| 1 | /* GDB symbol table format definitions. |
| 2 | Copyright (C) 1987, 1988 Free Software Foundation, Inc. |
| 3 | |
| 4 | This file is part of GNU CC. |
| 5 | |
| 6 | GNU CC is free software; you can redistribute it and/or modify |
| 7 | it under the terms of the GNU General Public License as published by |
| 8 | the Free Software Foundation; either version 1, or (at your option) |
| 9 | any later version. |
| 10 | |
| 11 | GNU CC is distributed in the hope that it will be useful, |
| 12 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | GNU General Public License for more details. |
| 15 | |
| 16 | You should have received a copy of the GNU General Public License |
| 17 | along with GNU CC; see the file COPYING. If not, write to |
| 18 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ |
| 19 | |
| 20 | /* Format of GDB symbol table data. |
| 21 | There is one symbol segment for each source file or |
| 22 | independant compilation. These segments are simply concatenated |
| 23 | to form the GDB symbol table. A zero word where the beginning |
| 24 | of a segment is expected indicates there are no more segments. |
| 25 | |
| 26 | Format of a symbol segment: |
| 27 | |
| 28 | The symbol segment begins with a word containing 1 |
| 29 | if it is in the format described here. Other formats may |
| 30 | be designed, with other code numbers. |
| 31 | |
| 32 | The segment contains many objects which point at each other. |
| 33 | The pointers are offsets in bytes from the beginning of the segment. |
| 34 | Thus, each segment can be loaded into core and its pointers relocated |
| 35 | to make valid in-core pointers. |
| 36 | |
| 37 | All the data objects in the segment can be found indirectly from |
| 38 | one of them, the root object, of type `struct symbol_root'. |
| 39 | It appears at the beginning of the segment. |
| 40 | |
| 41 | The total size of the segment, in bytes, appears as the `length' |
| 42 | field of this object. This size includes the size of the |
| 43 | root object. |
| 44 | |
| 45 | All the object data types are defined here to contain pointer types |
| 46 | appropriate for in-core use on a relocated symbol segment. |
| 47 | Casts to and from type int are required for working with |
| 48 | unrelocated symbol segments such as are found in the file. |
| 49 | |
| 50 | The ldsymaddr word is filled in by the loader to contain |
| 51 | the offset (in bytes) within the ld symbol table |
| 52 | of the first nonglobal symbol from this compilation. |
| 53 | This makes it possible to match those symbols |
| 54 | (which contain line number information) reliably with |
| 55 | the segment they go with. |
| 56 | |
| 57 | Core addresses within the program that appear in the symbol segment |
| 58 | are not relocated by the loader. They are inserted by the assembler |
| 59 | and apply to addresses as output by the assembler, so GDB must |
| 60 | relocate them when it loads the symbol segment. It gets the information |
| 61 | on how to relocate from the textrel, datarel, bssrel, databeg and bssbeg |
| 62 | words of the root object. |
| 63 | |
| 64 | The words textrel, datarel and bssrel |
| 65 | are filled in by ld with the amounts to relocate within-the-file |
| 66 | text, data and bss addresses by; databeg and bssbeg can be |
| 67 | used to tell which kind of relocation an address needs. */ |
| 68 | |
| 69 | enum language {language_c}; |
| 70 | |
| 71 | struct symbol_root |
| 72 | { |
| 73 | int format; /* Data format version */ |
| 74 | int length; /* # bytes in this symbol segment */ |
| 75 | int ldsymoff; /* Offset in ld symtab of this file's syms */ |
| 76 | int textrel; /* Relocation for text addresses */ |
| 77 | int datarel; /* Relocation for data addresses */ |
| 78 | int bssrel; /* Relocation for bss addresses */ |
| 79 | char *filename; /* Name of main source file compiled */ |
| 80 | char *filedir; /* Name of directory it was reached from */ |
| 81 | struct blockvector *blockvector; /* Vector of all symbol-naming blocks */ |
| 82 | struct typevector *typevector; /* Vector of all data types */ |
| 83 | enum language language; /* Code identifying the language used */ |
| 84 | char *version; /* Version info. Not fully specified */ |
| 85 | char *compilation; /* Compilation info. Not fully specified */ |
| 86 | int databeg; /* Address within the file of data start */ |
| 87 | int bssbeg; /* Address within the file of bss start */ |
| 88 | struct sourcevector *sourcevector; /* Vector of line-number info */ |
| 89 | }; |
| 90 | \f |
| 91 | /* All data types of symbols in the compiled program |
| 92 | are represented by `struct type' objects. |
| 93 | All of these objects are pointed to by the typevector. |
| 94 | The type vector may have empty slots that contain zero. */ |
| 95 | |
| 96 | struct typevector |
| 97 | { |
| 98 | int length; /* Number of types described */ |
| 99 | struct type *type[1]; |
| 100 | }; |
| 101 | |
| 102 | /* Different kinds of data types are distinguished by the `code' field. */ |
| 103 | |
| 104 | enum type_code |
| 105 | { |
| 106 | TYPE_CODE_UNDEF, /* Not used; catches errors */ |
| 107 | TYPE_CODE_PTR, /* Pointer type */ |
| 108 | TYPE_CODE_ARRAY, /* Array type, lower bound zero */ |
| 109 | TYPE_CODE_STRUCT, /* C struct or Pascal record */ |
| 110 | TYPE_CODE_UNION, /* C union or Pascal variant part */ |
| 111 | TYPE_CODE_ENUM, /* Enumeration type */ |
| 112 | TYPE_CODE_FUNC, /* Function type */ |
| 113 | TYPE_CODE_INT, /* Integer type */ |
| 114 | TYPE_CODE_FLT, /* Floating type */ |
| 115 | TYPE_CODE_VOID, /* Void type (values zero length) */ |
| 116 | TYPE_CODE_SET, /* Pascal sets */ |
| 117 | TYPE_CODE_RANGE, /* Range (integers within spec'd bounds) */ |
| 118 | TYPE_CODE_PASCAL_ARRAY, /* Array with explicit type of index */ |
| 119 | }; |
| 120 | |
| 121 | /* This appears in a type's flags word for an unsigned integer type. */ |
| 122 | #define TYPE_FLAG_UNSIGNED 1 |
| 123 | |
| 124 | /* Other flag bits are used with GDB. */ |
| 125 | |
| 126 | struct type |
| 127 | { |
| 128 | /* Code for kind of type */ |
| 129 | enum type_code code; |
| 130 | /* Name of this type, or zero if none. |
| 131 | This is used for printing only. |
| 132 | Type names specified as input are defined by symbols. */ |
| 133 | char *name; |
| 134 | /* Length in bytes of storage for a value of this type */ |
| 135 | int length; |
| 136 | /* For a pointer type, describes the type of object pointed to. |
| 137 | For an array type, describes the type of the elements. |
| 138 | For a function type, describes the type of the value. |
| 139 | Unused otherwise. */ |
| 140 | struct type *target_type; |
| 141 | /* Type that is a pointer to this type. |
| 142 | Zero if no such pointer-to type is known yet. |
| 143 | The debugger may add the address of such a type |
| 144 | if it has to construct one later. */ |
| 145 | struct type *pointer_type; |
| 146 | /* Type that is a function returning this type. |
| 147 | Zero if no such function type is known here. |
| 148 | The debugger may add the address of such a type |
| 149 | if it has to construct one later. */ |
| 150 | struct type *function_type; |
| 151 | /* Flags about this type. */ |
| 152 | short flags; |
| 153 | /* Number of fields described for this type */ |
| 154 | short nfields; |
| 155 | /* For structure and union types, a description of each field. |
| 156 | For set and pascal array types, there is one "field", |
| 157 | whose type is the domain type of the set or array. |
| 158 | For range types, there are two "fields", |
| 159 | the minimum and maximum values (both inclusive). |
| 160 | For enum types, each possible value is described by one "field". |
| 161 | For range types, there are two "fields", that record constant values |
| 162 | (inclusive) for the minimum and maximum. |
| 163 | |
| 164 | Using a pointer to a separate array of fields |
| 165 | allows all types to have the same size, which is useful |
| 166 | because we can allocate the space for a type before |
| 167 | we know what to put in it. */ |
| 168 | struct field |
| 169 | { |
| 170 | /* Position of this field, counting in bits from start of |
| 171 | containing structure. For a function type, this is the |
| 172 | position in the argument list of this argument. |
| 173 | For a range bound or enum value, this is the value itself. */ |
| 174 | int bitpos; |
| 175 | /* Size of this field, in bits, or zero if not packed. |
| 176 | For an unpacked field, the field's type's length |
| 177 | says how many bytes the field occupies. */ |
| 178 | int bitsize; |
| 179 | /* In a struct or enum type, type of this field. |
| 180 | In a function type, type of this argument. |
| 181 | In an array type, the domain-type of the array. */ |
| 182 | struct type *type; |
| 183 | /* Name of field, value or argument. |
| 184 | Zero for range bounds and array domains. */ |
| 185 | char *name; |
| 186 | } *fields; |
| 187 | }; |
| 188 | \f |
| 189 | /* All of the name-scope contours of the program |
| 190 | are represented by `struct block' objects. |
| 191 | All of these objects are pointed to by the blockvector. |
| 192 | |
| 193 | Each block represents one name scope. |
| 194 | Each lexical context has its own block. |
| 195 | |
| 196 | The first two blocks in the blockvector are special. |
| 197 | The first one contains all the symbols defined in this compilation |
| 198 | whose scope is the entire program linked together. |
| 199 | The second one contains all the symbols whose scope is the |
| 200 | entire compilation excluding other separate compilations. |
| 201 | In C, these correspond to global symbols and static symbols. |
| 202 | |
| 203 | Each block records a range of core addresses for the code that |
| 204 | is in the scope of the block. The first two special blocks |
| 205 | give, for the range of code, the entire range of code produced |
| 206 | by the compilation that the symbol segment belongs to. |
| 207 | |
| 208 | The blocks appear in the blockvector |
| 209 | in order of increasing starting-address, |
| 210 | and, within that, in order of decreasing ending-address. |
| 211 | |
| 212 | This implies that within the body of one function |
| 213 | the blocks appear in the order of a depth-first tree walk. */ |
| 214 | |
| 215 | struct blockvector |
| 216 | { |
| 217 | /* Number of blocks in the list. */ |
| 218 | int nblocks; |
| 219 | /* The blocks themselves. */ |
| 220 | struct block *block[1]; |
| 221 | }; |
| 222 | |
| 223 | struct block |
| 224 | { |
| 225 | /* Addresses in the executable code that are in this block. |
| 226 | Note: in an unrelocated symbol segment in a file, |
| 227 | these are always zero. They can be filled in from the |
| 228 | N_LBRAC and N_RBRAC symbols in the loader symbol table. */ |
| 229 | int startaddr, endaddr; |
| 230 | /* The symbol that names this block, |
| 231 | if the block is the body of a function; |
| 232 | otherwise, zero. |
| 233 | Note: In an unrelocated symbol segment in an object file, |
| 234 | this field may be zero even when the block has a name. |
| 235 | That is because the block is output before the name |
| 236 | (since the name resides in a higher block). |
| 237 | Since the symbol does point to the block (as its value), |
| 238 | it is possible to find the block and set its name properly. */ |
| 239 | struct symbol *function; |
| 240 | /* The `struct block' for the containing block, or 0 if none. */ |
| 241 | /* Note that in an unrelocated symbol segment in an object file |
| 242 | this pointer may be zero when the correct value should be |
| 243 | the second special block (for symbols whose scope is one compilation). |
| 244 | This is because the compiler ouptuts the special blocks at the |
| 245 | very end, after the other blocks. */ |
| 246 | struct block *superblock; |
| 247 | /* Number of local symbols. */ |
| 248 | int nsyms; |
| 249 | /* The symbols. */ |
| 250 | struct symbol *sym[1]; |
| 251 | }; |
| 252 | \f |
| 253 | /* Represent one symbol name; a variable, constant, function or typedef. */ |
| 254 | |
| 255 | /* Different name spaces for symbols. Looking up a symbol specifies |
| 256 | a namespace and ignores symbol definitions in other name spaces. |
| 257 | |
| 258 | VAR_NAMESPACE is the usual namespace. |
| 259 | In C, this contains variables, function names, typedef names |
| 260 | and enum type values. |
| 261 | |
| 262 | STRUCT_NAMESPACE is used in C to hold struct, union and enum type names. |
| 263 | Thus, if `struct foo' is used in a C program, |
| 264 | it produces a symbol named `foo' in the STRUCT_NAMESPACE. |
| 265 | |
| 266 | LABEL_NAMESPACE may be used for names of labels (for gotos); |
| 267 | currently it is not used and labels are not recorded at all. */ |
| 268 | |
| 269 | /* For a non-global symbol allocated statically, |
| 270 | the correct core address cannot be determined by the compiler. |
| 271 | The compiler puts an index number into the symbol's value field. |
| 272 | This index number can be matched with the "desc" field of |
| 273 | an entry in the loader symbol table. */ |
| 274 | |
| 275 | enum namespace |
| 276 | { |
| 277 | UNDEF_NAMESPACE, VAR_NAMESPACE, STRUCT_NAMESPACE, LABEL_NAMESPACE, |
| 278 | }; |
| 279 | |
| 280 | /* An address-class says where to find the value of the symbol in core. */ |
| 281 | |
| 282 | enum address_class |
| 283 | { |
| 284 | LOC_UNDEF, /* Not used; catches errors */ |
| 285 | LOC_CONST, /* Value is constant int */ |
| 286 | LOC_STATIC, /* Value is at fixed address */ |
| 287 | LOC_REGISTER, /* Value is in register */ |
| 288 | LOC_ARG, /* Value is at spec'd position in arglist */ |
| 289 | LOC_LOCAL, /* Value is at spec'd pos in stack frame */ |
| 290 | LOC_TYPEDEF, /* Value not used; definition in SYMBOL_TYPE |
| 291 | Symbols in the namespace STRUCT_NAMESPACE |
| 292 | all have this class. */ |
| 293 | LOC_LABEL, /* Value is address in the code */ |
| 294 | LOC_BLOCK, /* Value is address of a `struct block'. |
| 295 | Function names have this class. */ |
| 296 | LOC_EXTERNAL, /* Value is at address not in this compilation. |
| 297 | This is used for .comm symbols |
| 298 | and for extern symbols within functions. |
| 299 | Inside GDB, this is changed to LOC_STATIC once the |
| 300 | real address is obtained from a loader symbol. */ |
| 301 | LOC_CONST_BYTES /* Value is a constant byte-sequence. */ |
| 302 | }; |
| 303 | |
| 304 | struct symbol |
| 305 | { |
| 306 | /* Symbol name */ |
| 307 | char *name; |
| 308 | /* Name space code. */ |
| 309 | enum namespace namespace; |
| 310 | /* Address class */ |
| 311 | enum address_class class; |
| 312 | /* Data type of value */ |
| 313 | struct type *type; |
| 314 | /* constant value, or address if static, or register number, |
| 315 | or offset in arguments, or offset in stack frame. */ |
| 316 | union |
| 317 | { |
| 318 | long value; |
| 319 | struct block *block; /* for LOC_BLOCK */ |
| 320 | char *bytes; /* for LOC_CONST_BYTES */ |
| 321 | } |
| 322 | value; |
| 323 | }; |
| 324 | \f |
| 325 | /* Source-file information. |
| 326 | This describes the relation between source files and line numbers |
| 327 | and addresses in the program text. */ |
| 328 | |
| 329 | struct sourcevector |
| 330 | { |
| 331 | int length; /* Number of source files described */ |
| 332 | struct source *source[1]; /* Descriptions of the files */ |
| 333 | }; |
| 334 | |
| 335 | /* Line number and address of one line. */ |
| 336 | |
| 337 | struct line |
| 338 | { |
| 339 | int linenum; |
| 340 | int address; |
| 341 | }; |
| 342 | |
| 343 | /* All the information on one source file. */ |
| 344 | |
| 345 | struct source |
| 346 | { |
| 347 | char *name; /* Name of file */ |
| 348 | int nlines; /* Number of lines that follow */ |
| 349 | struct line lines[1]; /* Information on each line */ |
| 350 | }; |