69}, {0x0F71, 0x0F84}, {0x0F86, 0x0F8B}, {0x0F90, 0x0F95},
+               {0x0F99, 0x0FAD}, {0x0FB1, 0x0FB7}, {0x10A0, 0x10C5}, {0x10D0, 0x10F6},
+               {0x3041, 0x3093}, {0x309B, 0x309C}, {0x30A1, 0x30F6}, {0x30FB, 0x30FC},
+               {0x3105, 0x312C}, {0x4E00, 0x9FA5}, {0xAC00, 0xD7A3}, {0x0660, 0x0669},
+               {0x06F0, 0x06F9}, {0x0966, 0x096F}, {0x09E6, 0x09EF}, {0x0A66, 0x0A6F},
+               {0x0AE6, 0x0AEF}, {0x0B66, 0x0B6F}, {0x0BE7, 0x0BEF}, {0x0C66, 0x0C6F},
+               {0x0CE6, 0x0CEF}, {0x0D66, 0x0D6F}, {0x0E50, 0x0E59}, {0x0ED0, 0x0ED9},
+               {0x0F20, 0x0F33}, {0x02B0, 0x02B8}, {0x02BD, 0x02C1}, {0x02D0, 0x02D1},
+               {0x02E0, 0x02E4}, {0x203F, 0x2040}, {0x210A, 0x2113}, {0x2118, 0x211D},
+               {0x212A, 0x2131}, {0x2133, 0x2138}, {0x2160, 0x2182}, {0x3005, 0x3007},
+               {0x3021, 0x3029},
+       };
+       for (size_t i = 0; i < sizeof(ranges)/sizeof(ranges[0]); ++i) {
+               if (ranges[i][0] <= v && v <= ranges[i][1])
+                       return true;
+       }
+       for (size_t i = 0; i < sizeof(single_chars)/sizeof(single_chars[0]); ++i) {
+               if (v == single_chars[i])
+                       return true;
+       }
+       return false;
+}
+
+static bool is_universal_char_valid_identifier_c11(utf32 const v)
 {
        /* C11 Annex D.1 */
        if (                v == 0x000A8) return true;
@@ -499,14 +599,24 @@ static bool is_universal_char_valid_identifier(utf32 const v)
        return false;
 }
 
-static bool is_universal_char_valid_identifier_start(utf32 const v)
+static bool is_universal_char_valid_identifier(utf32 const v)
+{
+       if (c_mode & _C11)
+               return is_universal_char_valid_identifier_c11(v);
+       return is_universal_char_valid_identifier_c99(v);
+}
+
+static bool is_universal_char_invalid_identifier_start(utf32 const v)
 {
+       if (! (c_mode & _C11))
+               return false;
+
        /* C11 Annex D.2 */
-       if (0x0300 <= v && v <= 0x036F) return false;
-       if (0x1DC0 <= v && v <= 0x1DFF) return false;
-       if (0x20D0 <= v && v <= 0x20FF) return false;
-       if (0xFE20 <= v && v <= 0xFE2F) return false;
-       return true;
+       if (0x0300 <= v && v <= 0x036F) return true;
+       if (0x1DC0 <= v && v <= 0x1DFF) return true;
+       if (0x20D0 <= v && v <= 0x20FF) return true;
+       if (0xFE20 <= v && v <= 0xFE2F) return true;
+       return false;
 }
 
 /**
@@ -588,23 +698,35 @@ string_t make_string(char const *const string)
        return sym_make_string(STRING_ENCODING_CHAR);
 }
 
+static utf32 get_string_encoding_limit(string_encoding_t const enc)
+{
+       switch (enc) {
+       case STRING_ENCODING_CHAR:   return 0xFF;
+       case STRING_ENCODING_CHAR16: return 0xFFFF;
+       case STRING_ENCODING_CHAR32: return 0xFFFFFFFF;
+       case STRING_ENCODING_UTF8:   return 0xFFFFFFFF;
+       case STRING_ENCODING_WIDE:   return 0xFFFFFFFF; // FIXME depends on settings
+       }
+       panic("invalid string encoding");
+}
+
 static void parse_string(utf32 const delimiter, token_kind_t const kind,
                          string_encoding_t const enc,
                          char const *const context)
 {
-       const unsigned start_linenr = input.position.lineno;
-
        eat(delimiter);
 
+       utf32 const limit = get_string_encoding_limit(enc);
        while (true) {
                switch (input.c) {
                case '\\': {
                        if (resolve_escape_sequences) {
                                utf32 const tc = parse_escape_sequence();
+                               if (tc > limit) {
+                                       warningf(WARN_OTHER, &pp_token.base.pos,
+                                                "escape sequence out of range");
+                               }
                                if (enc == STRING_ENCODING_CHAR) {
-                                       if (tc >= 0x100) {
-                                               warningf(WARN_OTHER, &pp_token.base.source_position, "escape sequence out of range");
-                                       }
                                        obstack_1grow(&symbol_obstack, tc);
                                } else {
                                        obstack_grow_utf8(&symbol_obstack, tc);
@@ -619,16 +741,12 @@ static void parse_string(utf32 const delimiter, token_kind_t const kind,
                }
 
                case NEWLINE:
-                       errorf(&pp_token.base.source_position, "newline while parsing %s", context);
+                       errorf(&pp_token.base.pos, "newline while parsing %s", context);
                        break;
 
-               case EOF: {
-                       source_position_t source_position;
-                       source_position.input_name = pp_token.base.source_position.input_name;
-                       source_position.lineno     = start_linenr;
-                       errorf(&source_position, "EOF while parsing %s", context);
+               case EOF:
+                       errorf(&pp_token.base.pos, "EOF while parsing %s", context);
                        goto end_of_string;
-               }
 
                default:
                        if (input.c == delimiter) {
@@ -759,6 +877,67 @@ static void finished_expanding(pp_definition_t *definition)
        current_expansion = parent;
 }
 
+static void grow_string_escaped(struct obstack *obst, const string_t *string, char const *delimiter)
+{
+       char const *prefix = get_string_encoding_prefix(string->encoding);
+       obstack_printf(obst, "%s%s", prefix, delimiter);
+       size_t      size = string->size;
+       const char *str  = string->begin;
+       if (resolve_escape_sequences) {
+               obstack_grow(obst, str, size);
+       } else {
+               for (size_t i = 0; i < size; ++i) {
+                       const char c = str[i];
+                       if (c == '\\' || c == '"')
+                               obstack_1grow(obst, '\\');
+                       obstack_1grow(obst, c);
+               }
+       }
+       obstack_printf(obst, "%s", delimiter);
+}
+
+static void grow_token(struct obstack *obst, const token_t *token)
+{
+       switch (token->kind) {
+       case T_NUMBER:
+               obstack_grow(obst, token->literal.string.begin, token->literal.string.size);
+               break;
+
+       case T_STRING_LITERAL: {
+               char const *const delimiter = resolve_escape_sequences ? "\"" : "\\\"";
+               grow_string_escaped(obst, &token->literal.string, delimiter);
+               break;
+       }
+
+       case T_CHARACTER_CONSTANT:
+               grow_string_escaped(obst, &token->literal.string, "'");
+               break;
+
+       case T_IDENTIFIER:
+       default: {
+               const char *str = token->base.symbol->string;
+               size_t      len = strlen(str);
+               obstack_grow(obst, str, len);
+               break;
+       }
+       }
+}
+
+static void stringify(const pp_definition_t *definition)
+{
+       assert(obstack_object_size(&symbol_obstack) == 0);
+
+       size_t list_len = definition->list_len;
+       for (size_t p = 0; p < list_len; ++p) {
+               const saved_token_t *saved = &definition->token_list[p];
+               if (p > 0 && saved->had_whitespace)
+                       obstack_1grow(&symbol_obstack, ' ');
+               grow_token(&symbol_obstack, &saved->token);
+       }
+       pp_token.kind           = T_STRING_LITERAL;
+       pp_token.literal.string = sym_make_string(STRING_ENCODING_CHAR);
+}
+
 static inline void set_punctuator(token_kind_t const kind)
 {
        pp_token.kind        = kind;
@@ -791,11 +970,22 @@ restart:;
        }
        const saved_token_t *saved = &current_expansion->token_list[pos++];
        pp_token = saved->token;
+       if (pp_token.kind == '#') {
+               if (pos < current_expansion->list_len) {
+                       const saved_token_t *next = &current_expansion->token_list[pos];
+                       if (next->token.kind == T_MACRO_PARAMETER) {
+                               pp_definition_t *def = next->token.macro_parameter.def;
+                               assert(def != NULL && def->is_parameter);
+                               stringify(def);
+                               ++pos;
+                       }
+               }
+       }
 
        if (current_expansion->expand_pos > 0)
                info.had_whitespace = saved->had_whitespace;
-       pp_token.base.source_position = expansion_pos;
-       ++current_expansion->expand_pos;
+       current_expansion->expand_pos = pos;
+       pp_token.base.pos             = expansion_pos;
 
        return true;
 }
@@ -806,13 +996,11 @@ restart:;
  */
 static token_kind_t peek_expansion(void)
 {
-       pp_definition_t *expansion = current_expansion;
-       while (expansion != NULL && expansion->expand_pos >= expansion->list_len) {
-               expansion = expansion->parent_expansion;
+       for (pp_definition_t *e = current_expansion; e; e = e->parent_expansion) {
+               if (e->expand_pos < e->list_len)
+                       return e->token_list[e->expand_pos].token.kind;
        }
-       if (expansion == NULL)
-               return T_EOF;
-       return expansion->token_list[expansion->expand_pos].token.kind;
+       return T_EOF;
 }
 
 static void skip_line_comment(void)
@@ -838,7 +1026,7 @@ static void skip_multiline_comment(void)
 {
        info.had_whitespace = true;
 
-       unsigned start_linenr = input.position.lineno;
+       position_t const start_pos = input.pos;
        while (true) {
                switch (input.c) {
                case '/':
@@ -850,8 +1038,8 @@ static void skip_multiline_comment(void)
                case '*':
                        next_char();
                        if (input.c == '/') {
-                               if (input.position.lineno != input.output_line)
-                                       info.whitespace_at_line_begin = input.position.colno;
+                               if (input.pos.lineno != input.output_line)
+                                       info.whitespace_at_line_begin = input.pos.colno;
                                next_char();
                                return;
                        }
@@ -860,13 +1048,9 @@ static void skip_multiline_comment(void)
                case NEWLINE:
                        break;
 
-               case EOF: {
-                       source_position_t source_position;
-                       source_position.input_name = pp_token.base.source_position.input_name;
-                       source_position.lineno     = start_linenr;
-                       errorf(&source_position, "at end of file while looking for comment end");
+               case EOF:
+                       errorf(&start_pos, "at end of file while looking for comment end");
                        return;
-               }
 
                default:
                        next_char();
@@ -967,6 +1151,17 @@ static inline void eat_token(token_kind_t const kind)
        next_input_token();
 }
 
+static string_encoding_t identify_encoding_prefix(symbol_t *const sym)
+{
+       if (sym == symbol_L) return STRING_ENCODING_WIDE;
+       if (c_mode & _C11) {
+               if (sym == symbol_U)  return STRING_ENCODING_CHAR32;
+               if (sym == symbol_u)  return STRING_ENCODING_CHAR16;
+               if (sym == symbol_u8) return STRING_ENCODING_UTF8;
+       }
+       return STRING_ENCODING_CHAR;
+}
+
 static void parse_symbol(void)
 {
        assert(obstack_object_size(&symbol_obstack) == 0);
@@ -994,12 +1189,12 @@ universal:
                                utf32 const v = parse_universal_char(n);
                                if (!is_universal_char_valid_identifier(v)) {
                                        if (is_universal_char_valid(v)) {
-                                               errorf(&input.position,
+                                               errorf(&input.pos,
                                                           "universal character \\%c%0*X is not valid in an identifier",
                                                           n == 4 ? 'u' : 'U', (int)n, v);
                                        }
-                               } else if (obstack_object_size(&symbol_obstack) == 0 && !is_universal_char_valid_identifier_start(v)) {
-                                       errorf(&input.position,
+                               } else if (obstack_object_size(&symbol_obstack) == 0 && is_universal_char_invalid_identifier_start(v)) {
+                                       errorf(&input.pos,
                                                   "universal character \\%c%0*X is not valid as start of an identifier",
                                                   n == 4 ? 'u' : 'U', (int)n, v);
                                } else if (resolve_escape_sequences) {
@@ -1024,19 +1219,27 @@ end_symbol:
        obstack_1grow(&symbol_obstack, '\0');
        char *string = obstack_finish(&symbol_obstack);
 
-       /* might be a wide string or character constant ( L"string"/L'c' ) */
-       if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
-               obstack_free(&symbol_obstack, string);
-               parse_string_literal(STRING_ENCODING_WIDE);
-               return;
-       } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
-               obstack_free(&symbol_obstack, string);
-               parse_character_constant(STRING_ENCODING_WIDE);
-               return;
-       }
-
        symbol_t *symbol = symbol_table_insert(string);
 
+       /* Might be a prefixed string or character constant: L/U/u/u8"string". */
+       if (input.c == '"') {
+               string_encoding_t const enc = identify_encoding_prefix(symbol);
+               if (enc != STRING_ENCODING_CHAR) {
+                       parse_string_literal(enc);
+                       return;
+               }
+       } else if (input.c == '\'') {
+               string_encoding_t const enc = identify_encoding_prefix(symbol);
+               if (enc != STRING_ENCODING_CHAR) {
+                       if (enc == STRING_ENCODING_UTF8) {
+                               errorf(&pp_token.base.pos,
+                                      "'u8' is not a valid encoding for a chracter constant");
+                       }
+                       parse_character_constant(enc);
+                       return;
+               }
+       }
+
        pp_token.kind        = symbol->ID;
        pp_token.base.symbol = symbol;
 
@@ -1119,8 +1322,8 @@ static void next_input_token(void)
                info.had_whitespace = false;
        }
 restart:
-       pp_token.base.source_position = input.position;
-       pp_token.base.symbol          = NULL;
+       pp_token.base.pos    = input.pos;
+       pp_token.base.symbol = NULL;
 
        switch (input.c) {
        case ' ':
@@ -1300,10 +1503,11 @@ digraph_percentcolon:
                if (input_stack != NULL) {
                        fclose(close_pp_input());
                        pop_restore_input();
-                       fputc('\n', out);
+                       if (out)
+                               fputc('\n', out);
                        if (input.c == (utf32)EOF)
-                               --input.position.lineno;
-                       print_line_directive(&input.position, "2");
+                               --input.pos.lineno;
+                       print_line_directive(&input.pos, "2");
                        goto restart;
                } else {
                        info.at_line_begin = true;
@@ -1324,8 +1528,7 @@ digraph_percentcolon:
        default:
 dollar_sign:
                if (error_on_unknown_chars) {
-                       errorf(&pp_token.base.source_position,
-                              "unknown character '%lc' found\n", input.c);
+                       errorf(&pp_token.base.pos, "unknown character '%lc' found", input.c);
                        next_char();
                        goto restart;
                } else {
@@ -1372,7 +1575,7 @@ static void print_quoted_string(const char *const string)
        fputc('"', out);
 }
 
-static void print_line_directive(const source_position_t *pos, const char *add)
+static void print_line_directive(const position_t *pos, const char *add)
 {
        if (!out)
                return;
@@ -1393,22 +1596,30 @@ static void print_line_directive(const source_position_t *pos, const char *add)
 
 static bool emit_newlines(void)
 {
-       unsigned delta = pp_token.base.source_position.lineno - input.output_line;
+       if (!out)
+               return true;
+
+       unsigned delta = pp_token.base.pos.lineno - input.output_line;
        if (delta == 0)
                return false;
 
        if (delta >= 9) {
                fputc('\n', out);
-               print_line_directive(&pp_token.base.source_position, NULL);
+               print_line_directive(&pp_token.base.pos, NULL);
                fputc('\n', out);
        } else {
                for (unsigned i = 0; i < delta; ++i) {
                        fputc('\n', out);
                }
        }
-       input.output_line = pp_token.base.source_position.lineno;
+       input.output_line = pp_token.base.pos.lineno;
 
-       for (unsigned i = 0; i < info.whitespace_at_line_begin; ++i)
+       unsigned whitespace = info.whitespace_at_line_begin;
+       /* make sure there is at least 1 whitespace before a (macro-expanded)
+        * '#' at line begin. I'm not sure why this is good, but gcc does it. */
+       if (pp_token.kind == '#' && whitespace == 0)
+               ++whitespace;
+       for (unsigned i = 0; i < whitespace; ++i)
                fputc(' ', out);
 
        return true;
@@ -1515,14 +1726,21 @@ static bool pp_definitions_equal(const pp_definition_t *definition1,
        for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
                if (!pp_tokens_equal(&t1->token, &t2->token))
                        return false;
+               if (t1->had_whitespace != t2->had_whitespace)
+                       return false;
        }
        return true;
 }
 
+static void missing_macro_param_error(void)
+{
+       errorf(&pp_token.base.pos, "'#' is not followed by a macro parameter");
+}
+
 static bool is_defineable_token(char const *const context)
 {
        if (info.at_line_begin) {
-               errorf(&pp_token.base.source_position, "unexpected end of line after %s", context);
+               errorf(&pp_token.base.pos, "unexpected end of line after %s", context);
        }
 
        symbol_t *const symbol = pp_token.base.symbol;
@@ -1537,7 +1755,8 @@ dollar_sign:
 
                default:
 no_ident:
-                       errorf(&pp_token.base.source_position, "expected identifier after %s, got %K", context, &pp_token);
+                       errorf(&pp_token.base.pos, "expected identifier after %s, got %K",
+                              context, &pp_token);
                        return false;
                }
        }
@@ -1546,7 +1765,8 @@ no_ident:
        switch (symbol->pp_ID) {
        /* Â§6.10.8:4 */
        case TP_defined:
-               errorf(&pp_token.base.source_position, "%K cannot be used as macro name in %s", &pp_token, context);
+               errorf(&pp_token.base.pos, "%K cannot be used as macro name in %s",
+                      &pp_token, context);
                return false;
 
        default:
@@ -1571,8 +1791,8 @@ static void parse_define_directive(void)
        pp_definition_t *new_definition
                = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
        memset(new_definition, 0, sizeof(new_definition[0]));
-       new_definition->symbol          = symbol;
-       new_definition->source_position = input.position;
+       new_definition->symbol = symbol;
+       new_definition->pos    = input.pos;
 
        /* this is probably the only place where spaces are significant in the
         * lexer (except for the fact that they separate tokens). #define b(x)
@@ -1587,7 +1807,7 @@ static void parse_define_directive(void)
                                new_definition->is_variadic = true;
                                eat_token(T_DOTDOTDOT);
                                if (pp_token.kind != ')') {
-                                       errorf(&input.position,
+                                       errorf(&input.pos,
                                                        "'...' not at end of macro argument list");
                                        goto error_out;
                                }
@@ -1596,9 +1816,9 @@ static void parse_define_directive(void)
                        case T_IDENTIFIER: {
                                pp_definition_t parameter;
                                memset(&parameter, 0, sizeof(parameter));
-                               parameter.source_position = pp_token.base.source_position;
-                               parameter.symbol          = pp_token.base.symbol;
-                               parameter.is_parameter    = true;
+                               parameter.pos          = pp_token.base.pos;
+                               parameter.symbol       = pp_token.base.symbol;
+                               parameter.is_parameter = true;
                                obstack_grow(&pp_obstack, &parameter, sizeof(parameter));
                                eat_token(T_IDENTIFIER);
 
@@ -1608,7 +1828,7 @@ static void parse_define_directive(void)
                                }
 
                                if (pp_token.kind != ')') {
-                                       errorf(&pp_token.base.source_position,
+                                       errorf(&pp_token.base.pos,
                                               "expected ',' or ')' after identifier, got %K",
                                               &pp_token);
                                        goto error_out;
@@ -1621,7 +1841,7 @@ static void parse_define_directive(void)
                                goto finish_argument_list;
 
                        default:
-                               errorf(&pp_token.base.source_position,
+                               errorf(&pp_token.base.pos,
                                       "expected identifier, '...' or ')' in #define argument list, got %K",
                                       &pp_token);
                                goto error_out;
@@ -1635,19 +1855,18 @@ static void parse_define_directive(void)
                        = size / sizeof(new_definition->parameters[0]);
                new_definition->parameters = obstack_finish(&pp_obstack);
                for (size_t i = 0; i < new_definition->n_parameters; ++i) {
-                       pp_definition_t *param    = &new_definition->parameters[i];
-                       symbol_t        *symbol   = param->symbol;
-                       pp_definition_t *previous = symbol->pp_definition;
+                       pp_definition_t *const param     = &new_definition->parameters[i];
+                       symbol_t        *const param_sym = param->symbol;
+                       pp_definition_t *const previous  = param_sym->pp_definition;
                        if (previous != NULL
                            && previous->function_definition == new_definition) {
-                               errorf(&param->source_position,
-                                      "duplicate macro parameter '%Y'", symbol);
+                               errorf(&param->pos, "duplicate macro parameter '%Y'", param_sym);
                                param->symbol = sym_anonymous;
                                continue;
                        }
                        param->parent_expansion    = previous;
                        param->function_definition = new_definition;
-                       symbol->pp_definition      = param;
+                       param_sym->pp_definition   = param;
                }
        } else {
                next_input_token();
@@ -1655,22 +1874,29 @@ static void parse_define_directive(void)
 
        /* construct token list */
        assert(obstack_object_size(&pp_obstack) == 0);
+       bool next_must_be_param = false;
        while (!info.at_line_begin) {
                if (pp_token.kind == T_IDENTIFIER) {
-                       const symbol_t  *symbol     = pp_token.base.symbol;
-                       pp_definition_t *definition = symbol->pp_definition;
+                       pp_definition_t *const definition = pp_token.base.symbol->pp_definition;
                        if (definition != NULL
                            && definition->function_definition == new_definition) {
                            pp_token.kind                = T_MACRO_PARAMETER;
                            pp_token.macro_parameter.def = definition;
                        }
                }
+               if (next_must_be_param && pp_token.kind != T_MACRO_PARAMETER) {
+                       missing_macro_param_error();
+               }
                saved_token_t saved_token;
                saved_token.token = pp_token;
                saved_token.had_whitespace = info.had_whitespace;
                obstack_grow(&pp_obstack, &saved_token, sizeof(saved_token));
+               next_must_be_param
+                       = new_definition->has_parameters && pp_token.kind == '#';
                next_input_token();
        }
+       if (next_must_be_param)
+               missing_macro_param_error();
 
        new_definition->list_len   = obstack_object_size(&pp_obstack)
                / sizeof(new_definition->token_list[0]);
@@ -1678,21 +1904,23 @@ static void parse_define_directive(void)
 
        if (new_definition->has_parameters) {
                for (size_t i = 0; i < new_definition->n_parameters; ++i) {
-                       pp_definition_t *param      = &new_definition->parameters[i];
-                       symbol_t        *symbol     = param->symbol;
-                       if (symbol == sym_anonymous)
+                       pp_definition_t *const param     = &new_definition->parameters[i];
+                       symbol_t        *const param_sym = param->symbol;
+                       if (param_sym == sym_anonymous)
                                continue;
-                       assert(symbol->pp_definition == param);
+                       assert(param_sym->pp_definition == param);
                        assert(param->function_definition == new_definition);
-                       symbol->pp_definition   = param->parent_expansion;
-                       param->parent_expansion = NULL;
+                       param_sym->pp_definition = param->parent_expansion;
+                       param->parent_expansion  = NULL;
                }
        }
 
        pp_definition_t *old_definition = symbol->pp_definition;
        if (old_definition != NULL) {
                if (!pp_definitions_equal(old_definition, new_definition)) {
-                       warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
+                       warningf(WARN_OTHER, &input.pos,
+                                "multiple definition of macro '%Y' (first defined %P)",
+                                symbol, &old_definition->pos);
                } else {
                        /* reuse the old definition */
                        obstack_free(&pp_obstack, new_definition);
@@ -1728,7 +1956,7 @@ static void parse_undef_directive(void)
        next_input_token();
 
        if (!info.at_line_begin) {
-               warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
+               warningf(WARN_OTHER, &input.pos, "extra tokens at end of #undef directive");
        }
        eat_pp_directive();
 }
@@ -1737,36 +1965,39 @@ static void parse_undef_directive(void)
  * They're only allowed behind an #include so they're not recognized
  * by the normal next_preprocessing_token. We handle them as a special
  * exception here */
-static void parse_headername(void)
+static const char *parse_headername(bool *system_include)
 {
-       const source_position_t start_position = input.position;
-       string_t                string         = { NULL, 0, STRING_ENCODING_CHAR };
-       assert(obstack_object_size(&symbol_obstack) == 0);
-
        if (info.at_line_begin) {
                parse_error("expected headername after #include");
-               goto finish_error;
+               return NULL;
        }
 
-       /* check wether we have a "... or <... headername */
+       /* check whether we have a "... or <... headername */
+       position_t pos = input.pos;
        switch (input.c) {
        {
                utf32 delimiter;
-       case '<': delimiter = '>'; goto parse_name;
-       case '"': delimiter = '"'; goto parse_name;
+       case '<': delimiter = '>'; *system_include = true;  goto parse_name;
+       case '"': delimiter = '"'; *system_include = false; goto parse_name;
 parse_name:
+               assert(obstack_object_size(&symbol_obstack) == 0);
                next_char();
                while (true) {
                        switch (input.c) {
                        case NEWLINE:
                        case EOF:
-                               errorf(&pp_token.base.source_position, "header name without closing '%c'", (char)delimiter);
-                               goto finish_error;
+                               {
+                                       char *dummy = obstack_finish(&symbol_obstack);
+                                       obstack_free(&symbol_obstack, dummy);
+                               }
+                               errorf(&pp_token.base.pos,
+                                      "header name without closing '%c'", (char)delimiter);
+                               return NULL;
 
                        default:
                                if (input.c == delimiter) {
                                        next_char();
-                                       goto finished_headername;
+                                       goto finish_headername;
                                } else {
                                        obstack_1grow(&symbol_obstack, (char)input.c);
                                        next_char();
@@ -1778,29 +2009,82 @@ parse_name:
        }
 
        default:
-               /* TODO: do normal pp_token parsing and concatenate results */
-               panic("pp_token concat include not implemented yet");
-       }
+               next_preprocessing_token();
+               if (info.at_line_begin) {
+                       /* TODO: if we are already in the new line then we parsed more than
+                        * wanted. We reuse the token, but could produce following errors
+                        * misbehaviours... */
+                       goto error_invalid_input;
+               }
+               if (pp_token.kind == T_STRING_LITERAL) {
+                       *system_include = false;
+                       return pp_token.literal.string.begin;
+               } else if (pp_token.kind == '<') {
+                       *system_include = true;
+                       assert(obstack_object_size(&pp_obstack) == 0);
+                       while (true) {
+                               next_preprocessing_token();
+                               if (info.at_line_begin) {
+                                       /* TODO: we shouldn't have parsed/expanded something on the
+                                        * next line yet... */
+                                       char *dummy = obstack_finish(&pp_obstack);
+                                       obstack_free(&pp_obstack, dummy);
+                                       goto error_invalid_input;
+                               }
+                               if (pp_token.kind == '>')
+                                       break;
 
-finished_headername:
-       string = sym_make_string(STRING_ENCODING_CHAR);
+                               saved_token_t saved;
+                               saved.token          = pp_token;
+                               saved.had_whitespace = info.had_whitespace;
+                               obstack_grow(&pp_obstack, &saved, sizeof(saved));
+                       }
+                       size_t size = obstack_object_size(&pp_obstack);
+                       assert(size % sizeof(saved_token_t) == 0);
+                       size_t n_tokens = size / sizeof(saved_token_t);
+                       saved_token_t *tokens = obstack_finish(&pp_obstack);
+                       assert(obstack_object_size(&symbol_obstack) == 0);
+                       for (size_t i = 0; i < n_tokens; ++i) {
+                               const saved_token_t *saved = &tokens[i];
+                               if (i > 0 && saved->had_whitespace)
+                                       obstack_1grow(&symbol_obstack, ' ');
+                               grow_token(&symbol_obstack, &saved->token);
+                       }
+                       obstack_free(&pp_obstack, tokens);
+                       goto finish_headername;
+               } else {
+error_invalid_input:
+                       {
+                               char *dummy = obstack_finish(&symbol_obstack);
+                               obstack_free(&symbol_obstack, dummy);
+                       }
 
-finish_error:
-       pp_token.base.source_position = start_position;
-       pp_token.kind                 = T_HEADERNAME;
-       pp_token.literal.string       = string;
+                       errorf(&pp_token.base.pos,
+                              "expected \"FILENAME\" or <FILENAME> after #include");
+                       return NULL;
+               }
+       }
+
+finish_headername:
+       obstack_1grow(&symbol_obstack, '\0');
+       char *const  headername = obstack_finish(&symbol_obstack);
+       const char  *identified = identify_string(headername);
+       pp_token.base.pos = pos;
+       return identified;
 }
 
-static bool do_include(bool const system_include, bool const include_next, char const *const headername)
+static bool do_include(bool const bracket_include, bool const include_next, char const *const headername)
 {
        size_t const        headername_len = strlen(headername);
        searchpath_entry_t *entry;
        if (include_next) {
-               entry = input.path ? input.path->next : searchpath;
+               entry = input.path      ? input.path->next
+                     : bracket_include ? bracket_searchpath.first
+                     : quote_searchpath.first;
        } else {
-               if (!system_include) {
+               if (!bracket_include) {
                        /* put dirname of current input on obstack */
-                       const char *filename   = input.position.input_name;
+                       const char *filename   = input.pos.input_name;
                        const char *last_slash = strrchr(filename, '/');
                        const char *full_name;
                        if (last_slash != NULL) {
@@ -1815,12 +2099,13 @@ static bool do_include(bool const system_include, bool const include_next, char
 
                        FILE *file = fopen(full_name, "r");
                        if (file != NULL) {
-                               switch_pp_input(file, full_name, NULL);
+                               switch_pp_input(file, full_name, NULL, false);
                                return true;
                        }
+                       entry = quote_searchpath.first;
+               } else {
+                       entry = bracket_searchpath.first;
                }
-
-               entry = searchpath;
        }
 
        assert(obstack_object_size(&symbol_obstack) == 0);
@@ -1837,7 +2122,7 @@ static bool do_include(bool const system_include, bool const include_next, char
                FILE *file          = fopen(complete_path, "r");
                if (file != NULL) {
                        const char *filename = identify_string(complete_path);
-                       switch_pp_input(file, filename, entry);
+                       switch_pp_input(file, filename, entry, entry->is_system_path);
                        return true;
                } else {
                        obstack_free(&symbol_obstack, complete_path);
@@ -1854,25 +2139,24 @@ static void parse_include_directive(bool const include_next)
                return;
        }
 
-       /* don't eat the TP_include here!
-        * we need an alternative parsing for the next token */
+       /* do not eat the TP_include, since it would already parse the next token
+        * which needs special handling here. */
        skip_till_newline(true);
-       bool system_include = input.c == '<';
-       parse_headername();
-       string_t headername = pp_token.literal.string;
-       if (headername.begin == NULL) {
+       bool system_include;
+       const char *headername = parse_headername(&system_include);
+       if (headername == NULL) {
                eat_pp_directive();
                return;
        }
 
        bool had_nonwhitespace = skip_till_newline(false);
        if (had_nonwhitespace) {
-               warningf(WARN_OTHER, &pp_token.base.source_position,
+               warningf(WARN_OTHER, &input.pos,
                         "extra tokens at end of #include directive");
        }
 
        if (n_inputs > INCLUDE_LIMIT) {
-               errorf(&pp_token.base.source_position, "#include nested too deeply");
+               errorf(&pp_token.base.pos, "#include nested too deeply");
                /* eat \n or EOF */
                next_input_token();
                return;
@@ -1884,11 +2168,11 @@ static void parse_include_directive(bool const include_next)
        info.at_line_begin            = true;
        emit_newlines();
        push_input();
-       bool res = do_include(system_include, include_next, pp_token.literal.string.begin);
+       bool res = do_include(system_include, include_next, headername);
        if (res) {
                next_input_token();
        } else {
-               errorf(&pp_token.base.source_position, "failed including '%S': %s", &pp_token.literal.string, strerror(errno));
+               errorf(&pp_token.base.pos, "failed including '%s': %s", headername, strerror(errno));
                pop_restore_input();
        }
 }
@@ -1917,9 +2201,9 @@ void check_unclosed_conditionals(void)
                pp_conditional_t *conditional = conditional_stack;
 
                if (conditional->in_else) {
-                       errorf(&conditional->source_position, "unterminated #else");
+                       errorf(&conditional->pos, "unterminated #else");
                } else {
-                       errorf(&conditional->source_position, "unterminated condition");
+                       errorf(&conditional->pos, "unterminated condition");
                }
                pop_conditional();
        }
@@ -1933,35 +2217,33 @@ static void parse_ifdef_ifndef_directive(bool const is_ifdef)
        if (skip_mode) {
                eat_pp_directive();
                pp_conditional_t *conditional = push_conditional();
-               conditional->source_position  = pp_token.base.source_position;
-               conditional->skip             = true;
+               conditional->pos  = pp_token.base.pos;
+               conditional->skip = true;
                return;
        }
 
        if (pp_token.kind != T_IDENTIFIER || info.at_line_begin) {
-               errorf(&pp_token.base.source_position,
-                      "expected identifier after #%s, got %K",
+               errorf(&pp_token.base.pos, "expected identifier after #%s, got %K",
                       is_ifdef ? "ifdef" : "ifndef", &pp_token);
                eat_pp_directive();
 
                /* just take the true case in the hope to avoid further errors */
                condition = true;
        } else {
-               /* evaluate wether we are in true or false case */
+               /* evaluate whether we are in true or false case */
                condition = (bool)pp_token.base.symbol->pp_definition == is_ifdef;
                eat_token(T_IDENTIFIER);
 
                if (!info.at_line_begin) {
-                       errorf(&pp_token.base.source_position,
-                              "extra tokens at end of #%s",
+                       errorf(&pp_token.base.pos, "extra tokens at end of #%s",
                               is_ifdef ? "ifdef" : "ifndef");
                        eat_pp_directive();
                }
        }
 
        pp_conditional_t *conditional = push_conditional();
-       conditional->source_position  = pp_token.base.source_position;
-       conditional->condition        = condition;
+       conditional->pos       = pp_token.base.pos;
+       conditional->condition = condition;
 
        if (!condition) {
                skip_mode = true;
@@ -1974,21 +2256,21 @@ static void parse_else_directive(void)
 
        if (!info.at_line_begin) {
                if (!skip_mode) {
-                       warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
+                       warningf(WARN_OTHER, &pp_token.base.pos, "extra tokens at end of #else");
                }
                eat_pp_directive();
        }
 
        pp_conditional_t *conditional = conditional_stack;
        if (conditional == NULL) {
-               errorf(&pp_token.base.source_position, "#else without prior #if");
+               errorf(&pp_token.base.pos, "#else without prior #if");
                return;
        }
 
        if (conditional->in_else) {
-               errorf(&pp_token.base.source_position,
+               errorf(&pp_token.base.pos,
                       "#else after #else (condition started %P)",
-                      &conditional->source_position);
+                      &conditional->pos);
                skip_mode = true;
                return;
        }
@@ -1997,7 +2279,7 @@ static void parse_else_directive(void)
        if (!conditional->skip) {
                skip_mode = conditional->condition;
        }
-       conditional->source_position = pp_token.base.source_position;
+       conditional->pos = pp_token.base.pos;
 }
 
 static void parse_endif_directive(void)
@@ -2006,14 +2288,14 @@ static void parse_endif_directive(void)
 
        if (!info.at_line_begin) {
                if (!skip_mode) {
-                       warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
+                       warningf(WARN_OTHER, &pp_token.base.pos, "extra tokens at end of #endif");
                }
                eat_pp_directive();
        }
 
        pp_conditional_t *conditional = conditional_stack;
        if (conditional == NULL) {
-               errorf(&pp_token.base.source_position, "#endif without prior #if");
+               errorf(&pp_token.base.pos, "#endif without prior #if");
                return;
        }
 
@@ -2046,7 +2328,7 @@ static void parse_pragma_directive(void)
        }
 
        if (pp_token.kind != T_IDENTIFIER) {
-               warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
+               warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.pos,
                         "expected identifier after #pragma");
                eat_pp_directive();
                return;
@@ -2074,13 +2356,13 @@ static void parse_pragma_directive(void)
                        }
                        if (value == STDC_VALUE_UNKNOWN) {
                                kind = STDC_UNKNOWN;
-                               errorf(&pp_token.base.source_position, "bad STDC pragma argument");
+                               errorf(&pp_token.base.pos, "bad STDC pragma argument");
                        }
                }
        }
        eat_pp_directive();
        if (kind == STDC_UNKNOWN) {
-               warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
+               warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.pos,
                         "encountered unknown #pragma");
        }
 }
@@ -2095,12 +2377,12 @@ static void parse_line_directive(void)
                long const line = strtol(pp_token.literal.string.begin, &end, 0);
                if (*end == '\0') {
                        /* use offset -1 as this is about the next line */
-                       input.position.lineno = line - 1;
+                       input.pos.lineno = line - 1;
                        /* force output of line */
-                       input.output_line = input.position.lineno - 20;
+                       input.output_line = input.pos.lineno - 20;
                } else {
                        if (!skip_mode) {
-                               errorf(&input.position, "'%S' is not a valid line number",
+                               errorf(&input.pos, "'%S' is not a valid line number",
                                           &pp_token.literal.string);
                        }
                }
@@ -2110,8 +2392,8 @@ static void parse_line_directive(void)
        }
        if (pp_token.kind == T_STRING_LITERAL
            && pp_token.literal.string.encoding == STRING_ENCODING_CHAR) {
-               input.position.input_name       = pp_token.literal.string.begin;
-               input.position.is_system_header = false;
+               input.pos.input_name       = pp_token.literal.string.begin;
+               input.pos.is_system_header = false;
                next_input_token();
 
                /* attempt to parse numeric flags as outputted by gcc preprocessor */
@@ -2125,7 +2407,7 @@ static void parse_line_directive(void)
                         * currently we're only interested in "3"
                         */
                        if (streq(pp_token.literal.string.begin, "3")) {
-                               input.position.is_system_header = true;
+                               input.pos.is_system_header = true;
                        }
                        next_input_token();
                }
@@ -2144,7 +2426,7 @@ static void parse_error_directive(void)
        bool const old_resolve_escape_sequences = resolve_escape_sequences;
        resolve_escape_sequences = false;
 
-       source_position_t const pos = pp_token.base.source_position;
+       position_t const pos = pp_token.base.pos;
        do {
                if (info.had_whitespace && obstack_object_size(&pp_obstack) != 0)
                        obstack_1grow(&pp_obstack, ' ');
@@ -2215,7 +2497,7 @@ line_directive:
        } else {
 skip:
                if (!skip_mode) {
-                       errorf(&pp_token.base.source_position, "invalid preprocessing directive #%K", &pp_token);
+                       errorf(&pp_token.base.pos, "invalid preprocessing directive #%K", &pp_token);
                }
                eat_pp_directive();
        }
@@ -2272,7 +2554,7 @@ restart:
 
                                        if (next_token == '(') {
                                                if (current_expansion == NULL)
-                                                       expansion_pos = pp_token.base.source_position;
+                                                       expansion_pos = pp_token.base.pos;
                                                next_preprocessing_token();
                                                assert(pp_token.kind == '(');
 
@@ -2294,7 +2576,7 @@ restart:
                                        }
                                } else {
                                        if (current_expansion == NULL)
-                                               expansion_pos = pp_token.base.source_position;
+                                               expansion_pos = pp_token.base.pos;
                                        start_expanding(pp_definition);
                                        goto restart;
                                }
@@ -2322,7 +2604,7 @@ restart:
                        finish_current_argument();
                        current_call->expand_pos++;
                        if (current_call->expand_pos >= current_call->n_parameters) {
-                               errorf(&pp_token.base.source_position,
+                               errorf(&pp_token.base.pos,
                                           "too many arguments passed for macro '%Y'",
                                           current_call->symbol);
                                current_argument = NULL;
@@ -2355,24 +2637,21 @@ restart:
        }
 }
 
-
-static void prepend_include_path(const char *path)
+void append_include_path(searchpath_t *paths, const char *path)
 {
        searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
-       entry->path = path;
-       entry->next = searchpath;
-       searchpath  = entry;
+       entry->path           = path;
+       entry->is_system_path = paths->is_system_path;
+
+       *paths->anchor = entry;
+       paths->anchor  = &entry->next;
 }
 
-static void setup_include_path(void)
+static void append_env_paths(searchpath_t *paths, const char *envvar)
 {
-       /* built-in paths */
-       prepend_include_path("/usr/include");
-
-       /* parse environment variable */
-       const char *cpath = getenv("CPATH");
-       if (cpath != NULL && *cpath != '\0') {
-               const char *begin = cpath;
+       const char *val = getenv(envvar);
+       if (val != NULL && *val != '\0') {
+               const char *begin = val;
                const char *c;
                do {
                        c = begin;
@@ -2381,35 +2660,60 @@ static void setup_include_path(void)
 
                        size_t len = c-begin;
                        if (len == 0) {
-                               /* for gcc compatibility (Matze: I would expect that
+                               /* use "." for gcc compatibility (Matze: I would expect that
                                 * nothing happens for an empty entry...) */
-                               prepend_include_path(".");
+                               append_include_path(paths, ".");
                        } else {
                                char *const string = obstack_copy0(&config_obstack, begin, len);
-                               prepend_include_path(string);
+                               append_include_path(paths, string);
                        }
 
                        begin = c+1;
                        /* skip : */
                        if (*begin == ':')
                                ++begin;
-               } while(*c != '\0');
+               } while (*c != '\0');
        }
 }
 
+static void append_searchpath(searchpath_t *path, const searchpath_t *append)
+{
+       *path->anchor = append->first;
+}
+
+static void setup_include_path(void)
+{
+       /* built-in paths */
+       append_include_path(&system_searchpath, "/usr/include");
+
+       /* parse environment variable */
+       append_env_paths(&bracket_searchpath, "CPATH");
+       append_env_paths(&system_searchpath,
+                        c_mode & _CXX ? "CPLUS_INCLUDE_PATH" : "C_INCLUDE_PATH");
+
+       /* append system search path to bracket searchpath */
+       append_searchpath(&system_searchpath,  &after_searchpath);
+       append_searchpath(&bracket_searchpath, &system_searchpath);
+       append_searchpath(&quote_searchpath, &bracket_searchpath);
+}
+
 static void input_error(unsigned const delta_lines, unsigned const delta_cols, char const *const message)
 {
-       source_position_t pos = pp_token.base.source_position;
+       position_t pos = pp_token.base.pos;
        pos.lineno += delta_lines;
        pos.colno  += delta_cols;
        errorf(&pos, "%s", message);
 }
 
+void init_include_paths(void)
+{
+       obstack_init(&config_obstack);
+}
+
 void init_preprocessor(void)
 {
        init_symbols();
 
-       obstack_init(&config_obstack);
        obstack_init(&pp_obstack);
        obstack_init(&input_obstack);
        strset_init(&stringset);
@@ -2432,6 +2736,7 @@ int pptest_main(int argc, char **argv);
 int pptest_main(int argc, char **argv)
 {
        init_symbol_table();
+       init_include_paths();
        init_preprocessor();
        init_tokens();
 
@@ -2444,7 +2749,7 @@ int pptest_main(int argc, char **argv)
        for (int i = 1; i < argc; ++i) {
                const char *opt = argv[i];
                if (streq(opt, "-I")) {
-                       prepend_include_path(argv[++i]);
+                       append_include_path(&bracket_searchpath, argv[++i]);
                        continue;
                } else if (streq(opt, "-E")) {
                        /* ignore */
@@ -2484,7 +2789,7 @@ int pptest_main(int argc, char **argv)
                fprintf(stderr, "Couldn't open input '%s'\n", filename);
                return 1;
        }
-       switch_pp_input(file, filename, NULL);
+       switch_pp_input(file, filename, NULL, false);
 
        for (;;) {
                next_preprocessing_token();