|
|
1.1.1.2 ! root 1: /* Protoize program - Original version by Ron Guilmette at MCC. 1.1 root 2: 3: Copyright (C) 1989, 1992 Free Software Foundation, Inc. 4: 5: This file is part of GNU CC. 6: 7: GNU CC is free software; you can redistribute it and/or modify 8: it under the terms of the GNU General Public License as published by 9: the Free Software Foundation; either version 2, or (at your option) 10: any later version. 11: 12: GNU CC is distributed in the hope that it will be useful, 13: but WITHOUT ANY WARRANTY; without even the implied warranty of 14: MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 15: GNU General Public License for more details. 16: 17: You should have received a copy of the GNU General Public License 18: along with GNU CC; see the file COPYING. If not, write to 19: the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ 20: 21: /* Any reasonable C++ compiler should have all of the same features 22: as __STDC__ plus more, so make sure that __STDC__ is defined if 23: __cplusplus is defined. */ 24: 25: #if defined(__cplusplus) && !defined(__STDC__) 26: #define __STDC__ 1 27: #endif /* defined(__cplusplus) && !defined(__STDC__) */ 28: 29: #if defined(__GNUC__) || defined (__GNUG__) 30: #define VOLATILE volatile 31: #else 32: #define VOLATILE 33: #endif 34: 35: #ifndef __STDC__ 36: #define const 37: #endif 38: 39: #include "config.h" 40: 41: #ifdef POSIX /* We should be able to define _POSIX_SOURCE unconditionally, 42: but some systems respond in buggy ways to it, 43: including Sunos 4.1.1. Which we don't classify as POSIX. */ 44: /* In case this is a POSIX system with an ANSI C compiler, 45: ask for definition of all POSIX facilities. */ 46: #define _POSIX_SOURCE 47: #endif 48: 49: #if 0 50: /* Users are not supposed to use _POSIX_SOURCE to say the 51: system is a POSIX system. That is not what _POSIX_SOURCE means! -- rms */ 52: /* If the user asked for POSIX via _POSIX_SOURCE, turn on POSIX code. */ 53: #if defined(_POSIX_SOURCE) && !defined(POSIX) 54: #define POSIX 55: #endif 56: #endif /* 0 */ 57: 58: #include <stdio.h> 59: #include <ctype.h> 60: #include <errno.h> 61: #include <sys/types.h> 62: #include <sys/stat.h> 63: #include <sys/dir.h> 64: #if ! defined (USG) || defined (SVR4) 65: #include <sys/wait.h> 66: #endif 67: #include <setjmp.h> 68: #include "gvarargs.h" 69: 1.1.1.2 ! root 70: /* Include getopt.h for the sake of getopt_long. ! 71: We don't need the declaration of getopt, and it could conflict ! 72: with something from a system header file, so effectively nullify that. */ ! 73: #define getopt getopt_loser ! 74: #include "getopt.h" 1.1 root 75: 76: extern int errno; 77: extern char *sys_errlist[]; 78: extern char *version_string; 79: 80: /* Systems which are compatible only with POSIX 1003.1-1988 (but *not* 81: with POSIX 1003.1-1990), e.g. Ultrix 4.2, might not have 82: const qualifiers in the prototypes in the system include files. 83: Unfortunately, this can lead to GCC issuing lots of warnings for 84: calls to the following functions. To eliminate these warnings we 85: provide the following #defines. */ 86: 87: #define my_access(file,flag) access((char *)file, flag) 88: #define my_stat(file,pkt) stat((char *)file, pkt) 89: #define my_execvp(prog,argv) execvp((char *)prog, (char **)argv) 90: #define my_link(file1, file2) link((char *)file1, (char *)file2) 91: #define my_unlink(file) unlink((char *)file) 92: #define my_open(file, mode, flag) open((char *)file, mode, flag) 93: #define my_chmod(file, mode) chmod((char *)file, mode) 94: 1.1.1.2 ! root 95: char *getpwd (); 1.1 root 96: 97: /* Aliases for pointers to void. 98: These were made to facilitate compilation with other compilers. */ 99: 100: #ifdef __STDC__ 101: typedef void * pointer_type; 102: typedef const void * const_pointer_type; 103: #else 104: typedef char * pointer_type; 105: typedef char * const_pointer_type; 106: #endif 107: 108: #if defined(POSIX) 109: 110: #include <stdlib.h> 111: #include <unistd.h> 112: #include <signal.h> 113: #include <fcntl.h> 114: #include <string.h> 115: 116: #else /* !defined(POSIX) */ 117: 118: #define R_OK 4 /* Test for Read permission */ 119: #define W_OK 2 /* Test for Write permission */ 120: #define X_OK 1 /* Test for eXecute permission */ 121: #define F_OK 0 /* Test for existence of File */ 122: 123: #define O_RDONLY 0 124: #define O_WRONLY 1 125: 126: /* Declaring stat or __flsbuf with a prototype 127: causes conflicts with system headers on some systems. */ 128: 129: #ifndef abort 130: extern VOLATILE void abort (); 131: #endif 132: extern int kill (); 133: extern int creat (); 134: #if 0 /* These conflict with stdio.h on some systems. */ 135: extern int fprintf (FILE *, const char *, ...); 136: extern int printf (const char *, ...); 137: #endif /* 0 */ 138: extern void exit (); 139: extern pointer_type malloc (); 140: extern pointer_type realloc (); 141: extern void free (); 142: extern int read (); 143: extern int write (); 144: extern int close (); 145: extern int fflush (); 146: extern int atoi (); 147: extern int puts (); 148: extern int fputs (); 149: extern int fputc (); 150: #if 0 /* Causes trouble on some systems that define setjmp as a macro. */ 151: extern int setjmp (); 152: extern void longjmp (); 153: #endif 154: 155: extern char * strcat (); 156: extern int strcmp (); 157: extern char * strcpy (); 158: #if 0 /* size_t from sys/types.h may fail to match GCC. 159: If so, we would get a warning from this. 160: So do without the prototype. */ 161: extern size_t strlen (const char *); 162: #endif 163: extern int strncmp (); 164: extern char * strncpy (); 165: extern char * strrchr (); 166: 1.1.1.2 ! root 167: /* Fork is not declared because the declaration caused a conflict ! 168: on the HPPA. */ 1.1 root 169: #if !(defined (USG) || defined (VMS)) 170: #define fork vfork 1.1.1.2 ! root 171: #endif /* (defined (USG) || defined (VMS)) */ 1.1 root 172: 173: #endif /* !defined (POSIX) */ 174: 175: /* Look for these where the `const' qualifier is intentionally cast aside. */ 176: 177: #define NONCONST 178: 179: /* Define a STRINGIFY macro that's right for ANSI or traditional C. */ 180: 181: #ifdef __STDC__ 182: #define STRINGIFY(STRING) #STRING 183: #else 184: #define STRINGIFY(STRING) "STRING" 185: #endif 186: 187: /* POSIX systems will not have definitions for WIFEXITED or WEXITSTATUS. 188: Define them correctly and so that they work for all environments. */ 189: 190: #undef WIFEXITED 191: #define WIFEXITED(status_word) ((*((int *)&status_word) & 0xff) == 0x00) 192: 193: #undef WEXITSTATUS 194: #define WEXITSTATUS(status_word) ((*((int *)&status_word) & 0xff00) >> 8) 195: 196: /* Define a default place to find the SYSCALLS.X file. */ 197: 198: #ifndef STD_PROTO_DIR 199: #define STD_PROTO_DIR "/usr/local/lib" 200: #endif /* !defined (STD_PROTO_DIR) */ 201: 202: /* Suffix of aux_info files. */ 203: 204: static const char * const aux_info_suffix = ".X"; 205: 206: /* String to attach to filenames for saved versions of original files. */ 207: 208: static const char * const save_suffix = ".save"; 209: 210: #ifndef UNPROTOIZE 211: 212: /* File name of the file which contains descriptions of standard system 213: routines. Note that we never actually do anything with this file per se, 214: but we do read in its corresponding aux_info file. */ 215: 1.1.1.2 ! root 216: static const char syscalls_filename[] = "SYSCALLS.c"; 1.1 root 217: 218: /* Default place to find the above file. */ 219: 220: static const char * const default_syscalls_dir = STD_PROTO_DIR; 221: 222: /* Variable to hold the complete absolutized filename of the SYSCALLS.c.X 223: file. */ 224: 225: static char * syscalls_absolute_filename; 226: 227: #endif /* !defined (UNPROTOIZE) */ 228: 229: /* Type of the structure that holds information about macro unexpansions. */ 230: 231: struct unexpansion_struct { 232: const char *expanded; 233: const char *contracted; 234: }; 235: typedef struct unexpansion_struct unexpansion; 236: 237: /* A table of conversions that may need to be made for some (stupid) older 238: operating systems where these types are preprocessor macros rather than 239: typedefs (as they really ought to be). 240: 241: WARNING: The contracted forms must be as small (or smaller) as the 242: expanded forms, or else havoc will ensue. */ 243: 244: static const unexpansion unexpansions[] = { 245: { "struct _iobuf", "FILE" }, 246: { 0, 0 } 247: }; 248: 249: /* The number of "primary" slots in the hash tables for filenames and for 250: function names. This can be as big or as small as you like, except that 251: it must be a power of two. */ 252: 253: #define HASH_TABLE_SIZE (1 << 9) 254: 255: /* Bit mask to use when computing hash values. */ 256: 257: static const int hash_mask = (HASH_TABLE_SIZE - 1); 258: 259: /* Make a table of default system include directories 260: just as it is done in cccp.c. */ 261: 262: #ifndef STANDARD_INCLUDE_DIR 263: #define STANDARD_INCLUDE_DIR "/usr/include" 264: #endif 265: 266: #ifndef LOCAL_INCLUDE_DIR 267: #define LOCAL_INCLUDE_DIR "/usr/local/include" 268: #endif 269: 270: struct default_include { char *fname; int cplusplus; } include_defaults[] 271: #ifdef INCLUDE_DEFAULTS 272: = INCLUDE_DEFAULTS; 273: #else 274: = { 275: /* Pick up GNU C++ specific include files. */ 276: { GPLUSPLUS_INCLUDE_DIR, 1}, 277: { GCC_INCLUDE_DIR, 0}, 278: #ifdef CROSS_COMPILE 279: /* For cross-compilation, this dir name is generated 280: automatically in Makefile.in. */ 281: { CROSS_INCLUDE_DIR, 0 }, 282: #else /* not CROSS_COMPILE */ 283: { LOCAL_INCLUDE_DIR, 0}, 284: /* Some systems have an extra dir of include files. */ 285: #ifdef SYSTEM_INCLUDE_DIR 286: { SYSTEM_INCLUDE_DIR, 0}, 287: #endif 288: { STANDARD_INCLUDE_DIR, 0}, 289: #endif /* not CROSS_COMPILE */ 290: { 0, 0} 291: }; 292: #endif /* no INCLUDE_DEFAULTS */ 293: 294: /* Datatype for lists of directories or filenames. */ 295: struct string_list 296: { 297: char *name; 298: struct string_list *next; 299: }; 300: 301: /* List of directories in which files should be converted. */ 302: 303: struct string_list *directory_list; 304: 305: /* List of file names which should not be converted. 306: A file is excluded if the end of its name, following a /, 307: matches one of the names in this list. */ 308: 309: struct string_list *exclude_list; 310: 311: /* The name of the other style of variable-number-of-parameters functions 312: (i.e. the style that we want to leave unconverted because we don't yet 313: know how to convert them to this style. This string is used in warning 314: messages. */ 315: 316: /* Also define here the string that we can search for in the parameter lists 317: taken from the .X files which will unambiguously indicate that we have 318: found a varargs style function. */ 319: 320: #ifdef UNPROTOIZE 321: static const char * const other_var_style = "stdarg"; 322: #else /* !defined (UNPROTOIZE) */ 323: static const char * const other_var_style = "varargs"; 324: /* Note that this is a string containing the expansion of va_alist. 325: But in `main' we discard all but the first token. */ 326: static const char *varargs_style_indicator = STRINGIFY (va_alist); 327: #endif /* !defined (UNPROTOIZE) */ 328: 329: /* The following two types are used to create hash tables. In this program, 330: there are two hash tables which are used to store and quickly lookup two 331: different classes of strings. The first type of strings stored in the 332: first hash table are absolute filenames of files which protoize needs to 333: know about. The second type of strings (stored in the second hash table) 334: are function names. It is this second class of strings which really 335: inspired the use of the hash tables, because there may be a lot of them. */ 336: 337: typedef struct hash_table_entry_struct hash_table_entry; 338: 339: /* Do some typedefs so that we don't have to write "struct" so often. */ 340: 341: typedef struct def_dec_info_struct def_dec_info; 342: typedef struct file_info_struct file_info; 343: typedef struct f_list_chain_item_struct f_list_chain_item; 344: 345: /* In the struct below, note that the "_info" field has two different uses 346: depending on the type of hash table we are in (i.e. either the filenames 347: hash table or the function names hash table). In the filenames hash table 348: the info fields of the entries point to the file_info struct which is 349: associated with each filename (1 per filename). In the function names 350: hash table, the info field points to the head of a singly linked list of 351: def_dec_info entries which are all defs or decs of the function whose 352: name is pointed to by the "symbol" field. Keeping all of the defs/decs 353: for a given function name on a special list specifically for that function 354: name makes it quick and easy to find out all of the important information 355: about a given (named) function. */ 356: 357: struct hash_table_entry_struct { 358: hash_table_entry * hash_next; /* -> to secondary entries */ 359: const char * symbol; /* -> to the hashed string */ 360: union { 361: const def_dec_info * _ddip; 362: file_info * _fip; 363: } _info; 364: }; 365: #define ddip _info._ddip 366: #define fip _info._fip 367: 368: /* Define a type specifically for our two hash tables. */ 369: 370: typedef hash_table_entry hash_table[HASH_TABLE_SIZE]; 371: 372: /* The following struct holds all of the important information about any 373: single filename (e.g. file) which we need to know about. */ 374: 375: struct file_info_struct { 376: const hash_table_entry * hash_entry; /* -> to associated hash entry */ 377: const def_dec_info * defs_decs; /* -> to chain of defs/decs */ 378: time_t mtime; /* Time of last modification. */ 379: }; 380: 381: /* Due to the possibility that functions may return pointers to functions, 382: (which may themselves have their own parameter lists) and due to the 383: fact that returned pointers-to-functions may be of type "pointer-to- 384: function-returning-pointer-to-function" (ad nauseum) we have to keep 385: an entire chain of ANSI style formal parameter lists for each function. 386: 387: Normally, for any given function, there will only be one formals list 388: on the chain, but you never know. 389: 390: Note that the head of each chain of formals lists is pointed to by the 391: `f_list_chain' field of the corresponding def_dec_info record. 392: 393: For any given chain, the item at the head of the chain is the *leftmost* 394: parameter list seen in the actual C language function declaration. If 395: there are other members of the chain, then these are linked in left-to-right 396: order from the head of the chain. */ 397: 398: struct f_list_chain_item_struct { 399: const f_list_chain_item * chain_next; /* -> to next item on chain */ 400: const char * formals_list; /* -> to formals list string */ 401: }; 402: 403: /* The following struct holds all of the important information about any 404: single function definition or declaration which we need to know about. 405: Note that for unprotoize we don't need to know very much because we 406: never even create records for stuff that we don't intend to convert 407: (like for instance defs and decs which are already in old K&R format 408: and "implicit" function declarations). */ 409: 410: struct def_dec_info_struct { 411: const def_dec_info * next_in_file; /* -> to rest of chain for file */ 412: file_info * file; /* -> file_info for containing file */ 413: int line; /* source line number of def/dec */ 414: const char * ansi_decl; /* -> left end of ansi decl */ 415: hash_table_entry * hash_entry; /* -> hash entry for function name */ 416: unsigned int is_func_def; /* = 0 means this is a declaration */ 417: const def_dec_info * next_for_func; /* -> to rest of chain for func name */ 418: unsigned int f_list_count; /* count of formals lists we expect */ 419: char prototyped; /* = 0 means already prototyped */ 420: #ifndef UNPROTOIZE 421: const f_list_chain_item * f_list_chain; /* -> chain of formals lists */ 422: const def_dec_info * definition; /* -> def/dec containing related def */ 423: char is_static; /* = 0 means visiblilty is "extern" */ 424: char is_implicit; /* != 0 for implicit func decl's */ 425: char written; /* != 0 means written for implicit */ 426: #else /* !defined (UNPROTOIZE) */ 427: const char * formal_names; /* -> to list of names of formals */ 428: const char * formal_decls; /* -> to string of formal declartions */ 429: #endif /* !defined (UNPROTOIZE) */ 430: }; 431: 432: /* Pointer to the tail component of the filename by which this program was 433: invoked. Used everywhere in error and warning messages. */ 434: 435: static const char *pname; 436: 437: /* Error counter. Will be non-zero if we should give up at the next convenient 438: stopping point. */ 439: 440: static int errors = 0; 441: 442: /* Option flags. */ 443: /* ??? These comments should say what the flag mean as well as the options 444: that set them. */ 445: 446: /* File name to use for running gcc. Allows GCC 2 to be named 447: something other than gcc. */ 448: static char *compiler_file_name = "gcc"; 449: 450: static int version_flag = 0; /* Print our version number. */ 451: static int quiet_flag = 0; /* Don't print messages normally. */ 452: static int nochange_flag = 0; /* Don't convert, just say what files 453: we would have converted. */ 454: static int nosave_flag = 0; /* Don't save the old version. */ 455: static int keep_flag = 0; /* Don't delete the .X files. */ 456: static const char ** compile_params = 0; /* Option string for gcc. */ 457: #ifdef UNPROTOIZE 458: static const char *indent_string = " "; /* Indentation for newly 459: inserted parm decls. */ 460: #else /* !defined (UNPROTOIZE) */ 461: static int local_flag = 0; /* Insert new local decls (when?). */ 462: static int global_flag = 0; /* set by -g option */ 463: static int cplusplus_flag = 0; /* Rename converted files to *.C. */ 464: static const char* nondefault_syscalls_dir = 0; /* Dir to look for 465: SYSCALLS.c.X in. */ 466: #endif /* !defined (UNPROTOIZE) */ 467: 1.1.1.2 ! root 468: /* An index into the compile_params array where we should insert the source ! 469: file name when we are ready to exec the C compiler. A zero value indicates 1.1 root 470: that we have not yet called munge_compile_params. */ 471: 1.1.1.2 ! root 472: static int input_file_name_index = 0; ! 473: ! 474: /* An index into the compile_params array where we should insert the filename ! 475: for the aux info file, when we run the C compiler. */ ! 476: static int aux_info_file_name_index = 0; 1.1 root 477: 478: /* Count of command line arguments which were "filename" arguments. */ 479: 480: static int n_base_source_files = 0; 481: 482: /* Points to a malloc'ed list of pointers to all of the filenames of base 483: source files which were specified on the command line. */ 484: 485: static const char **base_source_filenames; 486: 487: /* Line number of the line within the current aux_info file that we 488: are currently processing. Used for error messages in case the prototypes 489: info file is corrupted somehow. */ 490: 491: static int current_aux_info_lineno; 492: 493: /* Pointer to the name of the source file currently being converted. */ 494: 495: static const char *convert_filename; 496: 497: /* Pointer to relative root string (taken from aux_info file) which indicates 498: where directory the user was in when he did the compilation step that 499: produced the containing aux_info file. */ 500: 501: static const char *invocation_filename; 502: 503: /* Pointer to the base of the input buffer that holds the original text for the 504: source file currently being converted. */ 505: 506: static const char *orig_text_base; 507: 508: /* Pointer to the byte just beyond the end of the input buffer that holds the 509: original text for the source file currently being converted. */ 510: 511: static const char *orig_text_limit; 512: 513: /* Pointer to the base of the input buffer that holds the cleaned text for the 514: source file currently being converted. */ 515: 516: static const char *clean_text_base; 517: 518: /* Pointer to the byte just beyond the end of the input buffer that holds the 519: cleaned text for the source file currently being converted. */ 520: 521: static const char *clean_text_limit; 522: 523: /* Pointer to the last byte in the cleaned text buffer that we have already 524: (virtually) copied to the output buffer (or decided to ignore). */ 525: 526: static const char * clean_read_ptr; 527: 528: /* Pointer to the base of the output buffer that holds the replacement text 529: for the source file currently being converted. */ 530: 531: static char *repl_text_base; 532: 533: /* Pointer to the byte just beyond the end of the output buffer that holds the 534: replacement text for the source file currently being converted. */ 535: 536: static char *repl_text_limit; 537: 538: /* Pointer to the last byte which has been stored into the output buffer. 539: The next byte to be stored should be stored just past where this points 540: to. */ 541: 542: static char * repl_write_ptr; 543: 544: /* Pointer into the cleaned text buffer for the source file we are currently 545: converting. This points to the first character of the line that we last 546: did a "seek_to_line" to (see below). */ 547: 548: static const char *last_known_line_start; 549: 550: /* Number of the line (in the cleaned text buffer) that we last did a 551: "seek_to_line" to. Will be one if we just read a new source file 552: into the cleaned text buffer. */ 553: 554: static int last_known_line_number; 555: 556: /* The filenames hash table. */ 557: 558: static hash_table filename_primary; 559: 560: /* The function names hash table. */ 561: 562: static hash_table function_name_primary; 563: 564: /* The place to keep the recovery address which is used only in cases where 565: we get hopelessly confused by something in the cleaned original text. */ 566: 567: static jmp_buf source_confusion_recovery; 568: 569: /* A pointer to the current directory filename (used by abspath). */ 570: 571: static char *cwd_buffer; 572: 573: /* A place to save the read pointer until we are sure that an individual 574: attempt at editing will succeed. */ 575: 576: static const char * saved_clean_read_ptr; 577: 578: /* A place to save the write pointer until we are sure that an individual 579: attempt at editing will succeed. */ 580: 581: static char * saved_repl_write_ptr; 582: 583: /* Forward declaration. */ 584: 585: static const char *shortpath (); 586: 587: /* Allocate some space, but check that the allocation was successful. */ 588: /* alloca.c uses this, so don't make it static. */ 589: 590: pointer_type 591: xmalloc (byte_count) 592: size_t byte_count; 593: { 594: pointer_type rv; 595: 596: rv = malloc (byte_count); 597: if (rv == NULL) 598: { 599: fprintf (stderr, "\n%s: fatal error: can't allocate %u more bytes of memory\n", 600: pname, byte_count); 601: exit (1); 602: return 0; /* avoid warnings */ 603: } 604: else 605: return rv; 606: } 607: 608: /* Reallocate some space, but check that the reallocation was successful. */ 609: 610: pointer_type 611: xrealloc (old_space, byte_count) 612: pointer_type old_space; 613: size_t byte_count; 614: { 615: pointer_type rv; 616: 617: rv = realloc (old_space, byte_count); 618: if (rv == NULL) 619: { 620: fprintf (stderr, "\n%s: fatal error: can't allocate %u more bytes of memory\n", 621: pname, byte_count); 622: exit (1); 623: return 0; /* avoid warnings */ 624: } 625: else 626: return rv; 627: } 628: 629: /* Deallocate the area pointed to by an arbitrary pointer, but first, strip 630: the `const' qualifier from it and also make sure that the pointer value 631: is non-null. */ 632: 633: void 634: xfree (p) 635: const_pointer_type p; 636: { 637: if (p) 638: free ((NONCONST pointer_type) p); 639: } 640: 641: /* Make a copy of a string INPUT with size SIZE. */ 642: 643: static char * 644: savestring (input, size) 645: const char *input; 646: int size; 647: { 648: char *output = (char *) xmalloc (size + 1); 649: strcpy (output, input); 650: return output; 651: } 652: 653: /* More 'friendly' abort that prints the line and file. 654: config.h can #define abort fancy_abort if you like that sort of thing. */ 655: 656: void 657: fancy_abort () 658: { 659: fprintf (stderr, "%s: internal abort\n", pname); 660: exit (1); 661: } 662: 663: /* Make a duplicate of a given string in a newly allocated area. */ 664: 665: static char * 666: dupstr (s) 667: const char *s; 668: { 669: return strcpy ((char *) xmalloc (strlen (s) + 1), s); 670: } 671: 672: /* Make a duplicate of the first N bytes of a given string in a newly 673: allocated area. */ 674: 675: static char * 676: dupnstr (s, n) 677: const char *s; 678: size_t n; 679: { 680: char *ret_val = strncpy ((char *) xmalloc (n + 1), s, n); 681: 682: ret_val[n] = '\0'; 683: return ret_val; 684: } 685: 686: /* Return a pointer to the first occurance of s2 within s1 or NULL if s2 687: does not occur within s1. Assume neither s1 nor s2 are null pointers. */ 688: 689: static const char * 690: substr (s1, s2) 691: const char *s1; 692: const char *const s2; 693: { 694: for (; *s1 ; s1++) 695: { 696: const char *p1; 697: const char *p2; 698: int c; 699: 700: for (p1 = s1, p2 = s2; c = *p2; p1++, p2++) 701: if (*p1 != c) 702: goto outer; 703: return s1; 704: outer: 705: ; 706: } 707: return 0; 708: } 709: 710: /* Get setup to recover in case the edit we are about to do goes awry. */ 711: 712: void 713: save_pointers () 714: { 715: saved_clean_read_ptr = clean_read_ptr; 716: saved_repl_write_ptr = repl_write_ptr; 717: } 718: 719: /* Call this routine to recover our previous state whenever something looks 720: too confusing in the source code we are trying to edit. */ 721: 722: void 723: restore_pointers () 724: { 725: clean_read_ptr = saved_clean_read_ptr; 726: repl_write_ptr = saved_repl_write_ptr; 727: } 728: 729: /* Return true if the given character is a legal identifier character. */ 730: 731: static int 732: is_id_char (ch) 733: char ch; 734: { 735: return (isalnum (ch) || (ch == '_') || (ch == '$')); 736: } 737: 738: /* Give a message indicating the proper way to invoke this program and then 739: exit with non-zero status. */ 740: 741: static void 742: usage () 743: { 744: #ifdef UNPROTOIZE 745: fprintf (stderr, "%s: usage '%s [ -VqfnkN ] [ -i <istring> ] [ filename ... ]'\n", 746: pname, pname); 747: #else /* !defined (UNPROTOIZE) */ 748: fprintf (stderr, "%s: usage '%s [ -VqfnkNlgC ] [ -B <diname> ] [ filename ... ]'\n", 749: pname, pname); 750: #endif /* !defined (UNPROTOIZE) */ 751: exit (1); 752: } 753: 754: /* Return true if the given filename (assumed to be an absolute filename) 755: designates a file residing anywhere beneath any one of the "system" 756: include directories. */ 757: 758: static int 759: in_system_include_dir (path) 760: const char *path; 761: { 762: struct default_include *p; 763: 764: if (path[0] != '/') 765: abort (); /* Must be an absolutized filename. */ 766: 767: for (p = include_defaults; p->fname; p++) 768: if (!strncmp (path, p->fname, strlen (p->fname)) 769: && path[strlen (p->fname)] == '/') 770: return 1; 771: return 0; 772: } 773: 774: #if 0 775: /* Return true if the given filename designates a file that the user has 776: read access to and for which the user has write access to the containing 777: directory. */ 778: 779: static int 780: file_could_be_converted (const char *path) 781: { 782: char *const dir_name = (char *) alloca (strlen (path) + 1); 783: 784: if (my_access (path, R_OK)) 785: return 0; 786: 787: { 788: char *dir_last_slash; 789: 790: strcpy (dir_name, path); 791: dir_last_slash = strrchr (dir_name, '/'); 792: if (dir_last_slash) 793: *dir_last_slash = '\0'; 794: else 795: abort (); /* Should have been an absolutized filename. */ 796: } 797: 798: if (my_access (path, W_OK)) 799: return 0; 800: 801: return 1; 802: } 803: 804: /* Return true if the given filename designates a file that we are allowed 805: to modify. Files which we should not attempt to modify are (a) "system" 806: include files, and (b) files which the user doesn't have write access to, 807: and (c) files which reside in directories which the user doesn't have 808: write access to. Unless requested to be quiet, give warnings about 809: files that we will not try to convert for one reason or another. An 810: exception is made for "system" include files, which we never try to 811: convert and for which we don't issue the usual warnings. */ 812: 813: static int 814: file_normally_convertable (const char *path) 815: { 816: char *const dir_name = alloca (strlen (path) + 1); 817: 818: if (in_system_include_dir (path)) 819: return 0; 820: 821: { 822: char *dir_last_slash; 823: 824: strcpy (dir_name, path); 825: dir_last_slash = strrchr (dir_name, '/'); 826: if (dir_last_slash) 827: *dir_last_slash = '\0'; 828: else 829: abort (); /* Should have been an absolutized filename. */ 830: } 831: 832: if (my_access (path, R_OK)) 833: { 834: if (!quiet_flag) 835: fprintf (stderr, "%s: warning: no read access for file `%s'\n", 836: pname, shortpath (NULL, path)); 837: return 0; 838: } 839: 840: if (my_access (path, W_OK)) 841: { 842: if (!quiet_flag) 843: fprintf (stderr, "%s: warning: no write access for file `%s'\n", 844: pname, shortpath (NULL, path)); 845: return 0; 846: } 847: 848: if (my_access (dir_name, W_OK)) 849: { 850: if (!quiet_flag) 851: fprintf (stderr, "%s: warning: no write access for dir containing `%s'\n", 852: pname, shortpath (NULL, path)); 853: return 0; 854: } 855: 856: return 1; 857: } 858: #endif /* 0 */ 859: 860: #ifndef UNPROTOIZE 861: 862: /* Return true if the given file_info struct refers to the special SYSCALLS.c.X 863: file. Return false otherwise. */ 864: 865: static int 866: is_syscalls_file (fi_p) 867: const file_info *fi_p; 868: { 1.1.1.2 ! root 869: char const *f = fi_p->hash_entry->symbol; ! 870: size_t fl = strlen (f), sysl = sizeof (syscalls_filename) - 1; ! 871: return sysl <= fl && strcmp (f + fl - sysl, syscalls_filename) == 0; 1.1 root 872: } 873: 874: #endif /* !defined (UNPROTOIZE) */ 875: 876: /* Check to see if this file will need to have anything done to it on this 877: run. If there is nothing in the given file which both needs conversion 878: and for which we have the necessary stuff to do the conversion, return 879: false. Otherwise, return true. 880: 881: Note that (for protoize) it is only valid to call this function *after* 882: the connections between declarations and definitions have all been made 883: by connect_defs_and_decs. */ 884: 885: static int 886: needs_to_be_converted (file_p) 887: const file_info *file_p; 888: { 889: const def_dec_info *ddp; 890: 891: #ifndef UNPROTOIZE 892: 893: if (is_syscalls_file (file_p)) 894: return 0; 895: 896: #endif /* !defined (UNPROTOIZE) */ 897: 898: for (ddp = file_p->defs_decs; ddp; ddp = ddp->next_in_file) 899: 900: if ( 901: 902: #ifndef UNPROTOIZE 903: 904: /* ... and if we a protoizing and this function is in old style ... */ 905: !ddp->prototyped 906: /* ... and if this a definition or is a decl with an associated def ... */ 907: && (ddp->is_func_def || (!ddp->is_func_def && ddp->definition)) 908: 909: #else /* defined (UNPROTOIZE) */ 910: 911: /* ... and if we are unprotoizing and this function is in new style ... */ 912: ddp->prototyped 913: 914: #endif /* defined (UNPROTOIZE) */ 915: ) 916: /* ... then the containing file needs converting. */ 917: return -1; 918: return 0; 919: } 920: 921: /* Return 1 if the file name NAME is in a directory 922: that should be converted. */ 923: 924: static int 925: directory_specified_p (name) 926: const char *name; 927: { 928: struct string_list *p; 929: 930: for (p = directory_list; p; p = p->next) 931: if (!strncmp (name, p->name, strlen (p->name)) 932: && name[strlen (p->name)] == '/') 933: { 934: const char *q = name + strlen (p->name) + 1; 935: 936: /* If there are more slashes, it's in a subdir, so 937: this match doesn't count. */ 938: while (*q) 939: if (*q++ == '/') 940: goto lose; 941: return 1; 942: 943: lose: ; 944: } 945: 946: return 0; 947: } 948: 949: /* Return 1 if the file named NAME should be excluded from conversion. */ 950: 951: static int 952: file_excluded_p (name) 953: const char *name; 954: { 955: struct string_list *p; 956: int len = strlen (name); 957: 958: for (p = exclude_list; p; p = p->next) 959: if (!strcmp (name + len - strlen (p->name), p->name) 960: && name[len - strlen (p->name) - 1] == '/') 961: return 1; 962: 963: return 0; 964: } 965: 966: /* Construct a new element of a string_list. 967: STRING is the new element value, and REST holds the remaining elements. */ 968: 969: static struct string_list * 970: string_list_cons (string, rest) 971: char *string; 972: struct string_list *rest; 973: { 974: struct string_list *temp 975: = (struct string_list *) xmalloc (sizeof (struct string_list)); 976: 977: temp->next = rest; 978: temp->name = string; 979: return temp; 980: } 981: 982: /* ??? The GNU convention for mentioning function args in its comments 983: is to capitalize them. So change "hash_tab_p" to HASH_TAB_P below. 984: Likewise for all the other functions. */ 985: 986: /* Given a hash table, apply some function to each node in the table. The 987: table to traverse is given as the "hash_tab_p" argument, and the 988: function to be applied to each node in the table is given as "func" 989: argument. */ 990: 991: static void 992: visit_each_hash_node (hash_tab_p, func) 993: const hash_table_entry *hash_tab_p; 994: void (*func)(); 995: { 996: const hash_table_entry *primary; 997: 998: for (primary = hash_tab_p; primary < &hash_tab_p[HASH_TABLE_SIZE]; primary++) 999: if (primary->symbol) 1000: { 1001: hash_table_entry *second; 1002: 1003: (*func)(primary); 1004: for (second = primary->hash_next; second; second = second->hash_next) 1005: (*func) (second); 1006: } 1007: } 1008: 1009: /* Initialize all of the fields of a new hash table entry, pointed 1010: to by the "p" parameter. Note that the space to hold the entry 1011: is assumed to have already been allocated before this routine is 1012: called. */ 1013: 1014: static hash_table_entry * 1015: add_symbol (p, s) 1016: hash_table_entry *p; 1017: const char *s; 1018: { 1019: p->hash_next = NULL; 1020: p->symbol = dupstr (s); 1021: p->ddip = NULL; 1022: p->fip = NULL; 1023: return p; 1024: } 1025: 1026: /* Look for a particular function name or filename in the particular 1027: hash table indicated by "hash_tab_p". If the name is not in the 1028: given hash table, add it. Either way, return a pointer to the 1029: hash table entry for the given name. */ 1030: 1031: static hash_table_entry * 1032: lookup (hash_tab_p, search_symbol) 1033: hash_table_entry *hash_tab_p; 1034: const char *search_symbol; 1035: { 1036: int hash_value = 0; 1037: const char *search_symbol_char_p = search_symbol; 1038: hash_table_entry *p; 1039: 1040: while (*search_symbol_char_p) 1041: hash_value += *search_symbol_char_p++; 1042: hash_value &= hash_mask; 1043: p = &hash_tab_p[hash_value]; 1044: if (! p->symbol) 1045: return add_symbol (p, search_symbol); 1046: if (!strcmp (p->symbol, search_symbol)) 1047: return p; 1048: while (p->hash_next) 1049: { 1050: p = p->hash_next; 1051: if (!strcmp (p->symbol, search_symbol)) 1052: return p; 1053: } 1054: p->hash_next = (hash_table_entry *) xmalloc (sizeof (hash_table_entry)); 1055: p = p->hash_next; 1056: return add_symbol (p, search_symbol); 1057: } 1058: 1059: /* Throw a def/dec record on the junk heap. 1060: 1061: Also, since we are not using this record anymore, free up all of the 1062: stuff it pointed to. */ 1063: 1064: static void 1065: free_def_dec (p) 1066: def_dec_info *p; 1067: { 1068: xfree (p->ansi_decl); 1069: 1070: #ifndef UNPROTOIZE 1071: { 1072: const f_list_chain_item * curr; 1073: const f_list_chain_item * next; 1074: 1075: for (curr = p->f_list_chain; curr; curr = next) 1076: { 1077: next = curr->chain_next; 1078: xfree (curr); 1079: } 1080: } 1081: #endif /* !defined (UNPROTOIZE) */ 1082: 1083: xfree (p); 1084: } 1085: 1086: /* Unexpand as many macro symbol as we can find. 1087: 1088: If the given line must be unexpanded, make a copy of it in the heap and 1089: return a pointer to the unexpanded copy. Otherwise return NULL. */ 1090: 1091: static char * 1092: unexpand_if_needed (aux_info_line) 1093: const char *aux_info_line; 1094: { 1095: static char *line_buf = 0; 1096: static int line_buf_size = 0; 1097: const unexpansion* unexp_p; 1098: int got_unexpanded = 0; 1099: const char *s; 1100: char *copy_p = line_buf; 1101: 1102: if (line_buf == 0) 1103: { 1104: line_buf_size = 1024; 1105: line_buf = (char *) xmalloc (line_buf_size); 1106: } 1107: 1108: copy_p = line_buf; 1109: 1110: /* Make a copy of the input string in line_buf, expanding as necessary. */ 1111: 1112: for (s = aux_info_line; *s != '\n'; ) 1113: { 1114: for (unexp_p = unexpansions; unexp_p->expanded; unexp_p++) 1115: { 1116: const char *in_p = unexp_p->expanded; 1117: size_t len = strlen (in_p); 1118: 1119: if (*s == *in_p && !strncmp (s, in_p, len) && !is_id_char (s[len])) 1120: { 1121: int size = strlen (unexp_p->contracted); 1122: got_unexpanded = 1; 1123: if (copy_p + size - line_buf >= line_buf_size) 1124: { 1125: int offset = copy_p - line_buf; 1126: line_buf_size *= 2; 1127: line_buf_size += size; 1128: line_buf = (char *) xrealloc (line_buf, line_buf_size); 1129: copy_p = line_buf + offset; 1130: } 1131: strcpy (copy_p, unexp_p->contracted); 1132: copy_p += size; 1133: 1134: /* Assume the there will not be another replacement required 1135: within the text just replaced. */ 1136: 1137: s += len; 1138: goto continue_outer; 1139: } 1140: } 1141: if (copy_p - line_buf == line_buf_size) 1142: { 1143: int offset = copy_p - line_buf; 1144: line_buf_size *= 2; 1145: line_buf = (char *) xrealloc (line_buf, line_buf_size); 1146: copy_p = line_buf + offset; 1147: } 1148: *copy_p++ = *s++; 1149: continue_outer: ; 1150: } 1151: if (copy_p + 2 - line_buf >= line_buf_size) 1152: { 1153: int offset = copy_p - line_buf; 1154: line_buf_size *= 2; 1155: line_buf = (char *) xrealloc (line_buf, line_buf_size); 1156: copy_p = line_buf + offset; 1157: } 1158: *copy_p++ = '\n'; 1159: *copy_p++ = '\0'; 1160: 1161: return (got_unexpanded ? dupstr (line_buf) : 0); 1162: } 1163: 1164: /* Return the absolutized filename for the given relative 1165: filename. Note that if that filename is already absolute, it may 1166: still be returned in a modified form because this routine also 1167: eliminates redundant slashes and single dots and eliminates double 1168: dots to get a shortest possible filename from the given input 1169: filename. The absolutization of relative filenames is made by 1170: assuming that the given filename is to be taken as relative to 1171: the first argument (cwd) or to the current directory if cwd is 1172: NULL. */ 1173: 1174: static char * 1175: abspath (cwd, rel_filename) 1176: const char *cwd; 1177: const char *rel_filename; 1178: { 1179: /* Setup the current working directory as needed. */ 1180: const char *cwd2 = (cwd) ? cwd : cwd_buffer; 1181: char *const abs_buffer 1.1.1.2 ! root 1182: = (char *) alloca (strlen (cwd2) + strlen (rel_filename) + 2); 1.1 root 1183: char *endp = abs_buffer; 1184: char *outp, *inp; 1185: 1.1.1.2 ! root 1186: /* Copy the filename (possibly preceded by the current working 1.1 root 1187: directory name) into the absolutization buffer. */ 1188: 1189: { 1190: const char *src_p; 1191: 1192: if (rel_filename[0] != '/') 1193: { 1194: src_p = cwd2; 1195: while (*endp++ = *src_p++) 1196: continue; 1197: *(endp-1) = '/'; /* overwrite null */ 1198: } 1199: src_p = rel_filename; 1200: while (*endp++ = *src_p++) 1201: continue; 1202: } 1203: 1204: /* Now make a copy of abs_buffer into abs_buffer, shortening the 1205: filename (by taking out slashes and dots) as we go. */ 1206: 1207: outp = inp = abs_buffer; 1208: *outp++ = *inp++; /* copy first slash */ 1.1.1.2 ! root 1209: #ifdef apollo ! 1210: if (inp[0] == '/') ! 1211: *outp++ = *inp++; /* copy second slash */ ! 1212: #endif 1.1 root 1213: for (;;) 1214: { 1215: if (!inp[0]) 1216: break; 1217: else if (inp[0] == '/' && outp[-1] == '/') 1218: { 1219: inp++; 1220: continue; 1221: } 1222: else if (inp[0] == '.' && outp[-1] == '/') 1223: { 1224: if (!inp[1]) 1225: break; 1226: else if (inp[1] == '/') 1227: { 1228: inp += 2; 1229: continue; 1230: } 1231: else if ((inp[1] == '.') && (inp[2] == 0 || inp[2] == '/')) 1232: { 1233: inp += (inp[2] == '/') ? 3 : 2; 1234: outp -= 2; 1235: while (outp >= abs_buffer && *outp != '/') 1236: outp--; 1237: if (outp < abs_buffer) 1238: { 1239: /* Catch cases like /.. where we try to backup to a 1240: point above the absolute root of the logical file 1241: system. */ 1242: 1243: fprintf (stderr, "%s: invalid file name: %s\n", 1244: pname, rel_filename); 1245: exit (1); 1246: } 1247: *++outp = '\0'; 1248: continue; 1249: } 1250: } 1251: *outp++ = *inp++; 1252: } 1253: 1254: /* On exit, make sure that there is a trailing null, and make sure that 1255: the last character of the returned string is *not* a slash. */ 1256: 1257: *outp = '\0'; 1258: if (outp[-1] == '/') 1259: *--outp = '\0'; 1260: 1261: /* Make a copy (in the heap) of the stuff left in the absolutization 1262: buffer and return a pointer to the copy. */ 1263: 1264: return dupstr (abs_buffer); 1265: } 1266: 1267: /* Given a filename (and possibly a directory name from which the filename 1268: is relative) return a string which is the shortest possible 1269: equivalent for the corresponding full (absolutized) filename. The 1270: shortest possible equivalent may be constructed by converting the 1271: absolutized filename to be a relative filename (i.e. relative to 1272: the actual current working directory). However if a relative filename 1273: is longer, then the full absolute filename is returned. 1274: 1275: KNOWN BUG: 1276: 1277: Note that "simple-minded" conversion of any given type of filename (either 1278: relative or absolute) may not result in a valid equivalent filename if any 1279: subpart of the original filename is actually a symbolic link. */ 1280: 1281: static const char * 1282: shortpath (cwd, filename) 1283: const char *cwd; 1284: const char *filename; 1285: { 1286: char *rel_buffer; 1287: char *rel_buf_p; 1288: char *cwd_p = cwd_buffer; 1289: char *path_p; 1290: int unmatched_slash_count = 0; 1.1.1.2 ! root 1291: size_t filename_len = strlen (filename); 1.1 root 1292: 1293: path_p = abspath (cwd, filename); 1.1.1.2 ! root 1294: rel_buf_p = rel_buffer = (char *) xmalloc (filename_len); 1.1 root 1295: 1296: while (*cwd_p && (*cwd_p == *path_p)) 1297: { 1298: cwd_p++; 1299: path_p++; 1300: } 1.1.1.2 ! root 1301: if (!*cwd_p && (!*path_p || *path_p == '/')) /* whole pwd matched */ 1.1 root 1302: { 1303: if (!*path_p) /* input *is* the current path! */ 1304: return "."; 1305: else 1306: return ++path_p; 1307: } 1308: else 1309: { 1310: if (*path_p) 1311: { 1312: --cwd_p; 1313: --path_p; 1314: while (*cwd_p != '/') /* backup to last slash */ 1315: { 1316: --cwd_p; 1317: --path_p; 1318: } 1319: cwd_p++; 1320: path_p++; 1321: unmatched_slash_count++; 1322: } 1323: while (*cwd_p) 1324: if (*cwd_p++ == '/') 1325: unmatched_slash_count++; 1326: while (unmatched_slash_count--) 1327: { 1.1.1.2 ! root 1328: if (rel_buffer + filename_len <= rel_buf_p + 3) ! 1329: return filename; 1.1 root 1330: *rel_buf_p++ = '.'; 1331: *rel_buf_p++ = '.'; 1332: *rel_buf_p++ = '/'; 1333: } 1.1.1.2 ! root 1334: ! 1335: do ! 1336: { ! 1337: if (rel_buffer + filename_len <= rel_buf_p) ! 1338: return filename; ! 1339: } ! 1340: while (*rel_buf_p++ = *path_p++); ! 1341: 1.1 root 1342: --rel_buf_p; 1343: if (*(rel_buf_p-1) == '/') 1344: *--rel_buf_p = '\0'; 1345: return rel_buffer; 1346: } 1347: } 1348: 1349: /* Lookup the given filename in the hash table for filenames. If it is a 1350: new one, then the hash table info pointer will be null. In this case, 1351: we create a new file_info record to go with the filename, and we initialize 1352: that record with some reasonable values. */ 1353: 1354: static file_info * 1355: find_file (filename, do_not_stat) 1356: char *filename; 1357: int do_not_stat; 1358: /* FILENAME was const, but that causes a warning on AIX when calling stat. 1359: That is probably a bug in AIX, but might as well avoid the warning. */ 1360: { 1361: hash_table_entry *hash_entry_p; 1362: 1363: hash_entry_p = lookup (filename_primary, filename); 1364: if (hash_entry_p->fip) 1365: return hash_entry_p->fip; 1366: else 1367: { 1368: struct stat stat_buf; 1369: file_info *file_p = (file_info *) xmalloc (sizeof (file_info)); 1370: 1371: /* If we cannot get status on any given source file, give a warning 1372: and then just set its time of last modification to infinity. */ 1373: 1374: if (do_not_stat) 1375: stat_buf.st_mtime = (time_t) 0; 1376: else 1377: { 1378: if (my_stat (filename, &stat_buf) == -1) 1379: { 1380: fprintf (stderr, "%s: error: can't get status of `%s': %s\n", 1381: pname, shortpath (NULL, filename), sys_errlist[errno]); 1382: stat_buf.st_mtime = (time_t) -1; 1383: } 1384: } 1385: 1386: hash_entry_p->fip = file_p; 1387: file_p->hash_entry = hash_entry_p; 1388: file_p->defs_decs = NULL; 1389: file_p->mtime = stat_buf.st_mtime; 1390: return file_p; 1391: } 1392: } 1393: 1394: /* Generate a fatal error because some part of the aux_info file is 1395: messed up. */ 1396: 1397: static void 1398: aux_info_corrupted () 1399: { 1400: fprintf (stderr, "\n%s: fatal error: aux info file corrupted at line %d\n", 1401: pname, current_aux_info_lineno); 1402: exit (1); 1403: } 1404: 1405: /* ??? This comment is vague. Say what the condition is for. */ 1406: /* Check to see that a condition is true. This is kind of like an assert. */ 1407: 1408: static void 1409: check_aux_info (cond) 1410: int cond; 1411: { 1412: if (! cond) 1413: aux_info_corrupted (); 1414: } 1415: 1416: /* Given a pointer to the closing right parenthesis for a particular formals 1417: list (in a aux_info file) find the corresponding left parenthesis and 1418: return a pointer to it. */ 1419: 1420: static const char * 1421: find_corresponding_lparen (p) 1422: const char *p; 1423: { 1424: const char *q; 1425: int paren_depth; 1426: 1427: for (paren_depth = 1, q = p-1; paren_depth; q--) 1428: { 1429: switch (*q) 1430: { 1431: case ')': 1432: paren_depth++; 1433: break; 1434: case '(': 1435: paren_depth--; 1436: break; 1437: } 1438: } 1439: return ++q; 1440: } 1441: 1442: /* Given a line from an aux info file, and a time at which the aux info 1443: file it came from was created, check to see if the item described in 1444: the line comes from a file which has been modified since the aux info 1445: file was created. If so, return non-zero, else return zero. */ 1446: 1447: static int 1448: referenced_file_is_newer (l, aux_info_mtime) 1449: const char *l; 1450: time_t aux_info_mtime; 1451: { 1452: const char *p; 1453: file_info *fi_p; 1454: char *filename; 1455: 1456: check_aux_info (l[0] == '/'); 1457: check_aux_info (l[1] == '*'); 1458: check_aux_info (l[2] == ' '); 1459: 1460: { 1461: const char *filename_start = p = l + 3; 1462: 1463: while (*p != ':') 1464: p++; 1465: filename = (char *) alloca ((size_t) (p - filename_start) + 1); 1466: strncpy (filename, filename_start, (size_t) (p - filename_start)); 1467: filename[p-filename_start] = '\0'; 1468: } 1469: 1470: /* Call find_file to find the file_info record associated with the file 1471: which contained this particular def or dec item. Note that this call 1472: may cause a new file_info record to be created if this is the first time 1473: that we have ever known about this particular file. */ 1474: 1475: fi_p = find_file (abspath (invocation_filename, filename), 0); 1476: 1477: return (fi_p->mtime > aux_info_mtime); 1478: } 1479: 1480: /* Given a line of info from the aux_info file, create a new 1481: def_dec_info record to remember all of the important information about 1482: a function definition or declaration. 1483: 1484: Link this record onto the list of such records for the particular file in 1.1.1.2 ! root 1485: which it occurred in proper (descending) line number order (for now). 1.1 root 1486: 1487: If there is an identical record already on the list for the file, throw 1488: this one away. Doing so takes care of the (useless and troublesome) 1489: duplicates which are bound to crop up due to multiple inclusions of any 1490: given individual header file. 1491: 1492: Finally, link the new def_dec record onto the list of such records 1493: pertaining to this particular function name. */ 1494: 1495: static void 1496: save_def_or_dec (l, is_syscalls) 1497: const char *l; 1498: int is_syscalls; 1499: { 1500: const char *p; 1501: const char *semicolon_p; 1502: def_dec_info *def_dec_p = (def_dec_info *) xmalloc (sizeof (def_dec_info)); 1503: 1504: #ifndef UNPROTOIZE 1505: def_dec_p->written = 0; 1506: #endif /* !defined (UNPROTOIZE) */ 1507: 1508: /* Start processing the line by picking off 5 pieces of information from 1509: the left hand end of the line. These are filename, line number, 1510: new/old/implicit flag (new = ANSI prototype format), definition or 1511: declaration flag, and extern/static flag). */ 1512: 1513: check_aux_info (l[0] == '/'); 1514: check_aux_info (l[1] == '*'); 1515: check_aux_info (l[2] == ' '); 1516: 1517: { 1518: const char *filename_start = p = l + 3; 1519: char *filename; 1520: 1521: while (*p != ':') 1522: p++; 1523: filename = (char *) alloca ((size_t) (p - filename_start) + 1); 1524: strncpy (filename, filename_start, (size_t) (p - filename_start)); 1525: filename[p-filename_start] = '\0'; 1526: 1527: /* Call find_file to find the file_info record associated with the file 1528: which contained this particular def or dec item. Note that this call 1529: may cause a new file_info record to be created if this is the first time 1530: that we have ever known about this particular file. 1531: 1532: Note that we started out by forcing all of the base source file names 1533: (i.e. the names of the aux_info files with the .X stripped off) into the 1534: filenames hash table, and we simultaneously setup file_info records for 1535: all of these base file names (even if they may be useless later). 1536: The file_info records for all of these "base" file names (properly) 1537: act as file_info records for the "original" (i.e. un-included) files 1.1.1.2 ! root 1538: which were submitted to gcc for compilation (when the -aux-info 1.1 root 1539: option was used). */ 1540: 1541: def_dec_p->file = find_file (abspath (invocation_filename, filename), is_syscalls); 1542: } 1543: 1544: { 1545: const char *line_number_start = ++p; 1546: char line_number[10]; 1547: 1548: while (*p != ':') 1549: p++; 1550: strncpy (line_number, line_number_start, (size_t) (p - line_number_start)); 1551: line_number[p-line_number_start] = '\0'; 1552: def_dec_p->line = atoi (line_number); 1553: } 1554: 1555: /* Check that this record describes a new-style, old-style, or implicit 1556: definition or declaration. */ 1557: 1558: p++; /* Skip over the `:'. */ 1559: check_aux_info ((*p == 'N') || (*p == 'O') || (*p == 'I')); 1560: 1561: /* Is this a new style (ANSI prototyped) definition or declaration? */ 1562: 1563: def_dec_p->prototyped = (*p == 'N'); 1564: 1565: #ifndef UNPROTOIZE 1566: 1567: /* Is this an implicit declaration? */ 1568: 1569: def_dec_p->is_implicit = (*p == 'I'); 1570: 1571: #endif /* !defined (UNPROTOIZE) */ 1572: 1573: p++; 1574: 1575: check_aux_info ((*p == 'C') || (*p == 'F')); 1576: 1577: /* Is this item a function definition (F) or a declaration (C). Note that 1578: we treat item taken from the syscalls file as though they were function 1579: definitions regardless of what the stuff in the file says. */ 1580: 1581: def_dec_p->is_func_def = ((*p++ == 'F') || is_syscalls); 1582: 1583: #ifndef UNPROTOIZE 1584: def_dec_p->definition = 0; /* Fill this in later if protoizing. */ 1585: #endif /* !defined (UNPROTOIZE) */ 1586: 1587: check_aux_info (*p++ == ' '); 1588: check_aux_info (*p++ == '*'); 1589: check_aux_info (*p++ == '/'); 1590: check_aux_info (*p++ == ' '); 1591: 1592: #ifdef UNPROTOIZE 1593: check_aux_info ((!strncmp (p, "static", 6)) || (!strncmp (p, "extern", 6))); 1594: #else /* !defined (UNPROTOIZE) */ 1595: if (!strncmp (p, "static", 6)) 1596: def_dec_p->is_static = -1; 1597: else if (!strncmp (p, "extern", 6)) 1598: def_dec_p->is_static = 0; 1599: else 1600: check_aux_info (0); /* Didn't find either `extern' or `static'. */ 1601: #endif /* !defined (UNPROTOIZE) */ 1602: 1603: { 1604: const char *ansi_start = p; 1605: 1606: p += 6; /* Pass over the "static" or "extern". */ 1607: 1608: /* We are now past the initial stuff. Search forward from here to find 1609: the terminating semicolon that should immediately follow the entire 1610: ANSI format function declaration. */ 1611: 1612: while (*++p != ';') 1613: continue; 1614: 1615: semicolon_p = p; 1616: 1617: /* Make a copy of the ansi declaration part of the line from the aux_info 1618: file. */ 1619: 1620: def_dec_p->ansi_decl 1621: = dupnstr (ansi_start, (size_t) ((semicolon_p+1) - ansi_start)); 1622: } 1623: 1624: /* Backup and point at the final right paren of the final argument list. */ 1625: 1626: p--; 1627: 1628: /* Now isolate a whole set of formal argument lists, one-by-one. Normally, 1629: there will only be one list to isolate, but there could be more. */ 1630: 1631: def_dec_p->f_list_count = 0; 1632: 1633: #ifndef UNPROTOIZE 1634: def_dec_p->f_list_chain = NULL; 1635: #endif /* !defined (UNPROTOIZE) */ 1636: 1637: for (;;) 1638: { 1639: const char *left_paren_p = find_corresponding_lparen (p); 1640: #ifndef UNPROTOIZE 1641: { 1642: f_list_chain_item *cip = 1643: (f_list_chain_item *) xmalloc (sizeof (f_list_chain_item)); 1644: 1645: cip->formals_list 1646: = dupnstr (left_paren_p + 1, (size_t) (p - (left_paren_p+1))); 1647: 1648: /* Add the new chain item at the head of the current list. */ 1649: 1650: cip->chain_next = def_dec_p->f_list_chain; 1651: def_dec_p->f_list_chain = cip; 1652: } 1653: #endif /* !defined (UNPROTOIZE) */ 1654: def_dec_p->f_list_count++; 1655: 1656: p = left_paren_p - 2; 1657: 1658: /* p must now point either to another right paren, or to the last 1659: character of the name of the function that was declared/defined. 1660: If p points to another right paren, then this indicates that we 1661: are dealing with multiple formals lists. In that case, there 1.1.1.2 ! root 1662: really should be another right paren preceding this right paren. */ 1.1 root 1663: 1664: if (*p != ')') 1665: break; 1666: else 1667: check_aux_info (*--p == ')'); 1668: } 1669: 1670: 1671: { 1672: const char *past_fn = p + 1; 1673: 1674: check_aux_info (*past_fn == ' '); 1675: 1676: /* Scan leftwards over the identifier that names the function. */ 1677: 1678: while (is_id_char (*p)) 1679: p--; 1680: p++; 1681: 1682: /* p now points to the leftmost character of the function name. */ 1683: 1684: { 1685: char *fn_string = (char *) alloca (past_fn - p + 1); 1686: 1687: strncpy (fn_string, p, (size_t) (past_fn - p)); 1688: fn_string[past_fn-p] = '\0'; 1689: def_dec_p->hash_entry = lookup (function_name_primary, fn_string); 1690: } 1691: } 1692: 1693: /* Look at all of the defs and decs for this function name that we have 1694: collected so far. If there is already one which is at the same 1695: line number in the same file, then we can discard this new def_dec_info 1696: record. 1697: 1698: As an extra assurance that any such pair of (nominally) identical 1699: function declarations are in fact identical, we also compare the 1700: ansi_decl parts of the lines from the aux_info files just to be on 1701: the safe side. 1702: 1703: This comparison will fail if (for instance) the user was playing 1704: messy games with the preprocessor which ultimately causes one 1705: function declaration in one header file to look differently when 1706: that file is included by two (or more) other files. */ 1707: 1708: { 1709: const def_dec_info *other; 1710: 1711: for (other = def_dec_p->hash_entry->ddip; other; other = other->next_for_func) 1712: { 1713: if (def_dec_p->line == other->line && def_dec_p->file == other->file) 1714: { 1715: if (strcmp (def_dec_p->ansi_decl, other->ansi_decl)) 1716: { 1717: fprintf (stderr, "%s: error: declaration of function `%s' at %s(%d) takes different forms\n", 1718: pname, 1719: def_dec_p->hash_entry->symbol, 1720: def_dec_p->file->hash_entry->symbol, 1721: def_dec_p->line); 1722: exit (1); 1723: } 1724: free_def_dec (def_dec_p); 1725: return; 1726: } 1727: } 1728: } 1729: 1730: #ifdef UNPROTOIZE 1731: 1732: /* If we are doing unprotoizing, we must now setup the pointers that will 1733: point to the K&R name list and to the K&R argument declarations list. 1734: 1735: Note that if this is only a function declaration, then we should not 1736: expect to find any K&R style formals list following the ANSI-style 1737: formals list. This is because GCC knows that such information is 1738: useless in the case of function declarations (function definitions 1739: are a different story however). 1740: 1741: Since we are unprotoizing, we don't need any such lists anyway. 1742: All we plan to do is to delete all characters between ()'s in any 1743: case. */ 1744: 1745: def_dec_p->formal_names = NULL; 1746: def_dec_p->formal_decls = NULL; 1747: 1748: if (def_dec_p->is_func_def) 1749: { 1750: p = semicolon_p; 1751: check_aux_info (*++p == ' '); 1752: check_aux_info (*++p == '/'); 1753: check_aux_info (*++p == '*'); 1754: check_aux_info (*++p == ' '); 1755: check_aux_info (*++p == '('); 1756: 1757: { 1758: const char *kr_names_start = ++p; /* Point just inside '('. */ 1759: 1760: while (*p++ != ')') 1761: continue; 1762: p--; /* point to closing right paren */ 1763: 1764: /* Make a copy of the K&R parameter names list. */ 1765: 1766: def_dec_p->formal_names 1767: = dupnstr (kr_names_start, (size_t) (p - kr_names_start)); 1768: } 1769: 1770: check_aux_info (*++p == ' '); 1771: p++; 1772: 1773: /* p now points to the first character of the K&R style declarations 1774: list (if there is one) or to the star-slash combination that ends 1775: the comment in which such lists get embedded. */ 1776: 1777: /* Make a copy of the K&R formal decls list and set the def_dec record 1778: to point to it. */ 1779: 1780: if (*p == '*') /* Are there no K&R declarations? */ 1781: { 1782: check_aux_info (*++p == '/'); 1783: def_dec_p->formal_decls = ""; 1784: } 1785: else 1786: { 1787: const char *kr_decls_start = p; 1788: 1789: while (p[0] != '*' || p[1] != '/') 1790: p++; 1791: p--; 1792: 1793: check_aux_info (*p == ' '); 1794: 1795: def_dec_p->formal_decls 1796: = dupnstr (kr_decls_start, (size_t) (p - kr_decls_start)); 1797: } 1798: 1799: /* Handle a special case. If we have a function definition marked as 1800: being in "old" style, and if it's formal names list is empty, then 1801: it may actually have the string "void" in its real formals list 1802: in the original source code. Just to make sure, we will get setup 1803: to convert such things anyway. 1804: 1805: This kludge only needs to be here because of an insurmountable 1806: problem with generating .X files. */ 1807: 1808: if (!def_dec_p->prototyped && !*def_dec_p->formal_names) 1809: def_dec_p->prototyped = 1; 1810: } 1811: 1812: /* Since we are unprotoizing, if this item is already in old (K&R) style, 1813: we can just ignore it. If that is true, throw away the itme now. */ 1814: 1815: if (!def_dec_p->prototyped) 1816: { 1817: free_def_dec (def_dec_p); 1818: return; 1819: } 1820: 1821: #endif /* defined (UNPROTOIZE) */ 1822: 1823: /* Add this record to the head of the list of records pertaining to this 1824: particular function name. */ 1825: 1826: def_dec_p->next_for_func = def_dec_p->hash_entry->ddip; 1827: def_dec_p->hash_entry->ddip = def_dec_p; 1828: 1829: /* Add this new def_dec_info record to the sorted list of def_dec_info 1830: records for this file. Note that we don't have to worry about duplicates 1831: (caused by multiple inclusions of header files) here because we have 1832: already eliminated duplicates above. */ 1833: 1834: if (!def_dec_p->file->defs_decs) 1835: { 1836: def_dec_p->file->defs_decs = def_dec_p; 1837: def_dec_p->next_in_file = NULL; 1838: } 1839: else 1840: { 1841: int line = def_dec_p->line; 1842: const def_dec_info *prev = NULL; 1843: const def_dec_info *curr = def_dec_p->file->defs_decs; 1844: const def_dec_info *next = curr->next_in_file; 1845: 1846: while (next && (line < curr->line)) 1847: { 1848: prev = curr; 1849: curr = next; 1850: next = next->next_in_file; 1851: } 1852: if (line >= curr->line) 1853: { 1854: def_dec_p->next_in_file = curr; 1855: if (prev) 1856: ((NONCONST def_dec_info *) prev)->next_in_file = def_dec_p; 1857: else 1858: def_dec_p->file->defs_decs = def_dec_p; 1859: } 1860: else /* assert (next == NULL); */ 1861: { 1862: ((NONCONST def_dec_info *) curr)->next_in_file = def_dec_p; 1863: /* assert (next == NULL); */ 1864: def_dec_p->next_in_file = next; 1865: } 1866: } 1867: } 1868: 1.1.1.2 ! root 1869: /* Set up the vector COMPILE_PARAMS which is the argument list for running GCC. ! 1870: Also set input_file_name_index and aux_info_file_name_index ! 1871: to the indices of the slots where the file names should go. */ ! 1872: ! 1873: /* We initialize the vector by removing -g, -O, -S, -c, and -o options, ! 1874: and adding '-aux-info AUXFILE -S -o /dev/null INFILE' at the end. */ 1.1 root 1875: 1876: static void 1877: munge_compile_params (params_list) 1878: const char *params_list; 1879: { 1.1.1.2 ! root 1880: /* Build up the contents in a temporary vector ! 1881: that is so big that to has to be big enough. */ ! 1882: char **temp_params ! 1883: = (char **) alloca ((strlen (params_list) + 6) * sizeof (char *)); 1.1 root 1884: int param_count = 0; 1885: const char *param; 1886: 1887: temp_params[param_count++] = compiler_file_name; 1888: for (;;) 1889: { 1890: while (isspace (*params_list)) 1891: params_list++; 1892: if (!*params_list) 1893: break; 1894: param = params_list; 1895: while (*params_list && !isspace (*params_list)) 1896: params_list++; 1897: if (param[0] != '-') 1898: temp_params[param_count++] 1899: = dupnstr (param, (size_t) (params_list - param)); 1900: else 1901: { 1902: switch (param[1]) 1903: { 1904: case 'g': 1905: case 'O': 1906: case 'S': 1907: case 'c': 1908: break; /* Don't copy these. */ 1909: case 'o': 1910: while (isspace (*params_list)) 1911: params_list++; 1912: while (*params_list && !isspace (*params_list)) 1913: params_list++; 1914: break; 1915: default: 1916: temp_params[param_count++] 1917: = dupnstr (param, (size_t) (params_list - param)); 1918: } 1919: } 1920: if (!*params_list) 1921: break; 1922: } 1.1.1.2 ! root 1923: temp_params[param_count++] = "-aux-info"; ! 1924: ! 1925: /* Leave room for the aux-info file name argument. */ ! 1926: aux_info_file_name_index = param_count; ! 1927: temp_params[param_count++] = NULL; ! 1928: 1.1 root 1929: temp_params[param_count++] = "-S"; 1930: temp_params[param_count++] = "-o"; 1931: temp_params[param_count++] = "/dev/null"; 1932: 1.1.1.2 ! root 1933: /* Leave room for the input file name argument. */ ! 1934: input_file_name_index = param_count; ! 1935: temp_params[param_count++] = NULL; ! 1936: /* Terminate the list. */ 1.1 root 1937: temp_params[param_count++] = NULL; 1938: 1939: /* Make a copy of the compile_params in heap space. */ 1940: 1941: compile_params 1942: = (const char **) xmalloc (sizeof (char *) * (param_count+1)); 1943: memcpy (compile_params, temp_params, sizeof (char *) * param_count); 1944: } 1945: 1946: /* Do a recompilation for the express purpose of generating a new aux_info 1947: file to go with a specific base source file. */ 1948: 1949: static int 1950: gen_aux_info_file (base_filename) 1951: const char *base_filename; 1952: { 1953: int child_pid; 1954: 1.1.1.2 ! root 1955: if (!input_file_name_index) 1.1 root 1956: munge_compile_params (""); 1957: 1.1.1.2 ! root 1958: /* Store the full source file name in the argument vector. */ ! 1959: compile_params[input_file_name_index] = shortpath (NULL, base_filename); ! 1960: /* Add .X to source file name to get aux-info file name. */ ! 1961: compile_params[aux_info_file_name_index] ! 1962: = dupnstr (compile_params[input_file_name_index], ! 1963: (2 + strlen (compile_params[input_file_name_index]))); ! 1964: strcat (compile_params[aux_info_file_name_index], ".X"); 1.1 root 1965: 1966: if (!quiet_flag) 1967: fprintf (stderr, "%s: compiling `%s'\n", 1.1.1.2 ! root 1968: pname, compile_params[input_file_name_index]); 1.1 root 1969: 1970: if (child_pid = fork ()) 1971: { 1972: if (child_pid == -1) 1973: { 1974: fprintf (stderr, "%s: error: could not fork process: %s\n", 1975: pname, sys_errlist[errno]); 1976: return 0; 1977: } 1978: 1979: #if 0 1980: /* Print out the command line that the other process is now executing. */ 1981: 1982: if (!quiet_flag) 1983: { 1984: const char **arg; 1985: 1986: fputs ("\t", stderr); 1987: for (arg = compile_params; *arg; arg++) 1988: { 1989: fputs (*arg, stderr); 1990: fputc (' ', stderr); 1991: } 1992: fputc ('\n', stderr); 1993: fflush (stderr); 1994: } 1995: #endif /* 0 */ 1996: 1997: { 1998: int wait_status; 1999: 2000: if (wait (&wait_status) == -1) 2001: { 2002: fprintf (stderr, "%s: error: wait for process failed: %s\n", 2003: pname, sys_errlist[errno]); 2004: return 0; 2005: } 2006: if (!WIFEXITED (wait_status)) 2007: { 1.1.1.2 ! root 2008: fprintf (stderr, "%s: error: subprocess %ld did not exit\n", ! 2009: pname, (long) child_pid); 1.1 root 2010: kill (child_pid, 9); 2011: return 0; 2012: } 1.1.1.2 ! root 2013: if (WEXITSTATUS (wait_status) != 0) ! 2014: { ! 2015: fprintf (stderr, "%s: error: %s: compilation failed\n", ! 2016: pname, base_filename); ! 2017: return 0; ! 2018: } ! 2019: return 1; 1.1 root 2020: } 2021: } 2022: else 2023: { 2024: if (my_execvp (compile_params[0], (char *const *) compile_params)) 2025: { 1.1.1.2 ! root 2026: int e = errno, f = fileno (stderr); ! 2027: write (f, pname, strlen (pname)); ! 2028: write (f, ": ", 2); ! 2029: write (f, compile_params[0], strlen (compile_params[0])); ! 2030: write (f, ": ", 2); ! 2031: write (f, sys_errlist[e], strlen (sys_errlist[e])); ! 2032: write (f, "\n", 1); ! 2033: _exit (1); 1.1 root 2034: } 2035: return 1; /* Never executed. */ 2036: } 2037: } 2038: 2039: /* Read in all of the information contained in a single aux_info file. 2040: Save all of the important stuff for later. */ 2041: 2042: static void 2043: process_aux_info_file (base_source_filename, keep_it, is_syscalls) 2044: const char *base_source_filename; 2045: int keep_it; 2046: int is_syscalls; 2047: { 1.1.1.2 ! root 2048: size_t base_len = strlen (base_source_filename); ! 2049: char * aux_info_filename ! 2050: = (char *) alloca (base_len + strlen (aux_info_suffix) + 1); 1.1 root 2051: char *aux_info_base; 2052: char *aux_info_limit; 1.1.1.2 ! root 2053: char *aux_info_relocated_name; 1.1 root 2054: const char *aux_info_second_line; 2055: time_t aux_info_mtime; 2056: size_t aux_info_size; 2057: 2058: /* Construct the aux_info filename from the base source filename. */ 2059: 2060: strcpy (aux_info_filename, base_source_filename); 2061: strcat (aux_info_filename, aux_info_suffix); 2062: 2063: /* Check that the aux_info file exists and is readable. If it does not 2064: exist, try to create it (once only). */ 2065: 2066: start_over: ; 2067: 2068: { 2069: int retries = 0; 2070: 2071: retry: 2072: if (my_access (aux_info_filename, R_OK) == -1) 2073: { 2074: if (errno == ENOENT && retries == 0) 2075: { 2076: if (is_syscalls) 2077: { 2078: fprintf (stderr, "%s: warning: missing SYSCALLS file `%s'\n", 2079: pname, aux_info_filename); 2080: return; 2081: } 2082: if (!gen_aux_info_file (base_source_filename)) 1.1.1.2 ! root 2083: { ! 2084: errors++; ! 2085: return; ! 2086: } 1.1 root 2087: retries++; 2088: goto retry; 2089: } 2090: else 2091: { 2092: fprintf (stderr, "%s: error: can't read aux info file `%s': %s\n", 2093: pname, shortpath (NULL, aux_info_filename), 2094: sys_errlist[errno]); 2095: errors++; 2096: return; 2097: } 2098: } 2099: } 2100: 2101: { 2102: struct stat stat_buf; 2103: 2104: /* Get some status information about this aux_info file. */ 2105: 2106: if (my_stat (aux_info_filename, &stat_buf) == -1) 2107: { 2108: fprintf (stderr, "%s: error: can't get status of aux info file `%s': %s\n", 2109: pname, shortpath (NULL, aux_info_filename), 2110: sys_errlist[errno]); 2111: errors++; 2112: return; 2113: } 2114: 2115: /* Check on whether or not this aux_info file is zero length. If it is, 2116: then just ignore it and return. */ 2117: 2118: if ((aux_info_size = stat_buf.st_size) == 0) 2119: return; 2120: 2121: /* Get the date/time of last modification for this aux_info file and 2122: remember it. We will have to check that any source files that it 2123: contains information about are at least this old or older. */ 2124: 2125: aux_info_mtime = stat_buf.st_mtime; 2126: } 2127: 2128: { 2129: int aux_info_file; 2130: 2131: /* Open the aux_info file. */ 2132: 2133: if ((aux_info_file = my_open (aux_info_filename, O_RDONLY, 0444 )) == -1) 2134: { 2135: fprintf (stderr, "%s: error: can't open aux info file `%s' for reading: %s\n", 2136: pname, shortpath (NULL, aux_info_filename), 2137: sys_errlist[errno]); 2138: return; 2139: } 2140: 2141: /* Allocate space to hold the aux_info file in memory. */ 2142: 2143: aux_info_base = xmalloc (aux_info_size + 1); 2144: aux_info_limit = aux_info_base + aux_info_size; 2145: *aux_info_limit = '\0'; 2146: 2147: /* Read the aux_info file into memory. */ 2148: 2149: if (read (aux_info_file, aux_info_base, aux_info_size) != aux_info_size) 2150: { 2151: fprintf (stderr, "%s: error: while reading aux info file `%s': %s\n", 2152: pname, shortpath (NULL, aux_info_filename), 2153: sys_errlist[errno]); 2154: free (aux_info_base); 2155: close (aux_info_file); 2156: return; 2157: } 2158: 2159: /* Close the aux info file. */ 2160: 2161: if (close (aux_info_file)) 2162: { 2163: fprintf (stderr, "%s: error: while closing aux info file `%s': %s\n", 2164: pname, shortpath (NULL, aux_info_filename), 2165: sys_errlist[errno]); 2166: free (aux_info_base); 2167: close (aux_info_file); 2168: return; 2169: } 2170: } 2171: 2172: /* Delete the aux_info file (unless requested not to). If the deletion 2173: fails for some reason, don't even worry about it. */ 2174: 2175: if (!keep_it) 2176: if (my_unlink (aux_info_filename) == -1) 2177: fprintf (stderr, "%s: error: can't delete aux info file `%s': %s\n", 2178: pname, shortpath (NULL, aux_info_filename), 2179: sys_errlist[errno]); 2180: 2181: /* Save a pointer into the first line of the aux_info file which 2182: contains the filename of the directory from which the compiler 2183: was invoked when the associated source file was compiled. 2184: This information is used later to help create complete 2185: filenames out of the (potentially) relative filenames in 2186: the aux_info file. */ 2187: 2188: { 2189: char *p = aux_info_base; 2190: 2191: while (*p != ':') 2192: p++; 2193: p++; 2194: while (*p == ' ') 2195: p++; 2196: invocation_filename = p; /* Save a pointer to first byte of path. */ 2197: while (*p != ' ') 2198: p++; 2199: *p++ = '/'; 2200: *p++ = '\0'; 2201: while (*p++ != '\n') 2202: continue; 2203: aux_info_second_line = p; 1.1.1.2 ! root 2204: aux_info_relocated_name = 0; ! 2205: if (invocation_filename[0] != '/') ! 2206: { ! 2207: /* INVOCATION_FILENAME is relative; ! 2208: append it to BASE_SOURCE_FILENAME's dir. */ ! 2209: char *dir_end; ! 2210: aux_info_relocated_name = xmalloc (base_len + (p-invocation_filename)); ! 2211: strcpy (aux_info_relocated_name, base_source_filename); ! 2212: dir_end = strrchr (aux_info_relocated_name, '/'); ! 2213: if (dir_end) ! 2214: dir_end++; ! 2215: else ! 2216: dir_end = aux_info_relocated_name; ! 2217: strcpy (dir_end, invocation_filename); ! 2218: invocation_filename = aux_info_relocated_name; ! 2219: } 1.1 root 2220: } 2221: 2222: 2223: { 2224: const char *aux_info_p; 2225: 2226: /* Do a pre-pass on the lines in the aux_info file, making sure that all 2227: of the source files referenced in there are at least as old as this 2228: aux_info file itself. If not, go back and regenerate the aux_info 2229: file anew. Don't do any of this for the syscalls file. */ 2230: 2231: if (!is_syscalls) 2232: { 2233: current_aux_info_lineno = 2; 2234: 2235: for (aux_info_p = aux_info_second_line; *aux_info_p; ) 2236: { 2237: if (referenced_file_is_newer (aux_info_p, aux_info_mtime)) 2238: { 2239: free (aux_info_base); 1.1.1.2 ! root 2240: xfree (aux_info_relocated_name); ! 2241: if (keep_it && my_unlink (aux_info_filename) == -1) 1.1 root 2242: { 2243: fprintf (stderr, "%s: error: can't delete file `%s': %s\n", 2244: pname, shortpath (NULL, aux_info_filename), 2245: sys_errlist[errno]); 2246: return; 2247: } 2248: goto start_over; 2249: } 2250: 2251: /* Skip over the rest of this line to start of next line. */ 2252: 2253: while (*aux_info_p != '\n') 2254: aux_info_p++; 2255: aux_info_p++; 2256: current_aux_info_lineno++; 2257: } 2258: } 2259: 2260: /* Now do the real pass on the aux_info lines. Save their information in 2261: the in-core data base. */ 2262: 2263: current_aux_info_lineno = 2; 2264: 2265: for (aux_info_p = aux_info_second_line; *aux_info_p;) 2266: { 2267: char *unexpanded_line = unexpand_if_needed (aux_info_p); 2268: 2269: if (unexpanded_line) 2270: { 2271: save_def_or_dec (unexpanded_line, is_syscalls); 2272: free (unexpanded_line); 2273: } 2274: else 2275: save_def_or_dec (aux_info_p, is_syscalls); 2276: 2277: /* Skip over the rest of this line and get to start of next line. */ 2278: 2279: while (*aux_info_p != '\n') 2280: aux_info_p++; 2281: aux_info_p++; 2282: current_aux_info_lineno++; 2283: } 2284: } 2285: 2286: free (aux_info_base); 1.1.1.2 ! root 2287: xfree (aux_info_relocated_name); 1.1 root 2288: } 2289: 2290: #ifndef UNPROTOIZE 2291: 2292: /* Check an individual filename for a .c suffix. If the filename has this 2293: suffix, rename the file such that its suffix is changed to .C. This 2294: function implements the -C option. */ 2295: 2296: static void 2297: rename_c_file (hp) 2298: const hash_table_entry *hp; 2299: { 2300: const char *filename = hp->symbol; 2301: int last_char_index = strlen (filename) - 1; 2302: char *const new_filename = (char *) alloca (strlen (filename) + 1); 2303: 2304: /* Note that we don't care here if the given file was converted or not. It 2305: is possible that the given file was *not* converted, simply because there 2306: was nothing in it which actually required conversion. Even in this case, 2307: we want to do the renaming. Note that we only rename files with the .c 2308: suffix. */ 2309: 2310: if (filename[last_char_index] != 'c' || filename[last_char_index-1] != '.') 2311: return; 2312: 2313: strcpy (new_filename, filename); 2314: new_filename[last_char_index] = 'C'; 2315: 2316: if (my_link (filename, new_filename) == -1) 2317: { 2318: fprintf (stderr, "%s: warning: can't link file `%s' to `%s': %s\n", 2319: pname, shortpath (NULL, filename), 2320: shortpath (NULL, new_filename), sys_errlist[errno]); 2321: errors++; 2322: return; 2323: } 2324: 2325: if (my_unlink (filename) == -1) 2326: { 2327: fprintf (stderr, "%s: warning: can't delete file `%s': %s\n", 2328: pname, shortpath (NULL, filename), sys_errlist[errno]); 2329: errors++; 2330: return; 2331: } 2332: } 2333: 2334: #endif /* !defined (UNPROTOIZE) */ 2335: 2336: /* Take the list of definitions and declarations attached to a particular 2337: file_info node and reverse the order of the list. This should get the 2338: list into an order such that the item with the lowest associated line 2339: number is nearest the head of the list. When these lists are originally 2340: built, they are in the opposite order. We want to traverse them in 2341: normal line number order later (i.e. lowest to highest) so reverse the 2342: order here. */ 2343: 2344: static void 2345: reverse_def_dec_list (hp) 2346: const hash_table_entry *hp; 2347: { 2348: file_info *file_p = hp->fip; 2349: const def_dec_info *prev = NULL; 2350: const def_dec_info *current = file_p->defs_decs; 2351: 2352: if (!( current = file_p->defs_decs)) 2353: return; /* no list to reverse */ 2354: 2355: prev = current; 2356: if (! (current = current->next_in_file)) 2357: return; /* can't reverse a single list element */ 2358: 2359: ((NONCONST def_dec_info *) prev)->next_in_file = NULL; 2360: 2361: while (current) 2362: { 2363: const def_dec_info *next = current->next_in_file; 2364: 2365: ((NONCONST def_dec_info *) current)->next_in_file = prev; 2366: prev = current; 2367: current = next; 2368: } 2369: 2370: file_p->defs_decs = prev; 2371: } 2372: 2373: #ifndef UNPROTOIZE 2374: 2375: /* Find the (only?) extern definition for a particular function name, starting 2376: from the head of the linked list of entries for the given name. If we 2377: cannot find an extern definition for the given function name, issue a 2378: warning and scrounge around for the next best thing, i.e. an extern 2379: function declaration with a prototype attached to it. Note that we only 2380: allow such substitutions for extern declarations and never for static 2381: declarations. That's because the only reason we allow them at all is 2382: to let un-prototyped function declarations for system-supplied library 2383: functions get their prototypes from our own extra SYSCALLS.c.X file which 2384: contains all of the correct prototypes for system functions. */ 2385: 2386: static const def_dec_info * 2387: find_extern_def (head, user) 2388: const def_dec_info *head; 2389: const def_dec_info *user; 2390: { 2391: const def_dec_info *dd_p; 2392: const def_dec_info *extern_def_p = NULL; 2393: int conflict_noted = 0; 2394: 2395: /* Don't act too stupid here. Somebody may try to convert an entire system 2396: in one swell fwoop (rather than one program at a time, as should be done) 2397: and in that case, we may find that there are multiple extern definitions 2398: of a given function name in the entire set of source files that we are 2399: converting. If however one of these definitions resides in exactly the 2400: same source file as the reference we are trying to satisfy then in that 2401: case it would be stupid for us to fail to realize that this one definition 2402: *must* be the precise one we are looking for. 2403: 2404: To make sure that we don't miss an opportunity to make this "same file" 2405: leap of faith, we do a prescan of the list of records relating to the 2406: given function name, and we look (on this first scan) *only* for a 2407: definition of the function which is in the same file as the reference 2408: we are currently trying to satisfy. */ 2409: 2410: for (dd_p = head; dd_p; dd_p = dd_p->next_for_func) 2411: if (dd_p->is_func_def && !dd_p->is_static && dd_p->file == user->file) 2412: return dd_p; 2413: 2414: /* Now, since we have not found a definition in the same file as the 2415: reference, we scan the list again and consider all possibilities from 2416: all files. Here we may get conflicts with the things listed in the 2417: SYSCALLS.c.X file, but if that happens it only means that the source 2418: code being converted contains its own definition of a function which 2419: could have been supplied by libc.a. In such cases, we should avoid 2420: issuing the normal warning, and defer to the definition given in the 2421: user's own code. */ 2422: 2423: for (dd_p = head; dd_p; dd_p = dd_p->next_for_func) 2424: if (dd_p->is_func_def && !dd_p->is_static) 2425: { 2426: if (!extern_def_p) /* Previous definition? */ 2427: extern_def_p = dd_p; /* Remember the first definition found. */ 2428: else 2429: { 2430: /* Ignore definition just found if it came from SYSCALLS.c.X. */ 2431: 2432: if (is_syscalls_file (dd_p->file)) 2433: continue; 2434: 2435: /* Quietly replace the definition previously found with the one 2436: just found if the previous one was from SYSCALLS.c.X. */ 2437: 2438: if (is_syscalls_file (extern_def_p->file)) 2439: { 2440: extern_def_p = dd_p; 2441: continue; 2442: } 2443: 2444: /* If we get here, then there is a conflict between two function 2445: declarations for the same function, both of which came from the 2446: user's own code. */ 2447: 2448: if (!conflict_noted) /* first time we noticed? */ 2449: { 2450: conflict_noted = 1; 2451: fprintf (stderr, "%s: error: conflicting extern definitions of '%s'\n", 2452: pname, head->hash_entry->symbol); 2453: if (!quiet_flag) 2454: { 2455: fprintf (stderr, "%s: declarations of '%s' will not be converted\n", 2456: pname, head->hash_entry->symbol); 2457: fprintf (stderr, "%s: conflict list for '%s' follows:\n", 2458: pname, head->hash_entry->symbol); 2459: fprintf (stderr, "%s: %s(%d): %s\n", 2460: pname, 2461: shortpath (NULL, extern_def_p->file->hash_entry->symbol), 2462: extern_def_p->line, extern_def_p->ansi_decl); 2463: } 2464: } 2465: if (!quiet_flag) 2466: fprintf (stderr, "%s: %s(%d): %s\n", 2467: pname, 2468: shortpath (NULL, dd_p->file->hash_entry->symbol), 2469: dd_p->line, dd_p->ansi_decl); 2470: } 2471: } 2472: 2473: /* We want to err on the side of caution, so if we found multiple conflicting 2474: definitions for the same function, treat this as being that same as if we 2475: had found no definitions (i.e. return NULL). */ 2476: 2477: if (conflict_noted) 2478: return NULL; 2479: 2480: if (!extern_def_p) 2481: { 2482: /* We have no definitions for this function so do the next best thing. 2483: Search for an extern declaration already in prototype form. */ 2484: 2485: for (dd_p = head; dd_p; dd_p = dd_p->next_for_func) 2486: if (!dd_p->is_func_def && !dd_p->is_static && dd_p->prototyped) 2487: { 2488: extern_def_p = dd_p; /* save a pointer to the definition */ 2489: if (!quiet_flag) 2490: fprintf (stderr, "%s: warning: using formals list from %s(%d) for function `%s'\n", 2491: pname, 2492: shortpath (NULL, dd_p->file->hash_entry->symbol), 2493: dd_p->line, dd_p->hash_entry->symbol); 2494: break; 2495: } 2496: 2497: /* Gripe about unprototyped function declarations that we found no 2498: corresponding definition (or other source of prototype information) 2499: for. 2500: 2501: Gripe even if the unprototyped declaration we are worried about 2502: exists in a file in one of the "system" include directories. We 2503: can gripe about these because we should have at least found a 2504: corresponding (pseudo) definition in the SYSCALLS.c.X file. If we 2505: didn't, then that means that the SYSCALLS.c.X file is missing some 2506: needed prototypes for this particular system. That is worth telling 2507: the user about! */ 2508: 2509: if (!extern_def_p) 2510: { 2511: const char *file = user->file->hash_entry->symbol; 2512: 2513: if (!quiet_flag) 2514: if (in_system_include_dir (file)) 2515: { 2516: /* Why copy this string into `needed' at all? 2517: Why not just use user->ansi_decl without copying? */ 2518: char *needed = (char *) alloca (strlen (user->ansi_decl) + 1); 2519: char *p; 2520: 2521: strcpy (needed, user->ansi_decl); 2522: p = (NONCONST char *) substr (needed, user->hash_entry->symbol) 2523: + strlen (user->hash_entry->symbol) + 2; 1.1.1.2 ! root 2524: /* Avoid having ??? in the string. */ ! 2525: *p++ = '?'; ! 2526: *p++ = '?'; ! 2527: *p++ = '?'; ! 2528: strcpy (p, ");"); 1.1 root 2529: 2530: fprintf (stderr, "%s: %d: `%s' used but missing from SYSCALLS\n", 2531: shortpath (NULL, file), user->line, 2532: needed+7); /* Don't print "extern " */ 2533: } 2534: else 2535: fprintf (stderr, "%s: %d: warning: no extern definition for `%s'\n", 2536: shortpath (NULL, file), user->line, 2537: user->hash_entry->symbol); 2538: } 2539: } 2540: return extern_def_p; 2541: } 2542: 2543: /* Find the (only?) static definition for a particular function name in a 2544: given file. Here we get the function-name and the file info indirectly 2545: from the def_dec_info record pointer which is passed in. */ 2546: 2547: static const def_dec_info * 2548: find_static_definition (user) 2549: const def_dec_info *user; 2550: { 2551: const def_dec_info *head = user->hash_entry->ddip; 2552: const def_dec_info *dd_p; 2553: int num_static_defs = 0; 2554: const def_dec_info *static_def_p = NULL; 2555: 2556: for (dd_p = head; dd_p; dd_p = dd_p->next_for_func) 2557: if (dd_p->is_func_def && dd_p->is_static && (dd_p->file == user->file)) 2558: { 2559: static_def_p = dd_p; /* save a pointer to the definition */ 2560: num_static_defs++; 2561: } 2562: if (num_static_defs == 0) 2563: { 2564: if (!quiet_flag) 2565: fprintf (stderr, "%s: warning: no static definition for `%s' in file `%s'\n", 2566: pname, head->hash_entry->symbol, 2567: shortpath (NULL, user->file->hash_entry->symbol)); 2568: } 2569: else if (num_static_defs > 1) 2570: { 2571: fprintf (stderr, "%s: error: multiple static defs of `%s' in file `%s'\n", 2572: pname, head->hash_entry->symbol, 2573: shortpath (NULL, user->file->hash_entry->symbol)); 2574: return NULL; 2575: } 2576: return static_def_p; 2577: } 2578: 2579: /* Find good prototype style formal argument lists for all of the function 2580: declarations which didn't have them before now. 2581: 2582: To do this we consider each function name one at a time. For each function 2583: name, we look at the items on the linked list of def_dec_info records for 2584: that particular name. 2585: 2586: Somewhere on this list we should find one (and only one) def_dec_info 2587: record which represents the actual function definition, and this record 2588: should have a nice formal argument list already associated with it. 2589: 2590: Thus, all we have to do is to connect up all of the other def_dec_info 2591: records for this particular function name to the special one which has 2592: the full-blown formals list. 2593: 2594: Of course it is a little more complicated than just that. See below for 2595: more details. */ 2596: 2597: static void 2598: connect_defs_and_decs (hp) 2599: const hash_table_entry *hp; 2600: { 2601: const def_dec_info *dd_p; 2602: const def_dec_info *extern_def_p = NULL; 2603: int first_extern_reference = 1; 2604: 2605: /* Traverse the list of definitions and declarations for this particular 2606: function name. For each item on the list, if it is a function 2607: definition (either old style or new style) then GCC has already been 2608: kind enough to produce a prototype for us, and it is associated with 2609: the item already, so declare the item as its own associated "definition". 2610: 2611: Also, for each item which is only a function declaration, but which 2612: nonetheless has its own prototype already (obviously supplied by the user) 2613: declare the item as it's own definition. 2614: 2615: Note that when/if there are multiple user-supplied prototypes already 2616: present for multiple declarations of any given function, these multiple 2617: prototypes *should* all match exactly with one another and with the 2618: prototype for the actual function definition. We don't check for this 2619: here however, since we assume that the compiler must have already done 1.1.1.2 ! root 2620: this consistency checking when it was creating the .X files. */ 1.1 root 2621: 2622: for (dd_p = hp->ddip; dd_p; dd_p = dd_p->next_for_func) 2623: if (dd_p->prototyped) 2624: ((NONCONST def_dec_info *) dd_p)->definition = dd_p; 2625: 2626: /* Traverse the list of definitions and declarations for this particular 2627: function name. For each item on the list, if it is an extern function 2628: declaration and if it has no associated definition yet, go try to find 2629: the matching extern definition for the declaration. 2630: 2631: When looking for the matching function definition, warn the user if we 2632: fail to find one. 2633: 2634: If we find more that one function definition also issue a warning. 2635: 2636: Do the search for the matching definition only once per unique function 2637: name (and only when absolutely needed) so that we can avoid putting out 2638: redundant warning messages, and so that we will only put out warning 2639: messages when there is actually a reference (i.e. a declaration) for 2640: which we need to find a matching definition. */ 2641: 2642: for (dd_p = hp->ddip; dd_p; dd_p = dd_p->next_for_func) 2643: if (!dd_p->is_func_def && !dd_p->is_static && !dd_p->definition) 2644: { 2645: if (first_extern_reference) 2646: { 2647: extern_def_p = find_extern_def (hp->ddip, dd_p); 2648: first_extern_reference = 0; 2649: } 2650: ((NONCONST def_dec_info *) dd_p)->definition = extern_def_p; 2651: } 2652: 2653: /* Traverse the list of definitions and declarations for this particular 2654: function name. For each item on the list, if it is a static function 2655: declaration and if it has no associated definition yet, go try to find 2656: the matching static definition for the declaration within the same file. 2657: 2658: When looking for the matching function definition, warn the user if we 2659: fail to find one in the same file with the declaration, and refuse to 2660: convert this kind of cross-file static function declaration. After all, 2661: this is stupid practice and should be discouraged. 2662: 2663: We don't have to worry about the possibility that there is more than one 2664: matching function definition in the given file because that would have 2665: been flagged as an error by the compiler. 2666: 2667: Do the search for the matching definition only once per unique 2668: function-name/source-file pair (and only when absolutely needed) so that 2669: we can avoid putting out redundant warning messages, and so that we will 2670: only put out warning messages when there is actually a reference (i.e. a 2671: declaration) for which we actually need to find a matching definition. */ 2672: 2673: for (dd_p = hp->ddip; dd_p; dd_p = dd_p->next_for_func) 2674: if (!dd_p->is_func_def && dd_p->is_static && !dd_p->definition) 2675: { 2676: const def_dec_info *dd_p2; 2677: const def_dec_info *static_def; 2678: 2679: /* We have now found a single static declaration for which we need to 2680: find a matching definition. We want to minimize the work (and the 2681: number of warnings), so we will find an appropriate (matching) 2682: static definition for this declaration, and then distribute it 2683: (as the definition for) any and all other static declarations 2684: for this function name which occur within the same file, and which 2685: do not already have definitions. 2686: 2687: Note that a trick is used here to prevent subsequent attempts to 2688: call find_static_definition for a given function-name & file 2689: if the first such call returns NULL. Essentially, we convert 2690: these NULL return values to -1, and put the -1 into the definition 2691: field for each other static declaration from the same file which 2692: does not already have an associated definition. 2693: This makes these other static declarations look like they are 2694: actually defined already when the outer loop here revisits them 2695: later on. Thus, the outer loop will skip over them. Later, we 2696: turn the -1's back to NULL's. */ 2697: 2698: ((NONCONST def_dec_info *) dd_p)->definition = 2699: (static_def = find_static_definition (dd_p)) 2700: ? static_def 2701: : (const def_dec_info *) -1; 2702: 2703: for (dd_p2 = dd_p->next_for_func; dd_p2; dd_p2 = dd_p2->next_for_func) 2704: if (!dd_p2->is_func_def && dd_p2->is_static 2705: && !dd_p2->definition && (dd_p2->file == dd_p->file)) 2706: ((NONCONST def_dec_info *)dd_p2)->definition = dd_p->definition; 2707: } 2708: 2709: /* Convert any dummy (-1) definitions we created in the step above back to 2710: NULL's (as they should be). */ 2711: 2712: for (dd_p = hp->ddip; dd_p; dd_p = dd_p->next_for_func) 2713: if (dd_p->definition == (def_dec_info *) -1) 2714: ((NONCONST def_dec_info *) dd_p)->definition = NULL; 2715: } 2716: 2717: #endif /* !defined (UNPROTOIZE) */ 2718: 2719: /* Give a pointer into the clean text buffer, return a number which is the 2720: original source line number that the given pointer points into. */ 2721: 2722: static int 2723: identify_lineno (clean_p) 2724: const char *clean_p; 2725: { 2726: int line_num = 1; 2727: const char *scan_p; 2728: 2729: for (scan_p = clean_text_base; scan_p <= clean_p; scan_p++) 2730: if (*scan_p == '\n') 2731: line_num++; 2732: return line_num; 2733: } 2734: 2735: /* Issue an error message and give up on doing this particular edit. */ 2736: 2737: static void 2738: declare_source_confusing (clean_p) 2739: const char *clean_p; 2740: { 2741: if (!quiet_flag) 2742: { 2743: if (clean_p == 0) 2744: fprintf (stderr, "%s: %d: warning: source too confusing\n", 2745: shortpath (NULL, convert_filename), last_known_line_number); 2746: else 2747: fprintf (stderr, "%s: %d: warning: source too confusing\n", 2748: shortpath (NULL, convert_filename), 2749: identify_lineno (clean_p)); 2750: } 2751: longjmp (source_confusion_recovery, 1); 2752: } 2753: 2754: /* Check that a condition which is expected to be true in the original source 2755: code is in fact true. If not, issue an error message and give up on 2756: converting this particular source file. */ 2757: 2758: static void 2759: check_source (cond, clean_p) 2760: int cond; 2761: const char *clean_p; 2762: { 2763: if (!cond) 2764: declare_source_confusing (clean_p); 2765: } 2766: 2767: /* If we think of the in-core cleaned text buffer as a memory mapped 2768: file (with the variable last_known_line_start acting as sort of a 2769: file pointer) then we can imagine doing "seeks" on the buffer. The 2770: following routine implements a kind of "seek" operation for the in-core 2771: (cleaned) copy of the source file. When finished, it returns a pointer to 2772: the start of a given (numbered) line in the cleaned text buffer. 2773: 2774: Note that protoize only has to "seek" in the forward direction on the 2775: in-core cleaned text file buffers, and it never needs to back up. 2776: 2777: This routine is made a little bit faster by remembering the line number 2778: (and pointer value) supplied (and returned) from the previous "seek". 2779: This prevents us from always having to start all over back at the top 2780: of the in-core cleaned buffer again. */ 2781: 2782: static const char * 2783: seek_to_line (n) 2784: int n; 2785: { 2786: if (n < last_known_line_number) 2787: abort (); 2788: 2789: while (n > last_known_line_number) 2790: { 2791: while (*last_known_line_start != '\n') 2792: check_source (++last_known_line_start < clean_text_limit, 0); 2793: last_known_line_start++; 2794: last_known_line_number++; 2795: } 2796: return last_known_line_start; 2797: } 2798: 2799: /* Given a pointer to a character in the cleaned text buffer, return a pointer 2800: to the next non-whitepace character which follows it. */ 2801: 2802: static const char * 2803: forward_to_next_token_char (ptr) 2804: const char *ptr; 2805: { 2806: for (++ptr; isspace (*ptr); check_source (++ptr < clean_text_limit, 0)) 2807: continue; 2808: return ptr; 2809: } 2810: 2811: /* Copy a chunk of text of length `len' and starting at `str' to the current 2812: output buffer. Note that all attempts to add stuff to the current output 2813: buffer ultimately go through here. */ 2814: 2815: static void 2816: output_bytes (str, len) 2817: const char *str; 2818: size_t len; 2819: { 2820: if ((repl_write_ptr + 1) + len >= repl_text_limit) 2821: { 2822: size_t new_size = (repl_text_limit - repl_text_base) << 1; 2823: char *new_buf = (char *) xrealloc (repl_text_base, new_size); 2824: 2825: repl_write_ptr = new_buf + (repl_write_ptr - repl_text_base); 2826: repl_text_base = new_buf; 2827: repl_text_limit = new_buf + new_size; 2828: } 2829: memcpy (repl_write_ptr + 1, str, len); 2830: repl_write_ptr += len; 2831: } 2832: 2833: /* Copy all bytes (except the trailing null) of a null terminated string to 2834: the current output buffer. */ 2835: 2836: static void 2837: output_string (str) 2838: const char *str; 2839: { 2840: output_bytes (str, strlen (str)); 2841: } 2842: 2843: /* Copy some characters from the original text buffer to the current output 2844: buffer. 2845: 2846: This routine takes a pointer argument `p' which is assumed to be a pointer 2847: into the cleaned text buffer. The bytes which are copied are the `original' 2848: equivalents for the set of bytes between the last value of `clean_read_ptr' 2849: and the argument value `p'. 2850: 2851: The set of bytes copied however, comes *not* from the cleaned text buffer, 2852: but rather from the direct counterparts of these bytes within the original 2853: text buffer. 2854: 2855: Thus, when this function is called, some bytes from the original text 2856: buffer (which may include original comments and preprocessing directives) 2857: will be copied into the output buffer. 2858: 2859: Note that the request implide when this routine is called includes the 2860: byte pointed to by the argument pointer `p'. */ 2861: 2862: static void 2863: output_up_to (p) 2864: const char *p; 2865: { 2866: size_t copy_length = (size_t) (p - clean_read_ptr); 2867: const char *copy_start = orig_text_base+(clean_read_ptr-clean_text_base)+1; 2868: 2869: if (copy_length == 0) 2870: return; 2871: 2872: output_bytes (copy_start, copy_length); 2873: clean_read_ptr = p; 2874: } 2875: 2876: /* Given a pointer to a def_dec_info record which represents some form of 2877: definition of a function (perhaps a real definition, or in lieu of that 2878: perhaps just a declaration with a full prototype) return true if this 2879: function is one which we should avoid converting. Return false 2880: otherwise. */ 2881: 2882: static int 2883: other_variable_style_function (ansi_header) 2884: const char *ansi_header; 2885: { 2886: #ifdef UNPROTOIZE 2887: 2888: /* See if we have a stdarg function, or a function which has stdarg style 2889: parameters or a stdarg style return type. */ 2890: 2891: return (int) substr (ansi_header, "..."); 2892: 2893: #else /* !defined (UNPROTOIZE) */ 2894: 2895: /* See if we have a varargs function, or a function which has varargs style 2896: parameters or a varargs style return type. */ 2897: 2898: const char *p; 2899: int len = strlen (varargs_style_indicator); 2900: 2901: for (p = ansi_header; p; ) 2902: { 2903: const char *candidate; 2904: 2905: if ((candidate = substr (p, varargs_style_indicator)) == 0) 2906: return 0; 2907: else 2908: if (!is_id_char (candidate[-1]) && !is_id_char (candidate[len])) 2909: return 1; 2910: else 2911: p = candidate + 1; 2912: } 2913: return 0; 2914: #endif /* !defined (UNPROTOIZE) */ 2915: } 2916: 2917: /* Do the editing operation specifically for a function "declaration". Note 2918: that editing for function "definitions" are handled in a separate routine 2919: below. */ 2920: 2921: static void 2922: edit_fn_declaration (def_dec_p, clean_text_p) 2923: const def_dec_info *def_dec_p; 2924: const char *VOLATILE clean_text_p; 2925: { 2926: const char *start_formals; 2927: const char *end_formals; 2928: const char *function_to_edit = def_dec_p->hash_entry->symbol; 2929: size_t func_name_len = strlen (function_to_edit); 2930: const char *end_of_fn_name; 2931: 2932: #ifndef UNPROTOIZE 2933: 2934: const f_list_chain_item *this_f_list_chain_item; 2935: const def_dec_info *definition = def_dec_p->definition; 2936: 2937: /* If we are protoizing, and if we found no corresponding definition for 2938: this particular function declaration, then just leave this declaration 2939: exactly as it is. */ 2940: 2941: if (!definition) 2942: return; 2943: 2944: /* If we are protoizing, and if the corresponding definition that we found 2945: for this particular function declaration defined an old style varargs 2946: function, then we want to issue a warning and just leave this function 2947: declaration unconverted. */ 2948: 2949: if (other_variable_style_function (definition->ansi_decl)) 2950: { 2951: if (!quiet_flag) 2952: fprintf (stderr, "%s: %d: warning: varargs function declaration not converted\n", 2953: shortpath (NULL, def_dec_p->file->hash_entry->symbol), 2954: def_dec_p->line); 2955: return; 2956: } 2957: 2958: #endif /* !defined (UNPROTOIZE) */ 2959: 2960: /* Setup here to recover from confusing source code detected during this 2961: particular "edit". */ 2962: 2963: save_pointers (); 2964: if (setjmp (source_confusion_recovery)) 2965: { 2966: restore_pointers (); 2967: fprintf (stderr, "%s: declaration of function `%s' not converted\n", 2968: pname, function_to_edit); 2969: return; 2970: } 2971: 2972: /* We are editing a function declaration. The line number we did a seek to 2973: contains the comma or semicolon which follows the declaration. Our job 2974: now is to scan backwards looking for the function name. This name *must* 2975: be followed by open paren (ignoring whitespace, of course). We need to 2976: replace everything between that open paren and the corresponding closing 2977: paren. If we are protoizing, we need to insert the prototype-style 2978: formals lists. If we are unprotoizing, we need to just delete everything 2979: between the pairs of opening and closing parens. */ 2980: 2981: /* First move up to the end of the line. */ 2982: 2983: while (*clean_text_p != '\n') 2984: check_source (++clean_text_p < clean_text_limit, 0); 2985: clean_text_p--; /* Point to just before the newline character. */ 2986: 2987: /* Now we can scan backwards for the function name. */ 2988: 2989: do 2990: { 2991: for (;;) 2992: { 2993: /* Scan leftwards until we find some character which can be 2994: part of an identifier. */ 2995: 2996: while (!is_id_char (*clean_text_p)) 2997: check_source (--clean_text_p > clean_read_ptr, 0); 2998: 2999: /* Scan backwards until we find a char that cannot be part of an 3000: identifier. */ 3001: 3002: while (is_id_char (*clean_text_p)) 3003: check_source (--clean_text_p > clean_read_ptr, 0); 3004: 3005: /* Having found an "id break", see if the following id is the one 3006: that we are looking for. If so, then exit from this loop. */ 3007: 3008: if (!strncmp (clean_text_p+1, function_to_edit, func_name_len)) 3009: { 3010: char ch = *(clean_text_p + 1 + func_name_len); 3011: 3012: /* Must also check to see that the name in the source text 3013: ends where it should (in order to prevent bogus matches 3014: on similar but longer identifiers. */ 3015: 3016: if (! is_id_char (ch)) 3017: break; /* exit from loop */ 3018: } 3019: } 3020: 3021: /* We have now found the first perfect match for the function name in 3022: our backward search. This may or may not be the actual function 3023: name at the start of the actual function declaration (i.e. we could 3024: have easily been mislead). We will try to avoid getting fooled too 3025: often by looking forward for the open paren which should follow the 3026: identifier we just found. We ignore whitespace while hunting. If 3027: the next non-whitespace byte we see is *not* an open left paren, 3028: then we must assume that we have been fooled and we start over 3029: again accordingly. Note that there is no guarrantee, that even if 3030: we do see the open paren, that we are in the right place. 3031: Programmers do the strangest things sometimes! */ 3032: 3033: end_of_fn_name = clean_text_p + strlen (def_dec_p->hash_entry->symbol); 3034: start_formals = forward_to_next_token_char (end_of_fn_name); 3035: } 3036: while (*start_formals != '('); 3037: 3038: /* start_of_formals now points to the opening left paren which immediately 3039: follows the name of the function. */ 3040: 3041: /* Note that there may be several formals lists which need to be modified 3042: due to the possibility that the return type of this function is a 3043: pointer-to-function type. If there are several formals lists, we 3044: convert them in left-to-right order here. */ 3045: 3046: #ifndef UNPROTOIZE 3047: this_f_list_chain_item = definition->f_list_chain; 3048: #endif /* !defined (UNPROTOIZE) */ 3049: 3050: for (;;) 3051: { 3052: { 3053: int depth; 3054: 3055: end_formals = start_formals + 1; 3056: depth = 1; 3057: for (; depth; check_source (++end_formals < clean_text_limit, 0)) 3058: { 3059: switch (*end_formals) 3060: { 3061: case '(': 3062: depth++; 3063: break; 3064: case ')': 3065: depth--; 3066: break; 3067: } 3068: } 3069: end_formals--; 3070: } 3071: 3072: /* end_formals now points to the closing right paren of the formals 3073: list whose left paren is pointed to by start_formals. */ 3074: 3075: /* Now, if we are protoizing, we insert the new ANSI-style formals list 3076: attached to the associated definition of this function. If however 3077: we are unprotoizing, then we simply delete any formals list which 3078: may be present. */ 3079: 3080: output_up_to (start_formals); 3081: #ifndef UNPROTOIZE 3082: if (this_f_list_chain_item) 3083: { 3084: output_string (this_f_list_chain_item->formals_list); 3085: this_f_list_chain_item = this_f_list_chain_item->chain_next; 3086: } 3087: else 3088: { 3089: if (!quiet_flag) 3090: fprintf (stderr, "%s: warning: too many parameter lists in declaration of `%s'\n", 3091: pname, def_dec_p->hash_entry->symbol); 3092: check_source (0, end_formals); /* leave the declaration intact */ 3093: } 3094: #endif /* !defined (UNPROTOIZE) */ 3095: clean_read_ptr = end_formals - 1; 3096: 3097: /* Now see if it looks like there may be another formals list associated 3098: with the function declaration that we are converting (following the 3099: formals list that we just converted. */ 3100: 3101: { 3102: const char *another_r_paren = forward_to_next_token_char (end_formals); 3103: 3104: if ((*another_r_paren != ')') 3105: || (*(start_formals = forward_to_next_token_char (another_r_paren)) != '(')) 3106: { 3107: #ifndef UNPROTOIZE 3108: if (this_f_list_chain_item) 3109: { 3110: if (!quiet_flag) 3111: fprintf (stderr, "\n%s: warning: too few parameter lists in declaration of `%s'\n", 3112: pname, def_dec_p->hash_entry->symbol); 3113: check_source (0, start_formals); /* leave the decl intact */ 3114: } 3115: #endif /* !defined (UNPROTOIZE) */ 3116: break; 3117: 3118: } 3119: } 3120: 3121: /* There does appear to be yet another formals list, so loop around 3122: again, and convert it also. */ 3123: } 3124: } 3125: 3126: /* Edit a whole group of formals lists, starting with the rightmost one 3127: from some set of formals lists. This routine is called once (from the 3128: outside) for each function declaration which is converted. It is 3129: recursive however, and it calls itself once for each remaining formal 3130: list that lies to the left of the one it was originally called to work 3131: on. Thus, a whole set gets done in right-to-left order. 3132: 3133: This routine returns non-zero if it thinks that it should not be trying 3134: to convert this particular function definition (because the name of the 3135: function doesn't match the one expected). */ 3136: 3137: static int 3138: edit_formals_lists (end_formals, f_list_count, def_dec_p) 3139: const char *end_formals; 3140: unsigned int f_list_count; 3141: const def_dec_info *def_dec_p; 3142: { 3143: const char *start_formals; 3144: int depth; 3145: 3146: start_formals = end_formals - 1; 3147: depth = 1; 3148: for (; depth; check_source (--start_formals > clean_read_ptr, 0)) 3149: { 3150: switch (*start_formals) 3151: { 3152: case '(': 3153: depth--; 3154: break; 3155: case ')': 3156: depth++; 3157: break; 3158: } 3159: } 3160: start_formals++; 3161: 3162: /* start_formals now points to the opening left paren of the formals list. */ 3163: 3164: f_list_count--; 3165: 3166: if (f_list_count) 3167: { 3168: const char *next_end; 3169: 3170: /* There should be more formal lists to the left of here. */ 3171: 3172: next_end = start_formals - 1; 3173: check_source (next_end > clean_read_ptr, 0); 3174: while (isspace (*next_end)) 3175: check_source (--next_end > clean_read_ptr, 0); 3176: check_source (*next_end == ')', next_end); 3177: check_source (--next_end > clean_read_ptr, 0); 3178: check_source (*next_end == ')', next_end); 3179: if (edit_formals_lists (next_end, f_list_count, def_dec_p)) 3180: return 1; 3181: } 3182: 3183: /* Check that the function name in the header we are working on is the same 3184: as the one we would expect to find. If not, issue a warning and return 3185: non-zero. */ 3186: 3187: if (f_list_count == 0) 3188: { 3189: const char *expected = def_dec_p->hash_entry->symbol; 3190: const char *func_name_start; 3191: const char *func_name_limit; 3192: size_t func_name_len; 3193: 3194: for (func_name_limit = start_formals-1; isspace (*func_name_limit); ) 3195: check_source (--func_name_limit > clean_read_ptr, 0); 3196: 3197: for (func_name_start = func_name_limit++; 3198: is_id_char (*func_name_start); 3199: func_name_start--) 3200: check_source (func_name_start > clean_read_ptr, 0); 3201: func_name_start++; 3202: func_name_len = func_name_limit - func_name_start; 3203: if (func_name_len == 0) 3204: check_source (0, func_name_start); 3205: if (func_name_len != strlen (expected) 3206: || strncmp (func_name_start, expected, func_name_len)) 3207: { 3208: fprintf (stderr, "%s: %d: warning: found `%s' but expected `%s'\n", 3209: shortpath (NULL, def_dec_p->file->hash_entry->symbol), 3210: identify_lineno (func_name_start), 3211: dupnstr (func_name_start, func_name_len), 3212: expected); 3213: return 1; 3214: } 3215: } 3216: 3217: output_up_to (start_formals); 3218: 3219: #ifdef UNPROTOIZE 3220: if (f_list_count == 0) 3221: output_string (def_dec_p->formal_names); 3222: #else /* !defined (UNPROTOIZE) */ 3223: { 3224: unsigned f_list_depth; 3225: const f_list_chain_item *flci_p = def_dec_p->f_list_chain; 3226: 3227: /* At this point, the current value of f_list count says how many 3228: links we have to follow through the f_list_chain to get to the 3229: particular formals list that we need to output next. */ 3230: 3231: for (f_list_depth = 0; f_list_depth < f_list_count; f_list_depth++) 3232: flci_p = flci_p->chain_next; 3233: output_string (flci_p->formals_list); 3234: } 3235: #endif /* !defined (UNPROTOIZE) */ 3236: 3237: clean_read_ptr = end_formals - 1; 3238: return 0; 3239: } 3240: 3241: /* Given a pointer to a byte in the clean text buffer which points to the 3242: beginning of a line that contains a "follower" token for a function 3243: definition header, do whatever is necessary to find the right closing 3244: paren for the rightmost formals list of the function definition header. 3245: */ 3246: 3247: static const char * 3248: find_rightmost_formals_list (clean_text_p) 3249: const char *clean_text_p; 3250: { 3251: const char *end_formals; 3252: 3253: /* We are editing a function definition. The line number we did a seek 3254: to contains the first token which immediately follows the entire set of 3255: formals lists which are part of this particular function definition 3256: header. 3257: 3258: Our job now is to scan leftwards in the clean text looking for the 3259: right-paren which is at the end of the function header's rightmost 3260: formals list. 3261: 3262: If we ignore whitespace, this right paren should be the first one we 3263: see which is (ignoring whitespace) immediately followed either by the 3264: open curly-brace beginning the function body or by an alphabetic 3265: character (in the case where the function definition is in old (K&R) 3266: style and there are some declarations of formal parameters). */ 3267: 3268: /* It is possible that the right paren we are looking for is on the 3269: current line (together with its following token). Just in case that 3270: might be true, we start out here by skipping down to the right end of 3271: the current line before starting our scan. */ 3272: 3273: for (end_formals = clean_text_p; *end_formals != '\n'; end_formals++) 3274: continue; 3275: end_formals--; 3276: 3277: #ifdef UNPROTOIZE 3278: 3279: /* Now scan backwards while looking for the right end of the rightmost 3280: formals list associated with this function definition. */ 3281: 3282: { 3283: char ch; 3284: const char *l_brace_p; 3285: 3286: /* Look leftward and try to find a right-paren. */ 3287: 3288: while (*end_formals != ')') 3289: { 3290: if (isspace (*end_formals)) 3291: while (isspace (*end_formals)) 3292: check_source (--end_formals > clean_read_ptr, 0); 3293: else 3294: check_source (--end_formals > clean_read_ptr, 0); 3295: } 3296: 3297: ch = *(l_brace_p = forward_to_next_token_char (end_formals)); 3298: /* Since we are unprotoizing an ANSI-style (prototyped) function 3299: definition, there had better not be anything (except whitespace) 3300: between the end of the ANSI formals list and the beginning of the 3301: function body (i.e. the '{'). */ 3302: 3303: check_source (ch == '{', l_brace_p); 3304: } 3305: 3306: #else /* !defined (UNPROTOIZE) */ 3307: 3308: /* Now scan backwards while looking for the right end of the rightmost 3309: formals list associated with this function definition. */ 3310: 3311: while (1) 3312: { 3313: char ch; 3314: const char *l_brace_p; 3315: 3316: /* Look leftward and try to find a right-paren. */ 3317: 3318: while (*end_formals != ')') 3319: { 3320: if (isspace (*end_formals)) 3321: while (isspace (*end_formals)) 3322: check_source (--end_formals > clean_read_ptr, 0); 3323: else 3324: check_source (--end_formals > clean_read_ptr, 0); 3325: } 3326: 3327: ch = *(l_brace_p = forward_to_next_token_char (end_formals)); 3328: 3329: /* Since it is possible that we found a right paren before the starting 3330: '{' of the body which IS NOT the one at the end of the real K&R 3331: formals list (say for instance, we found one embedded inside one of 3332: the old K&R formal parameter declarations) we have to check to be 3333: sure that this is in fact the right paren that we were looking for. 3334: 3335: The one we were looking for *must* be followed by either a '{' or 3336: by an alphabetic character, while others *cannot* legally be followed 3337: by such characters. */ 3338: 3339: if ((ch == '{') || isalpha (ch)) 3340: break; 3341: 3342: /* At this point, we have found a right paren, but we know that it is 3343: not the one we were looking for, so backup one character and keep 3344: looking. */ 3345: 3346: check_source (--end_formals > clean_read_ptr, 0); 3347: } 3348: 3349: #endif /* !defined (UNPROTOIZE) */ 3350: 3351: return end_formals; 3352: } 3353: 3354: #ifndef UNPROTOIZE 3355: 3356: /* Insert into the output file a totally new declaration for a function 3357: which (up until now) was being called from within the current block 3358: without having been declared at any point such that the declaration 3359: was visible (i.e. in scope) at the point of the call. 3360: 3361: We need to add in explicit declarations for all such function calls 3362: in order to get the full benefit of prototype-based function call 3363: parameter type checking. */ 3364: 3365: static void 3366: add_local_decl (def_dec_p, clean_text_p) 3367: const def_dec_info *def_dec_p; 3368: const char *clean_text_p; 3369: { 3370: const char *start_of_block; 3371: const char *function_to_edit = def_dec_p->hash_entry->symbol; 3372: 3373: /* Don't insert new local explicit declarations unless explicitly requested 3374: to do so. */ 3375: 3376: if (!local_flag) 3377: return; 3378: 3379: /* Setup here to recover from confusing source code detected during this 3380: particular "edit". */ 3381: 3382: save_pointers (); 3383: if (setjmp (source_confusion_recovery)) 3384: { 3385: restore_pointers (); 3386: fprintf (stderr, "%s: local declaration for function `%s' not inserted\n", 3387: pname, function_to_edit); 3388: return; 3389: } 3390: 3391: /* We have already done a seek to the start of the line which should 3392: contain *the* open curly brace which begins the block in which we need 3393: to insert an explicit function declaration (to replace the implicit one). 3394: 3395: Now we scan that line, starting from the left, until we find the 3396: open curly brace we are looking for. Note that there may actually be 3397: multiple open curly braces on the given line, but we will be happy 3398: with the leftmost one no matter what. */ 3399: 3400: start_of_block = clean_text_p; 3401: while (*start_of_block != '{' && *start_of_block != '\n') 3402: check_source (++start_of_block < clean_text_limit, 0); 3403: 3404: /* Note that the line from the original source could possibly 3405: contain *no* open curly braces! This happens if the line contains 3406: a macro call which expands into a chunk of text which includes a 3407: block (and that block's associated open and close curly braces). 3408: In cases like this, we give up, issue a warning, and do nothing. */ 3409: 3410: if (*start_of_block != '{') 3411: { 3412: if (!quiet_flag) 3413: fprintf (stderr, 3414: "\n%s: %d: warning: can't add declaration of `%s' into macro call\n", 3415: def_dec_p->file->hash_entry->symbol, def_dec_p->line, 3416: def_dec_p->hash_entry->symbol); 3417: return; 3418: } 3419: 3420: /* Figure out what a nice (pretty) indentation would be for the new 3421: declaration we are adding. In order to do this, we must scan forward 3422: from the '{' until we find the first line which starts with some 3423: non-whitespace characters (i.e. real "token" material). */ 3424: 3425: { 3426: const char *ep = forward_to_next_token_char (start_of_block) - 1; 3427: const char *sp; 3428: 3429: /* Now we have ep pointing at the rightmost byte of some existing indent 3430: stuff. At least that is the hope. 3431: 3432: We can now just scan backwards and find the left end of the existing 3433: indentation string, and then copy it to the output buffer. */ 3434: 3435: for (sp = ep; isspace (*sp) && *sp != '\n'; sp--) 3436: continue; 3437: 3438: /* Now write out the open { which began this block, and any following 3439: trash up to and including the last byte of the existing indent that 3440: we just found. */ 3441: 3442: output_up_to (ep); 3443: 3444: /* Now we go ahead and insert the new declaration at this point. 3445: 3446: If the definition of the given function is in the same file that we 3447: are currently editing, and if its full ANSI declaration normally 3448: would start with the keyword `extern', suppress the `extern'. */ 3449: 3450: { 3451: const char *decl = def_dec_p->definition->ansi_decl; 3452: 3453: if ((*decl == 'e') && (def_dec_p->file == def_dec_p->definition->file)) 3454: decl += 7; 3455: output_string (decl); 3456: } 3457: 1.1.1.2 ! root 3458: /* Finally, write out a new indent string, just like the preceding one 1.1 root 3459: that we found. This will typically include a newline as the first 3460: character of the indent string. */ 3461: 3462: output_bytes (sp, (size_t) (ep - sp) + 1); 3463: } 3464: } 3465: 3466: /* Given a pointer to a file_info record, and a pointer to the beginning 3467: of a line (in the clean text buffer) which is assumed to contain the 3468: first "follower" token for the first function definition header in the 3469: given file, find a good place to insert some new global function 3470: declarations (which will replace scattered and imprecise implicit ones) 3471: and then insert the new explicit declaration at that point in the file. */ 3472: 3473: static void 3474: add_global_decls (file_p, clean_text_p) 3475: const file_info *file_p; 3476: const char *clean_text_p; 3477: { 3478: const def_dec_info *dd_p; 3479: const char *scan_p; 3480: 3481: /* Setup here to recover from confusing source code detected during this 3482: particular "edit". */ 3483: 3484: save_pointers (); 3485: if (setjmp (source_confusion_recovery)) 3486: { 3487: restore_pointers (); 3488: fprintf (stderr, "%s: global declarations for file `%s' not inserted\n", 3489: pname, shortpath (NULL, file_p->hash_entry->symbol)); 3490: return; 3491: } 3492: 3493: /* Start by finding a good location for adding the new explicit function 3494: declarations. To do this, we scan backwards, ignoring whitespace 3495: and comments and other junk until we find either a semicolon, or until 3496: we hit the beginning of the file. */ 3497: 3498: scan_p = find_rightmost_formals_list (clean_text_p); 3499: for (;; --scan_p) 3500: { 3501: if (scan_p < clean_text_base) 3502: break; 3503: check_source (scan_p > clean_read_ptr, 0); 3504: if (*scan_p == ';') 3505: break; 3506: } 3507: 3508: /* scan_p now points either to a semicolon, or to just before the start 3509: of the whole file. */ 3510: 3511: /* Now scan forward for the first non-whitespace character. In theory, 3512: this should be the first character of the following function definition 3513: header. We will put in the added declarations just prior to that. */ 3514: 3515: scan_p++; 3516: while (isspace (*scan_p)) 3517: scan_p++; 3518: scan_p--; 3519: 3520: output_up_to (scan_p); 3521: 3522: /* Now write out full prototypes for all of the things that had been 3523: implicitly declared in this file (but only those for which we were 3524: actually able to find unique matching definitions). Avoid duplicates 3525: by marking things that we write out as we go. */ 3526: 3527: { 3528: int some_decls_added = 0; 3529: 3530: for (dd_p = file_p->defs_decs; dd_p; dd_p = dd_p->next_in_file) 3531: if (dd_p->is_implicit && dd_p->definition && !dd_p->definition->written) 3532: { 3533: const char *decl = dd_p->definition->ansi_decl; 3534: 3535: /* If the function for which we are inserting a declaration is 3536: actually defined later in the same file, then suppress the 3537: leading `extern' keyword (if there is one). */ 3538: 3539: if (*decl == 'e' && (dd_p->file == dd_p->definition->file)) 3540: decl += 7; 3541: 3542: output_string ("\n"); 3543: output_string (decl); 3544: some_decls_added = 1; 3545: ((NONCONST def_dec_info *) dd_p->definition)->written = 1; 3546: } 3547: if (some_decls_added) 3548: output_string ("\n\n"); 3549: } 3550: 3551: /* Unmark all of the definitions that we just marked. */ 3552: 3553: for (dd_p = file_p->defs_decs; dd_p; dd_p = dd_p->next_in_file) 3554: if (dd_p->definition) 3555: ((NONCONST def_dec_info *) dd_p->definition)->written = 0; 3556: } 3557: 3558: #endif /* !defined (UNPROTOIZE) */ 3559: 3560: /* Do the editing operation specifically for a function "definition". Note 3561: that editing operations for function "declarations" are handled by a 3562: separate routine above. */ 3563: 3564: static void 3565: edit_fn_definition (def_dec_p, clean_text_p) 3566: const def_dec_info *def_dec_p; 3567: const char *clean_text_p; 3568: { 3569: const char *end_formals; 3570: const char *function_to_edit = def_dec_p->hash_entry->symbol; 3571: 3572: /* Setup here to recover from confusing source code detected during this 3573: particular "edit". */ 3574: 3575: save_pointers (); 3576: if (setjmp (source_confusion_recovery)) 3577: { 3578: restore_pointers (); 3579: fprintf (stderr, "%s: definition of function `%s' not converted\n", 3580: pname, function_to_edit); 3581: return; 3582: } 3583: 3584: end_formals = find_rightmost_formals_list (clean_text_p); 3585: 3586: /* end_of_formals now points to the closing right paren of the rightmost 3587: formals list which is actually part of the `header' of the function 3588: definition that we are converting. */ 3589: 3590: /* If the header of this function definition looks like it declares a 3591: function with a variable number of arguments, and if the way it does 3592: that is different from that way we would like it (i.e. varargs vs. 3593: stdarg) then issue a warning and leave the header unconverted. */ 3594: 3595: if (other_variable_style_function (def_dec_p->ansi_decl)) 3596: { 3597: if (!quiet_flag) 3598: fprintf (stderr, "%s: %d: warning: definition of %s not converted\n", 3599: shortpath (NULL, def_dec_p->file->hash_entry->symbol), 3600: identify_lineno (end_formals), 3601: other_var_style); 3602: output_up_to (end_formals); 3603: return; 3604: } 3605: 3606: if (edit_formals_lists (end_formals, def_dec_p->f_list_count, def_dec_p)) 3607: { 3608: restore_pointers (); 3609: fprintf (stderr, "%s: definition of function `%s' not converted\n", 3610: pname, function_to_edit); 3611: return; 3612: } 3613: 3614: /* Have to output the last right paren because this never gets flushed by 3615: edit_formals_list. */ 3616: 3617: output_up_to (end_formals); 3618: 3619: #ifdef UNPROTOIZE 3620: { 3621: const char *decl_p; 3622: const char *semicolon_p; 3623: const char *limit_p; 3624: const char *scan_p; 3625: int had_newlines = 0; 3626: 3627: /* Now write out the K&R style formal declarations, one per line. */ 3628: 3629: decl_p = def_dec_p->formal_decls; 3630: limit_p = decl_p + strlen (decl_p); 3631: for (;decl_p < limit_p; decl_p = semicolon_p + 2) 3632: { 3633: for (semicolon_p = decl_p; *semicolon_p != ';'; semicolon_p++) 3634: continue; 3635: output_string ("\n"); 3636: output_string (indent_string); 3637: output_bytes (decl_p, (size_t) ((semicolon_p + 1) - decl_p)); 3638: } 3639: 3640: /* If there are no newlines between the end of the formals list and the 3641: start of the body, we should insert one now. */ 3642: 3643: for (scan_p = end_formals+1; *scan_p != '{'; ) 3644: { 3645: if (*scan_p == '\n') 3646: { 3647: had_newlines = 1; 3648: break; 3649: } 3650: check_source (++scan_p < clean_text_limit, 0); 3651: } 3652: if (!had_newlines) 3653: output_string ("\n"); 3654: } 3655: #else /* !defined (UNPROTOIZE) */ 3656: /* If we are protoizing, there may be some flotsum & jetsum (like comments 3657: and preprocessing directives) after the old formals list but before 3658: the following { and we would like to preserve that stuff while effectively 3659: deleting the existing K&R formal parameter declarations. We do so here 3660: in a rather tricky way. Basically, we white out any stuff *except* 3661: the comments/pp-directives in the original text buffer, then, if there 3662: is anything in this area *other* than whitespace, we output it. */ 3663: { 3664: const char *end_formals_orig; 3665: const char *start_body; 3666: const char *start_body_orig; 3667: const char *scan; 3668: const char *scan_orig; 3669: int have_flotsum = 0; 3670: int have_newlines = 0; 3671: 3672: for (start_body = end_formals + 1; *start_body != '{';) 3673: check_source (++start_body < clean_text_limit, 0); 3674: 3675: end_formals_orig = orig_text_base + (end_formals - clean_text_base); 3676: start_body_orig = orig_text_base + (start_body - clean_text_base); 3677: scan = end_formals + 1; 3678: scan_orig = end_formals_orig + 1; 3679: for (; scan < start_body; scan++, scan_orig++) 3680: { 3681: if (*scan == *scan_orig) 3682: { 3683: have_newlines |= (*scan_orig == '\n'); 3684: /* Leave identical whitespace alone. */ 3685: if (!isspace (*scan_orig)) 3686: *((NONCONST char *)scan_orig) = ' '; /* identical - so whiteout */ 3687: } 3688: else 3689: have_flotsum = 1; 3690: } 3691: if (have_flotsum) 3692: output_bytes (end_formals_orig + 1, 3693: (size_t) (start_body_orig - end_formals_orig) - 1); 3694: else 3695: if (have_newlines) 3696: output_string ("\n"); 3697: else 3698: output_string (" "); 3699: clean_read_ptr = start_body - 1; 3700: } 3701: #endif /* !defined (UNPROTOIZE) */ 3702: } 3703: 3704: /* Clean up the clean text buffer. Do this by converting comments and 3705: preprocessor directives into spaces. Also convert line continuations 3706: into whitespace. Also, whiteout string and character literals. */ 3707: 3708: static void 3709: do_cleaning (new_clean_text_base, new_clean_text_limit) 3710: char *new_clean_text_base; 3711: char *new_clean_text_limit; 3712: { 3713: char *scan_p; 3714: int non_whitespace_since_newline = 0; 3715: 3716: for (scan_p = new_clean_text_base; scan_p < new_clean_text_limit; scan_p++) 3717: { 3718: switch (*scan_p) 3719: { 3720: case '/': /* Handle comments. */ 3721: if (scan_p[1] != '*') 3722: goto regular; 3723: non_whitespace_since_newline = 1; 3724: scan_p[0] = ' '; 3725: scan_p[1] = ' '; 3726: scan_p += 2; 3727: while (scan_p[1] != '/' || scan_p[0] != '*') 3728: { 3729: if (!isspace (*scan_p)) 3730: *scan_p = ' '; 3731: if (++scan_p >= new_clean_text_limit) 3732: abort (); 3733: } 3734: *scan_p++ = ' '; 3735: *scan_p = ' '; 3736: break; 3737: 3738: case '#': /* Handle pp directives. */ 3739: if (non_whitespace_since_newline) 3740: goto regular; 3741: *scan_p = ' '; 3742: while (scan_p[1] != '\n' || scan_p[0] == '\\') 3743: { 3744: if (!isspace (*scan_p)) 3745: *scan_p = ' '; 3746: if (++scan_p >= new_clean_text_limit) 3747: abort (); 3748: } 3749: *scan_p++ = ' '; 3750: break; 3751: 3752: case '\'': /* Handle character literals. */ 3753: non_whitespace_since_newline = 1; 3754: while (scan_p[1] != '\'' || scan_p[0] == '\\') 3755: { 3756: if (scan_p[0] == '\\' && !isspace (scan_p[1])) 3757: scan_p[1] = ' '; 3758: if (!isspace (*scan_p)) 3759: *scan_p = ' '; 3760: if (++scan_p >= new_clean_text_limit) 3761: abort (); 3762: } 3763: *scan_p++ = ' '; 3764: break; 3765: 3766: case '"': /* Handle string literals. */ 3767: non_whitespace_since_newline = 1; 3768: while (scan_p[1] != '"' || scan_p[0] == '\\') 3769: { 3770: if (scan_p[0] == '\\' && !isspace (scan_p[1])) 3771: scan_p[1] = ' '; 3772: if (!isspace (*scan_p)) 3773: *scan_p = ' '; 3774: if (++scan_p >= new_clean_text_limit) 3775: abort (); 3776: } 3777: *scan_p++ = ' '; 3778: break; 3779: 3780: case '\\': /* Handle line continuations. */ 3781: if (scan_p[1] != '\n') 3782: goto regular; 3783: *scan_p = ' '; 3784: break; 3785: 3786: case '\n': 3787: non_whitespace_since_newline = 0; /* Reset. */ 3788: break; 3789: 3790: case ' ': 3791: case '\v': 3792: case '\t': 3793: case '\r': 3794: case '\f': 3795: case '\b': 3796: break; /* Whitespace characters. */ 3797: 3798: default: 3799: regular: 3800: non_whitespace_since_newline = 1; 3801: break; 3802: } 3803: } 3804: } 3805: 3806: /* Given a pointer to the closing right parenthesis for a particular formals 3807: list (in the clean text buffer) find the corresponding left parenthesis 3808: and return a pointer to it. */ 3809: 3810: static const char * 3811: careful_find_l_paren (p) 3812: const char *p; 3813: { 3814: const char *q; 3815: int paren_depth; 3816: 3817: for (paren_depth = 1, q = p-1; paren_depth; check_source (--q >= clean_text_base, 0)) 3818: { 3819: switch (*q) 3820: { 3821: case ')': 3822: paren_depth++; 3823: break; 3824: case '(': 3825: paren_depth--; 3826: break; 3827: } 3828: } 3829: return ++q; 3830: } 3831: 3832: /* Scan the clean text buffer for cases of function definitions that we 3833: don't really know about because they were preprocessed out when the 3834: aux info files were created. 3835: 3836: In this version of protoize/unprotoize we just give a warning for each 3837: one found. A later version may be able to at least unprotoize such 3838: missed items. 3839: 3840: Note that we may easily find all function definitions simply by 3841: looking for places where there is a left paren which is (ignoring 3842: whitespace) immediately followed by either a left-brace or by an 3843: upper or lower case letter. Whenever we find this combination, we 3844: have also found a function definition header. 3845: 3846: Finding function *declarations* using syntactic clues is much harder. 3847: I will probably try to do this in a later version though. */ 3848: 3849: static void 3850: scan_for_missed_items (file_p) 3851: const file_info *file_p; 3852: { 3853: static const char *scan_p; 3854: const char *limit = clean_text_limit - 3; 3855: static const char *backup_limit; 3856: 3857: backup_limit = clean_text_base - 1; 3858: 3859: for (scan_p = clean_text_base; scan_p < limit; scan_p++) 3860: { 3861: if (*scan_p == ')') 3862: { 3863: static const char *last_r_paren; 3864: const char *ahead_p; 3865: 3866: last_r_paren = scan_p; 3867: 3868: for (ahead_p = scan_p + 1; isspace (*ahead_p); ) 3869: check_source (++ahead_p < limit, limit); 3870: 3871: scan_p = ahead_p - 1; 3872: 3873: if (isalpha (*ahead_p) || *ahead_p == '{') 3874: { 3875: const char *last_l_paren; 3876: const int lineno = identify_lineno (ahead_p); 3877: 3878: if (setjmp (source_confusion_recovery)) 3879: continue; 3880: 3881: /* We know we have a function definition header. Now skip 3882: leftwards over all of its associated formals lists. */ 3883: 3884: do 3885: { 3886: last_l_paren = careful_find_l_paren (last_r_paren); 3887: for (last_r_paren = last_l_paren-1; isspace (*last_r_paren); ) 3888: check_source (--last_r_paren >= backup_limit, backup_limit); 3889: } 3890: while (*last_r_paren == ')'); 3891: 3892: if (is_id_char (*last_r_paren)) 3893: { 3894: const char *id_limit = last_r_paren + 1; 3895: const char *id_start; 3896: size_t id_length; 3897: const def_dec_info *dd_p; 3898: 3899: for (id_start = id_limit-1; is_id_char (*id_start); ) 3900: check_source (--id_start >= backup_limit, backup_limit); 3901: id_start++; 3902: backup_limit = id_start; 3903: if ((id_length = (size_t) (id_limit - id_start)) == 0) 3904: goto not_missed; 3905: 3906: { 3907: char *func_name = (char *) alloca (id_length + 1); 3908: static const char * const stmt_keywords[] 3909: = { "if", "while", "for", "switch", "return", 0 }; 3910: const char * const *stmt_keyword; 3911: 3912: strncpy (func_name, id_start, id_length); 3913: func_name[id_length] = '\0'; 3914: 3915: /* We must check here to see if we are actually looking at 3916: a statement rather than an actual function call. */ 3917: 3918: for (stmt_keyword = stmt_keywords; *stmt_keyword; stmt_keyword++) 3919: if (!strcmp (func_name, *stmt_keyword)) 3920: goto not_missed; 3921: 3922: #if 0 3923: fprintf (stderr, "%s: found definition of `%s' at %s(%d)\n", 3924: pname, 3925: func_name, 3926: shortpath (NULL, file_p->hash_entry->symbol), 3927: identify_lineno (id_start)); 3928: #endif /* 0 */ 3929: /* We really should check for a match of the function name 3930: here also, but why bother. */ 3931: 3932: for (dd_p = file_p->defs_decs; dd_p; dd_p = dd_p->next_in_file) 3933: if (dd_p->is_func_def && dd_p->line == lineno) 3934: goto not_missed; 3935: 3936: /* If we make it here, then we did not know about this 3937: function definition. */ 3938: 3939: fprintf (stderr, "%s: %d: warning: `%s' was #if 0\n", 3940: shortpath (NULL, file_p->hash_entry->symbol), 3941: identify_lineno (id_start), func_name); 3942: fprintf (stderr, "%s: function definition not converted\n", 3943: pname); 3944: } 3945: not_missed: ; 3946: } 3947: } 3948: } 3949: } 3950: } 3951: 3952: /* Do all editing operations for a single source file (either a "base" file 3953: or an "include" file). To do this we read the file into memory, keep a 3954: virgin copy there, make another cleaned in-core copy of the original file 3955: (i.e. one in which all of the comments and preprocessor directives have 3956: been replaced with whitespace), then use these two in-core copies of the 3957: file to make a new edited in-core copy of the file. Finally, rename the 3958: original file (as a way of saving it), and then write the edited version 3959: of the file from core to a disk file of the same name as the original. 3960: 3961: Note that the trick of making a copy of the original sans comments & 3962: preprocessor directives make the editing a whole lot easier. */ 3963: 3964: static void 3965: edit_file (hp) 3966: const hash_table_entry *hp; 3967: { 3968: struct stat stat_buf; 3969: const file_info *file_p = hp->fip; 3970: char *new_orig_text_base; 3971: char *new_orig_text_limit; 3972: char *new_clean_text_base; 3973: char *new_clean_text_limit; 3974: size_t orig_size; 3975: size_t repl_size; 3976: int first_definition_in_file; 3977: 3978: /* If we are not supposed to be converting this file, or if there is 3979: nothing in there which needs converting, just skip this file. */ 3980: 3981: if (!needs_to_be_converted (file_p)) 3982: return; 3983: 3984: convert_filename = file_p->hash_entry->symbol; 3985: 3986: /* Convert a file if it is in a directory where we want conversion 3987: and the file is not excluded. */ 3988: 3989: if (!directory_specified_p (convert_filename) 3990: || file_excluded_p (convert_filename)) 3991: { 3992: if (!quiet_flag 3993: #ifdef UNPROTOIZE 3994: /* Don't even mention "system" include files unless we are 3995: protoizing. If we are protoizing, we mention these as a 1.1.1.2 ! root 3996: gentle way of prodding the user to convert his "system" 1.1 root 3997: include files to prototype format. */ 3998: && !in_system_include_dir (convert_filename) 3999: #endif /* defined (UNPROTOIZE) */ 4000: ) 4001: fprintf (stderr, "%s: file `%s' not converted\n", 4002: pname, shortpath (NULL, convert_filename)); 4003: return; 4004: } 4005: 4006: /* Let the user know what we are up to. */ 4007: 4008: if (nochange_flag) 4009: fprintf (stderr, "%s: would convert file `%s'\n", 4010: pname, shortpath (NULL, convert_filename)); 4011: else 4012: fprintf (stderr, "%s: converting file `%s'\n", 4013: pname, shortpath (NULL, convert_filename)); 4014: fflush (stderr); 4015: 4016: /* Find out the size (in bytes) of the original file. */ 4017: 4018: /* The cast avoids an erroneous warning on AIX. */ 4019: if (my_stat ((char *)convert_filename, &stat_buf) == -1) 4020: { 4021: fprintf (stderr, "%s: error: can't get status for file `%s': %s\n", 4022: pname, shortpath (NULL, convert_filename), sys_errlist[errno]); 4023: return; 4024: } 4025: orig_size = stat_buf.st_size; 4026: 4027: /* Allocate a buffer to hold the original text. */ 4028: 4029: orig_text_base = new_orig_text_base = (char *) xmalloc (orig_size + 2); 4030: orig_text_limit = new_orig_text_limit = new_orig_text_base + orig_size; 4031: 4032: /* Allocate a buffer to hold the cleaned-up version of the original text. */ 4033: 4034: clean_text_base = new_clean_text_base = (char *) xmalloc (orig_size + 2); 4035: clean_text_limit = new_clean_text_limit = new_clean_text_base + orig_size; 4036: clean_read_ptr = clean_text_base - 1; 4037: 4038: /* Allocate a buffer that will hopefully be large enough to hold the entire 4039: converted output text. As an initial guess for the maximum size of the 4040: output buffer, use 125% of the size of the original + some extra. This 4041: buffer can be expanded later as needed. */ 4042: 4043: repl_size = orig_size + (orig_size >> 2) + 4096; 4044: repl_text_base = (char *) xmalloc (repl_size + 2); 4045: repl_text_limit = repl_text_base + repl_size - 1; 4046: repl_write_ptr = repl_text_base - 1; 4047: 4048: { 4049: int input_file; 4050: 4051: /* Open the file to be converted in READ ONLY mode. */ 4052: 4053: if ((input_file = my_open (convert_filename, O_RDONLY, 0444)) == -1) 4054: { 4055: fprintf (stderr, "%s: error: can't open file `%s' for reading: %s\n", 4056: pname, shortpath (NULL, convert_filename), 4057: sys_errlist[errno]); 4058: return; 4059: } 4060: 4061: /* Read the entire original source text file into the original text buffer 4062: in one swell fwoop. Then figure out where the end of the text is and 4063: make sure that it ends with a newline followed by a null. */ 4064: 4065: if (read (input_file, new_orig_text_base, orig_size) != orig_size) 4066: { 4067: close (input_file); 4068: fprintf (stderr, "\n%s: error: while reading input file `%s': %s\n", 4069: pname, shortpath (NULL, convert_filename), 4070: sys_errlist[errno]); 4071: return; 4072: } 4073: 4074: close (input_file); 4075: } 4076: 4077: if (orig_size == 0 || orig_text_limit[-1] != '\n') 4078: { 4079: *new_orig_text_limit++ = '\n'; 4080: orig_text_limit++; 4081: } 4082: 4083: /* Create the cleaned up copy of the original text. */ 4084: 4085: memcpy (new_clean_text_base, orig_text_base, 4086: (size_t) (orig_text_limit - orig_text_base)); 4087: do_cleaning (new_clean_text_base, new_clean_text_limit); 4088: 4089: #if 0 4090: { 4091: int clean_file; 4092: size_t clean_size = orig_text_limit - orig_text_base; 4093: char *const clean_filename = (char *) alloca (strlen (convert_filename) + 6 + 1); 4094: 4095: /* Open (and create) the clean file. */ 4096: 4097: strcpy (clean_filename, convert_filename); 4098: strcat (clean_filename, ".clean"); 4099: if ((clean_file = creat (clean_filename, 0666)) == -1) 4100: { 4101: fprintf (stderr, "%s: error: can't create/open clean file `%s': %s\n", 4102: pname, shortpath (NULL, clean_filename), 4103: sys_errlist[errno]); 4104: return; 4105: } 4106: 4107: /* Write the clean file. */ 4108: 4109: if (write (clean_file, new_clean_text_base, clean_size) != clean_size) 4110: fprintf (stderr, "%s: error: while writing file `%s': %s\n", 4111: pname, shortpath (NULL, clean_filename), sys_errlist[errno]); 4112: 4113: close (clean_file); 4114: } 4115: #endif /* 0 */ 4116: 4117: /* Do a simplified scan of the input looking for things that were not 4118: mentioned in the aux info files because of the fact that they were 4119: in a region of the source which was preprocessed-out (via #if or 4120: via #ifdef). */ 4121: 4122: scan_for_missed_items (file_p); 4123: 4124: /* Setup to do line-oriented forward seeking in the clean text buffer. */ 4125: 4126: last_known_line_number = 1; 4127: last_known_line_start = clean_text_base; 4128: 4129: /* Now get down to business and make all of the necessary edits. */ 4130: 4131: { 4132: const def_dec_info *def_dec_p; 4133: 4134: first_definition_in_file = 1; 4135: def_dec_p = file_p->defs_decs; 4136: for (; def_dec_p; def_dec_p = def_dec_p->next_in_file) 4137: { 4138: const char *clean_text_p = seek_to_line (def_dec_p->line); 4139: 4140: /* clean_text_p now points to the first character of the line which 4141: contains the `terminator' for the declaration or definition that 4142: we are about to process. */ 4143: 4144: #ifndef UNPROTOIZE 4145: 4146: if (global_flag && def_dec_p->is_func_def && first_definition_in_file) 4147: { 4148: add_global_decls (def_dec_p->file, clean_text_p); 4149: first_definition_in_file = 0; 4150: } 4151: 4152: /* Don't edit this item if it is already in prototype format or if it 4153: is a function declaration and we have found no corresponding 4154: definition. */ 4155: 4156: if (def_dec_p->prototyped 4157: || (!def_dec_p->is_func_def && !def_dec_p->definition)) 4158: continue; 4159: 4160: #endif /* !defined (UNPROTOIZE) */ 4161: 4162: if (def_dec_p->is_func_def) 4163: edit_fn_definition (def_dec_p, clean_text_p); 4164: else 4165: #ifndef UNPROTOIZE 4166: if (def_dec_p->is_implicit) 4167: add_local_decl (def_dec_p, clean_text_p); 4168: else 4169: #endif /* !defined (UNPROTOIZE) */ 4170: edit_fn_declaration (def_dec_p, clean_text_p); 4171: } 4172: } 4173: 4174: /* Finalize things. Output the last trailing part of the original text. */ 4175: 4176: output_up_to (clean_text_limit - 1); 4177: 4178: /* If this is just a test run, stop now and just deallocate the buffers. */ 4179: 4180: if (nochange_flag) 4181: { 4182: free (new_orig_text_base); 4183: free (new_clean_text_base); 4184: free (repl_text_base); 4185: return; 4186: } 4187: 4188: /* Change the name of the original input file. This is just a quick way of 4189: saving the original file. */ 4190: 4191: if (!nosave_flag) 4192: { 4193: char *new_filename = 4194: (char *) xmalloc (strlen (convert_filename) + strlen (save_suffix) + 2); 4195: 4196: strcpy (new_filename, convert_filename); 4197: strcat (new_filename, save_suffix); 4198: if (my_link (convert_filename, new_filename) == -1) 4199: { 4200: if (errno == EEXIST) 4201: { 4202: if (!quiet_flag) 4203: fprintf (stderr, "%s: warning: file `%s' already saved in `%s'\n", 4204: pname, 4205: shortpath (NULL, convert_filename), 4206: shortpath (NULL, new_filename)); 4207: } 4208: else 4209: { 4210: fprintf (stderr, "%s: error: can't link file `%s' to `%s': %s\n", 4211: pname, 4212: shortpath (NULL, convert_filename), 4213: shortpath (NULL, new_filename), 4214: sys_errlist[errno]); 4215: return; 4216: } 4217: } 4218: } 4219: 4220: if (my_unlink (convert_filename) == -1) 4221: { 4222: fprintf (stderr, "%s: error: can't delete file `%s': %s\n", 4223: pname, shortpath (NULL, convert_filename), sys_errlist[errno]); 4224: return; 4225: } 4226: 4227: { 4228: int output_file; 4229: 4230: /* Open (and create) the output file. */ 4231: 4232: if ((output_file = creat (convert_filename, 0666)) == -1) 4233: { 4234: fprintf (stderr, "%s: error: can't create/open output file `%s': %s\n", 4235: pname, shortpath (NULL, convert_filename), 4236: sys_errlist[errno]); 4237: return; 4238: } 4239: 4240: /* Write the output file. */ 4241: 4242: { 4243: unsigned int out_size = (repl_write_ptr + 1) - repl_text_base; 4244: 4245: if (write (output_file, repl_text_base, out_size) != out_size) 4246: fprintf (stderr, "%s: error: while writing file `%s': %s\n", 4247: pname, shortpath (NULL, convert_filename), 4248: sys_errlist[errno]); 4249: } 4250: 4251: close (output_file); 4252: } 4253: 4254: /* Deallocate the conversion buffers. */ 4255: 4256: free (new_orig_text_base); 4257: free (new_clean_text_base); 4258: free (repl_text_base); 4259: 4260: /* Change the mode of the output file to match the original file. */ 4261: 4262: /* The cast avoids an erroneous warning on AIX. */ 4263: if (my_chmod ((char *)convert_filename, stat_buf.st_mode) == -1) 4264: fprintf (stderr, "%s: error: can't change mode of file `%s': %s\n", 4265: pname, shortpath (NULL, convert_filename), sys_errlist[errno]); 4266: 4267: /* Note: We would try to change the owner and group of the output file 4268: to match those of the input file here, except that may not be a good 4269: thing to do because it might be misleading. Also, it might not even 4270: be possible to do that (on BSD systems with quotas for instance). */ 4271: } 4272: 4273: /* Do all of the individual steps needed to do the protoization (or 4274: unprotoization) of the files referenced in the aux_info files given 4275: in the command line. */ 4276: 4277: static void 4278: do_processing () 4279: { 4280: const char * const *base_pp; 4281: const char * const * const end_pps 4282: = &base_source_filenames[n_base_source_files]; 4283: 4284: #ifndef UNPROTOIZE 4285: int syscalls_len; 4286: #endif /* !defined (UNPROTOIZE) */ 4287: 4288: /* One-by-one, check (and create if necessary), open, and read all of the 4289: stuff in each aux_info file. After reading each aux_info file, the 4290: aux_info_file just read will be automatically deleted unless the 4291: keep_flag is set. */ 4292: 4293: for (base_pp = base_source_filenames; base_pp < end_pps; base_pp++) 4294: process_aux_info_file (*base_pp, keep_flag, 0); 4295: 4296: #ifndef UNPROTOIZE 4297: 4298: /* Also open and read the special SYSCALLS.c aux_info file which gives us 4299: the prototypes for all of the standard system-supplied functions. */ 4300: 4301: if (nondefault_syscalls_dir) 4302: { 4303: syscalls_absolute_filename 4304: = (char *) xmalloc (strlen (nondefault_syscalls_dir) 1.1.1.2 ! root 4305: + sizeof (syscalls_filename) + 1); 1.1 root 4306: strcpy (syscalls_absolute_filename, nondefault_syscalls_dir); 4307: } 4308: else 4309: { 4310: syscalls_absolute_filename 4311: = (char *) xmalloc (strlen (default_syscalls_dir) 1.1.1.2 ! root 4312: + sizeof (syscalls_filename) + 1); 1.1 root 4313: strcpy (syscalls_absolute_filename, default_syscalls_dir); 4314: } 4315: 4316: syscalls_len = strlen (syscalls_absolute_filename); 4317: if (*(syscalls_absolute_filename + syscalls_len - 1) != '/') 4318: { 4319: *(syscalls_absolute_filename + syscalls_len++) = '/'; 4320: *(syscalls_absolute_filename + syscalls_len) = '\0'; 4321: } 4322: strcat (syscalls_absolute_filename, syscalls_filename); 4323: 4324: /* Call process_aux_info_file in such a way that it does not try to 4325: delete the SYSCALLS aux_info file. */ 4326: 4327: process_aux_info_file (syscalls_absolute_filename, 1, 1); 4328: 4329: #endif /* !defined (UNPROTOIZE) */ 4330: 4331: /* When we first read in all of the information from the aux_info files 4332: we saved in it decending line number order, because that was likely to 4333: be faster. Now however, we want the chains of def & dec records to 4334: appear in ascending line number order as we get further away from the 4335: file_info record that they hang from. The following line causes all of 4336: these lists to be rearranged into ascending line number order. */ 4337: 4338: visit_each_hash_node (filename_primary, reverse_def_dec_list); 4339: 4340: #ifndef UNPROTOIZE 4341: 4342: /* Now do the "real" work. The following line causes each declaration record 4343: to be "visited". For each of these nodes, an attempt is made to match 4344: up the function declaration with a corresponding function definition, 4345: which should have a full prototype-format formals list with it. Once 4346: these match-ups are made, the conversion of the function declarations 4347: to prototype format can be made. */ 4348: 4349: visit_each_hash_node (function_name_primary, connect_defs_and_decs); 4350: 4351: #endif /* !defined (UNPROTOIZE) */ 4352: 4353: /* Now convert each file that can be converted (and needs to be). */ 4354: 4355: visit_each_hash_node (filename_primary, edit_file); 4356: 4357: #ifndef UNPROTOIZE 4358: 4359: /* If we are working in cplusplus mode, try to rename all .c files to .C 4360: files. Don't panic if some of the renames don't work. */ 4361: 4362: if (cplusplus_flag && !nochange_flag) 4363: visit_each_hash_node (filename_primary, rename_c_file); 4364: 4365: #endif /* !defined (UNPROTOIZE) */ 4366: } 4367: 4368: static struct option longopts[] = 4369: { 4370: {"version", 0, 0, 'V'}, 4371: {"file_name", 0, 0, 'p'}, 4372: {"quiet", 0, 0, 'q'}, 4373: {"silent", 0, 0, 'q'}, 4374: {"force", 0, 0, 'f'}, 4375: {"keep", 0, 0, 'k'}, 4376: {"nosave", 0, 0, 'N'}, 4377: {"nochange", 0, 0, 'n'}, 4378: {"compiler-options", 1, 0, 'c'}, 4379: {"exclude", 1, 0, 'x'}, 4380: {"directory", 1, 0, 'd'}, 4381: #ifdef UNPROTOIZE 4382: {"indent", 1, 0, 'i'}, 4383: #else 4384: {"local", 0, 0, 'l'}, 4385: {"global", 0, 0, 'g'}, 4386: {"c++", 0, 0, 'C'}, 4387: {"syscalls-dir", 1, 0, 'B'}, 4388: #endif 4389: {0, 0, 0, 0} 4390: }; 4391: 4392: int 4393: main (argc, argv) 4394: int argc; 4395: char **const argv; 4396: { 4397: int longind; 4398: int c; 1.1.1.2 ! root 4399: char *params = ""; 1.1 root 4400: 4401: pname = strrchr (argv[0], '/'); 4402: pname = pname ? pname+1 : argv[0]; 4403: 1.1.1.2 ! root 4404: cwd_buffer = getpwd (); ! 4405: if (!cwd_buffer) 1.1 root 4406: { 1.1.1.2 ! root 4407: fprintf (stderr, "%s: cannot get working directory: %s\n", ! 4408: pname, sys_errlist[errno]); ! 4409: exit (1); 1.1 root 4410: } 4411: 4412: /* By default, convert the files in the current directory. */ 4413: directory_list = string_list_cons (cwd_buffer, NULL); 4414: 4415: while ((c = getopt_long (argc, argv, 4416: #ifdef UNPROTOIZE 4417: "c:d:i:knNp:qVx:", 4418: #else 4419: "B:c:Cd:gklnNp:qVx:", 4420: #endif 4421: longopts, &longind)) != EOF) 4422: { 4423: if (c == 0) /* Long option. */ 4424: c = longopts[longind].val; 4425: switch (c) 4426: { 4427: case 'p': 4428: compiler_file_name = optarg; 4429: break; 4430: case 'd': 4431: directory_list 4432: = string_list_cons (abspath (NULL, optarg), directory_list); 4433: break; 4434: case 'x': 4435: exclude_list = string_list_cons (optarg, exclude_list); 4436: break; 4437: 4438: case 'V': 4439: version_flag = 1; 4440: break; 4441: case 'q': 4442: quiet_flag = 1; 4443: break; 4444: #if 0 4445: case 'f': 4446: force_flag = 1; 4447: break; 4448: #endif 4449: case 'n': 4450: nochange_flag = 1; 4451: keep_flag = 1; 4452: break; 4453: case 'N': 4454: nosave_flag = 1; 4455: break; 4456: case 'k': 4457: keep_flag = 1; 4458: break; 4459: case 'c': 1.1.1.2 ! root 4460: params = optarg; 1.1 root 4461: break; 4462: #ifdef UNPROTOIZE 4463: case 'i': 4464: indent_string = optarg; 4465: break; 4466: #else /* !defined (UNPROTOIZE) */ 4467: case 'l': 4468: local_flag = 1; 4469: break; 4470: case 'g': 4471: global_flag = 1; 4472: break; 4473: case 'C': 4474: cplusplus_flag = 1; 4475: break; 4476: case 'B': 4477: nondefault_syscalls_dir = optarg; 4478: break; 4479: #endif /* !defined (UNPROTOIZE) */ 4480: default: 4481: usage (); 4482: } 4483: } 4484: 1.1.1.2 ! root 4485: /* Set up compile_params based on -p and -c options. */ ! 4486: munge_compile_params (params); ! 4487: 1.1 root 4488: n_base_source_files = argc - optind; 4489: 4490: /* Now actually make a list of the base source filenames. */ 4491: 4492: base_source_filenames = 4493: (const char **) xmalloc ((n_base_source_files + 1) * sizeof (char *)); 4494: n_base_source_files = 0; 4495: for (; optind < argc; optind++) 4496: { 4497: const char *path = abspath (NULL, argv[optind]); 4498: int len = strlen (path); 4499: 4500: if (path[len-1] == 'c' && path[len-2] == '.') 4501: base_source_filenames[n_base_source_files++] = path; 4502: else 4503: { 4504: fprintf (stderr, "%s: input file names must have .c suffixes: %s\n", 4505: pname, shortpath (NULL, path)); 4506: errors++; 4507: } 4508: } 4509: 4510: #ifndef UNPROTOIZE 4511: /* We are only interested in the very first identifier token in the 4512: definition of `va_list', so if there is more junk after that first 4513: identifier token, delete it from the `varargs_style_indicator'. */ 4514: { 4515: const char *cp; 4516: 4517: for (cp = varargs_style_indicator; isalnum (*cp) || *cp == '_'; cp++) 4518: continue; 4519: if (*cp != 0) 4520: varargs_style_indicator = savestring (varargs_style_indicator, 4521: cp - varargs_style_indicator); 4522: } 4523: #endif /* !defined (UNPROTOIZE) */ 4524: 4525: if (errors) 4526: usage (); 4527: else 4528: { 4529: if (version_flag) 4530: fprintf (stderr, "%s: %s\n", pname, version_string); 4531: do_processing (); 4532: } 4533: if (errors) 4534: exit (1); 4535: else 4536: exit (0); 4537: return 1; 4538: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.