busybox/scripts/bb_mkdep.c

1789 lines
43 KiB
C

/* vi: set sw=4 ts=4: */
/*
* Another fast dependencies generator for Makefiles, Version 4.2
*
* Copyright (C) 2005,2006 by Vladimir Oleynik <dzo@simtreas.ru>
*
* mmaping file may be originally by Linus Torvalds.
*
* infix parser/evaluator for #if expression
* Copyright (c) 2001 Aaron Lehmann <aaronl@vitelus.com>
* Copyright (c) 2001 Manuel Novoa III <mjn3@codepoet.org>
* Copyright (c) 2003 Vladimir Oleynik <dzo@simtreas.ru>
*
* bb_simplify_path()
* Copyright (C) 2005 Manuel Novoa III <mjn3@codepoet.org>
*
* xmalloc() bb_xstrdup() bb_error_d():
* Copyright (C) 1999-2004 by Erik Andersen <andersen@codepoet.org>
*
* llist routine
* Copyright (C) 2003 Glenn McGrath
* Copyright (C) Vladimir Oleynik <dzo@simtreas.ru>
*
* (c) 2005,2006 Bernhard Fischer:
* - commentary typos,
* - move "memory exhausted" into msg_enomem,
* - more verbose --help output.
*
* This program does:
* 1) find #define KEY VALUE or #undef KEY from include/bb_config.h
* 2) recursive find and scan *.[ch] files, but skips scan of include/config/
* 3) find #include "*.h" and KEYs using, if not as #define and #undef
* 4) generate dependencies to stdout
* pwd/file.o: include/config/key*.h found_include_*.h
* path/inc.h: include/config/key*.h found_included_include_*.h
* 5) save include/config/key*.h if changed after previous usage
* This program does not generate dependencies for #include <...>
* Config file can have #if #elif #else #ifdef #ifndef #endif lines
*/
#define LOCAL_INCLUDE_PATH "include"
#define INCLUDE_CONFIG_PATH LOCAL_INCLUDE_PATH"/config"
#define INCLUDE_CONFIG_KEYS_PATH LOCAL_INCLUDE_PATH"/bb_config.h"
#define bb_mkdep_full_options \
"\nOptions:" \
"\n\t-I local_include_path include paths, default: \"" LOCAL_INCLUDE_PATH "\"" \
"\n\t-d don't generate depend" \
"\n\t-w show warning if include files not found" \
"\n\t-k include/config default: \"" INCLUDE_CONFIG_PATH "\"" \
"\n\t-c include/config.h configs, default: \"" INCLUDE_CONFIG_KEYS_PATH "\"" \
"\n\tdirs_to_scan default \".\""
#define bb_mkdep_terse_options "Usage: [-I local_include_paths] [-dw] " \
"[-k path_for_stored_keys] [dirs]"
#define _GNU_SOURCE
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/mman.h>
#include <getopt.h>
#include <dirent.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdarg.h>
#include <unistd.h>
#include <errno.h>
#include <fcntl.h>
#include <limits.h>
#ifdef __GNUC__
#define ATTRIBUTE __attribute__
#else
#define ATTRIBUTE(a) /* nothing */
#endif
#if !(defined __USE_ISOC99 || (defined __GLIBC_HAVE_LONG_LONG && defined __USE_MISC))
#define strtoll strtol
#endif
/* partial and simplified libbb routine */
static void bb_error_d(const char *s, ...) ATTRIBUTE ((noreturn, format (printf, 1, 2)));
static char * bb_asprint(const char *format, ...) ATTRIBUTE ((format (printf, 1, 2)));
static char *bb_simplify_path(const char *path);
/* stolen from libbb as is */
typedef struct llist_s {
char *data;
struct llist_s *link;
} llist_t;
static const char msg_enomem[] = "memory exhausted";
/* inline realization for fast works */
static inline void *xmalloc(size_t size)
{
void *p = malloc(size);
if(p == NULL)
bb_error_d(msg_enomem);
return p;
}
static inline char *bb_xstrdup(const char *s)
{
char *r = strdup(s);
if(r == NULL)
bb_error_d(msg_enomem);
return r;
}
static int dontgenerate_dep; /* flag -d usaged */
static int noiwarning; /* flag -w usaged */
static llist_t *configs; /* list of -c usaged and them stat() after parsed */
static llist_t *Iop; /* list of -I include usaged */
static char *pwd; /* current work directory */
static size_t replace; /* replace current work directory with build dir */
static const char *kp; /* KEY path, argument of -k used */
static size_t kp_len;
static struct stat st_kp; /* stat(kp) */
typedef struct BB_KEYS {
char *keyname;
size_t key_sz;
const char *value;
const char *checked;
char *stored_path;
const char *src_have_this_key;
struct BB_KEYS *next;
} bb_key_t;
static bb_key_t *key_top; /* list of found KEYs */
static bb_key_t *Ifound; /* list of parsed includes */
static void parse_conf_opt(const char *opt, const char *val, size_t key_sz);
static void parse_inc(const char *include, const char *fname);
static inline bb_key_t *check_key(bb_key_t *k, const char *nk, size_t key_sz)
{
bb_key_t *cur;
for(cur = k; cur; cur = cur->next) {
if(key_sz == cur->key_sz && memcmp(cur->keyname, nk, key_sz) == 0) {
cur->checked = cur->stored_path;
return cur;
}
}
return NULL;
}
static inline const char *lookup_key(const char *nk, size_t key_sz)
{
bb_key_t *cur;
for(cur = key_top; cur; cur = cur->next) {
if(key_sz == cur->key_sz && memcmp(cur->keyname, nk, key_sz) == 0) {
return cur->value;
}
}
return NULL;
}
/* for lexical analyser */
static int pagesizem1; /* padding mask = getpagesize() - 1 */
/* for speed tricks */
static char first_chars[1+UCHAR_MAX]; /* + L_EOF */
static char isalnums[1+UCHAR_MAX]; /* + L_EOF */
/* trick for fast find "define", "include", "undef",
"if((n)def)" "else", "endif" */
static const char * const preproc[] = {
/* 0 1 2 3 4 5 6 7 8 9 */
"", "efine", "lif", "lse", "ndif", "f", "fdef", "fndef", "nclude", "ndef" };
static const unsigned char first_chars_deiu[UCHAR_MAX] = {
[(int)'d'] = (unsigned char)(1|0x10), /* define */
[(int)'e'] = (unsigned char)(2|0x40), /* elif, else, endif */
[(int)'i'] = (unsigned char)(5|0x80), /* if ifdef ifndef include */
[(int)'u'] = (unsigned char)(9|0x90), /* undef */
};
#define CONFIG_MODE 0
#define IF0_MODE 1 /* #if 0 */
#define IF1_MODE 2 /* #if 1 */
#define ELSE0_MODE 4 /* #else found after #if 0 */
#define ELSE1_MODE 8 /* #else found after #if 1 */
#define ELIF1_MODE 16 /* #elif found after #if 1 */
#define FALSE_MODES (IF0_MODE|ELSE1_MODE|ELIF1_MODE)
#define yy_error_d(s) bb_error_d("%s:%d hmm, %s", fname, line, s)
/* state */
#define S 0 /* start state */
#define STR '"' /* string */
#define CHR '\'' /* char */
#define REM '/' /* block comment */
#define BS '\\' /* back slash */
#define POUND '#' /* # */
#define D '1' /* #define preprocessor's directive */
#define EI '2' /* #elif preprocessor's directive */
#define E '3' /* #else preprocessor's directive */
#define EF '4' /* #endif preprocessor's directive */
#define F '5' /* #if preprocessor's directive */
#define IFD '6' /* #ifdef preprocessor's directive */
#define IFND '7' /* #ifndef preprocessor's directive */
#define I '8' /* #include preprocessor's directive */
#define U '9' /* #undef preprocessor's directive */
#define DK 'K' /* #define KEY... (config mode) */
#define ANY '*' /* any unparsed chars */
/* [A-Z_a-z] */
#define ID(c) ((c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z') || c == '_')
/* [A-Z_a-z0-9] */
#define ISALNUM(c) (ID(c) || (c >= '0' && c <= '9'))
#define L_EOF (1+UCHAR_MAX)
#define getc0() do { c = (optr >= oend) ? L_EOF : *optr++; } while(0)
#define getc1() do { getc0(); if(c == BS) { getc0(); \
if(c == '\n') { line++; continue; } \
else { optr--; c = BS; } \
} break; } while(1)
static char id_s[4096];
#define put_id(ic) do { if(id_len == sizeof(id_s)) goto too_long; \
id[id_len++] = ic; } while(0)
static char ifcpp_stack[1024];
static int ptr_ifcpp_stack;
#define push_mode() do { \
if(ptr_ifcpp_stack == (int)sizeof(ifcpp_stack)) \
yy_error_d("#if* stack overflow"); \
ifcpp_stack[ptr_ifcpp_stack++] = (char)mode; \
} while(0)
#define pop_mode() do { \
if(ptr_ifcpp_stack == 0) \
yy_error_d("unexpected #endif"); \
mode = ifcpp_stack[--ptr_ifcpp_stack]; \
} while(0)
/* #if expression */
typedef long long arith_t;
static arith_t arith (const char *expr, int *perrcode);
/* The code uses a simple two-stack algorithm. See
* http://www.onthenet.com.au/~grahamis/int2008/week02/lect02.html
* for a detailed explanation of the infix-to-postfix algorithm on which
* this is based (this code differs in that it applies operators immediately
* to the stack instead of adding them to a queue to end up with an
* expression). */
#define arith_isspace(arithval) \
(arithval == ' ' || arithval == '\v' || arithval == '\t' || arithval == '\f')
typedef unsigned char operator;
/* An operator's token id is a bit of a bitfield. The lower 5 bits are the
* precedence, and 3 high bits are an ID unique across operators of that
* precedence. The ID portion is so that multiple operators can have the
* same precedence, ensuring that the leftmost one is evaluated first.
* Consider * and /. */
#define tok_decl(prec,id) (((id)<<5)|(prec))
#define PREC(op) ((op) & 0x1F)
#define TOK_LPAREN tok_decl(0,0)
#define TOK_COMMA tok_decl(1,0)
/* conditional is right associativity too */
#define TOK_CONDITIONAL tok_decl(2,0)
#define TOK_CONDITIONAL_SEP tok_decl(2,1)
#define TOK_OR tok_decl(3,0)
#define TOK_AND tok_decl(4,0)
#define TOK_BOR tok_decl(5,0)
#define TOK_BXOR tok_decl(6,0)
#define TOK_BAND tok_decl(7,0)
#define TOK_EQ tok_decl(8,0)
#define TOK_NE tok_decl(8,1)
#define TOK_LT tok_decl(9,0)
#define TOK_GT tok_decl(9,1)
#define TOK_GE tok_decl(9,2)
#define TOK_LE tok_decl(9,3)
#define TOK_LSHIFT tok_decl(10,0)
#define TOK_RSHIFT tok_decl(10,1)
#define TOK_ADD tok_decl(11,0)
#define TOK_SUB tok_decl(11,1)
#define TOK_MUL tok_decl(12,0)
#define TOK_DIV tok_decl(12,1)
#define TOK_REM tok_decl(12,2)
/* For now unary operators. */
#define UNARYPREC 13
#define TOK_BNOT tok_decl(UNARYPREC,0)
#define TOK_NOT tok_decl(UNARYPREC,1)
#define TOK_UMINUS tok_decl(UNARYPREC+1,0)
#define TOK_UPLUS tok_decl(UNARYPREC+1,1)
#define SPEC_PREC (UNARYPREC+2)
#define TOK_NUM tok_decl(SPEC_PREC, 0)
#define TOK_RPAREN tok_decl(SPEC_PREC, 1)
#define NUMPTR (*numstackptr)
typedef struct ARITCH_VAR_NUM {
arith_t val;
arith_t contidional_second_val;
char contidional_second_val_initialized;
} v_n_t;
typedef struct CHK_VAR_RECURSIVE_LOOPED {
const char *var;
size_t var_sz;
struct CHK_VAR_RECURSIVE_LOOPED *next;
} chk_var_recursive_looped_t;
static chk_var_recursive_looped_t *prev_chk_var_recursive;
static int arith_lookup_val(const char *var, size_t key_sz, arith_t *pval)
{
const char * p = lookup_key(var, key_sz);
if(p) {
int errcode;
/* recursive try as expression */
chk_var_recursive_looped_t *cur;
chk_var_recursive_looped_t cur_save;
for(cur = prev_chk_var_recursive; cur; cur = cur->next) {
if(cur->var_sz == key_sz && memcmp(cur->var, var, key_sz) == 0) {
/* expression recursion loop detected */
return -5;
}
}
/* save current lookuped var name */
cur = prev_chk_var_recursive;
cur_save.var = var;
cur_save.var_sz = key_sz;
cur_save.next = cur;
prev_chk_var_recursive = &cur_save;
*pval = arith (p, &errcode);
/* restore previous ptr after recursiving */
prev_chk_var_recursive = cur;
return errcode;
} else {
/* disallow undefined var */
fprintf(stderr, "%.*s ", (int)key_sz, var);
return -4;
}
}
/* "applying" a token means performing it on the top elements on the integer
* stack. For a unary operator it will only change the top element, but a
* binary operator will pop two arguments and push a result */
static inline int
arith_apply(operator op, v_n_t *numstack, v_n_t **numstackptr)
{
v_n_t *numptr_m1;
arith_t numptr_val, rez;
if (NUMPTR == numstack) goto err; /* There is no operator that can work
without arguments */
numptr_m1 = NUMPTR - 1;
rez = numptr_m1->val;
if (op == TOK_UMINUS)
rez *= -1;
else if (op == TOK_NOT)
rez = !rez;
else if (op == TOK_BNOT)
rez = ~rez;
else if (op != TOK_UPLUS) {
/* Binary operators */
/* check and binary operators need two arguments */
if (numptr_m1 == numstack) goto err;
/* ... and they pop one */
--NUMPTR;
numptr_val = rez;
if (op == TOK_CONDITIONAL) {
if(! numptr_m1->contidional_second_val_initialized) {
/* protect $((expr1 ? expr2)) without ": expr" */
goto err;
}
rez = numptr_m1->contidional_second_val;
} else if(numptr_m1->contidional_second_val_initialized) {
/* protect $((expr1 : expr2)) without "expr ? " */
goto err;
}
numptr_m1 = NUMPTR - 1;
if (op == TOK_CONDITIONAL) {
numptr_m1->contidional_second_val = rez;
}
rez = numptr_m1->val;
if (op == TOK_BOR)
rez |= numptr_val;
else if (op == TOK_OR)
rez = numptr_val || rez;
else if (op == TOK_BAND)
rez &= numptr_val;
else if (op == TOK_BXOR)
rez ^= numptr_val;
else if (op == TOK_AND)
rez = rez && numptr_val;
else if (op == TOK_EQ)
rez = (rez == numptr_val);
else if (op == TOK_NE)
rez = (rez != numptr_val);
else if (op == TOK_GE)
rez = (rez >= numptr_val);
else if (op == TOK_RSHIFT)
rez >>= numptr_val;
else if (op == TOK_LSHIFT)
rez <<= numptr_val;
else if (op == TOK_GT)
rez = (rez > numptr_val);
else if (op == TOK_LT)
rez = (rez < numptr_val);
else if (op == TOK_LE)
rez = (rez <= numptr_val);
else if (op == TOK_MUL)
rez *= numptr_val;
else if (op == TOK_ADD)
rez += numptr_val;
else if (op == TOK_SUB)
rez -= numptr_val;
else if (op == TOK_COMMA)
rez = numptr_val;
else if (op == TOK_CONDITIONAL_SEP) {
if (numptr_m1 == numstack) {
/* protect $((expr : expr)) without "expr ? " */
goto err;
}
numptr_m1->contidional_second_val_initialized = op;
numptr_m1->contidional_second_val = numptr_val;
}
else if (op == TOK_CONDITIONAL) {
rez = rez ?
numptr_val : numptr_m1->contidional_second_val;
}
else if(numptr_val==0) /* zero divisor check */
return -2;
else if (op == TOK_DIV)
rez /= numptr_val;
else if (op == TOK_REM)
rez %= numptr_val;
}
numptr_m1->val = rez;
return 0;
err: return(-1);
}
/* longest must first */
static const char op_tokens[] = {
'<','<', 0, TOK_LSHIFT,
'>','>', 0, TOK_RSHIFT,
'|','|', 0, TOK_OR,
'&','&', 0, TOK_AND,
'!','=', 0, TOK_NE,
'<','=', 0, TOK_LE,
'>','=', 0, TOK_GE,
'=','=', 0, TOK_EQ,
'!', 0, TOK_NOT,
'<', 0, TOK_LT,
'>', 0, TOK_GT,
'|', 0, TOK_BOR,
'&', 0, TOK_BAND,
'*', 0, TOK_MUL,
'/', 0, TOK_DIV,
'%', 0, TOK_REM,
'+', 0, TOK_ADD,
'-', 0, TOK_SUB,
'^', 0, TOK_BXOR,
'~', 0, TOK_BNOT,
',', 0, TOK_COMMA,
'?', 0, TOK_CONDITIONAL,
':', 0, TOK_CONDITIONAL_SEP,
')', 0, TOK_RPAREN,
'(', 0, TOK_LPAREN,
0
};
/* ptr to ")" */
#define endexpression &op_tokens[sizeof(op_tokens)-7]
/*
* Return of a legal variable name (a letter or underscore followed by zero or
* more letters, underscores, and digits).
*/
static inline char *
endofname(const char *name)
{
char *p;
p = (char *) name;
if (! ID(*p))
return p;
while (*++p) {
if (! ISALNUM(*p))
break;
}
return p;
}
static arith_t arith (const char *expr, int *perrcode)
{
char arithval; /* Current character under analysis */
operator lasttok, op;
operator prec;
const char *p = endexpression;
int errcode;
size_t datasizes = strlen(expr) + 2;
/* Stack of integers */
/* The proof that there can be no more than strlen(startbuf)/2+1 integers
* in any given correct or incorrect expression is left as an exercise to
* the reader. */
v_n_t *numstack = alloca(((datasizes)/2)*sizeof(v_n_t)),
*numstackptr = numstack;
/* Stack of operator tokens */
operator *stack = alloca((datasizes) * sizeof(operator)),
*stackptr = stack;
*stackptr++ = lasttok = TOK_LPAREN; /* start off with a left paren */
*perrcode = errcode = 0;
while(1) {
if ((arithval = *expr) == 0) {
if (p == endexpression) {
/* Null expression. */
err:
return (*perrcode = -1);
}
/* This is only reached after all tokens have been extracted from the
* input stream. If there are still tokens on the operator stack, they
* are to be applied in order. At the end, there should be a final
* result on the integer stack */
if (expr != endexpression + 1) {
/* If we haven't done so already, */
/* append a closing right paren */
expr = endexpression;
/* and let the loop process it. */
continue;
}
/* At this point, we're done with the expression. */
if (numstackptr != numstack+1) {
/* ... but if there isn't, it's bad */
goto err;
}
ret:
*perrcode = errcode;
return numstack->val;
} else {
/* Continue processing the expression. */
if (arith_isspace(arithval)) {
/* Skip whitespace */
goto prologue;
}
if((p = endofname(expr)) != expr) {
size_t var_name_size = (p-expr);
if(var_name_size == 7 &&
strncmp(expr, "defined", var_name_size) == 0) {
int brace_form = 0;
const char *v;
while(arith_isspace(*p)) p++;
if(*p == '(') {
p++;
while(arith_isspace(*p)) p++;
brace_form = 1;
}
expr = p;
if((p = endofname(expr)) == expr)
goto err;
var_name_size = (p-expr);
while(arith_isspace(*p)) p++;
if(brace_form && *p++ != ')')
goto err;
v = lookup_key(expr, var_name_size);
numstackptr->val = (v != NULL) ? 1 : 0;
} else {
errcode = arith_lookup_val(expr, var_name_size,
&(numstackptr->val));
if(errcode) goto ret;
}
expr = p;
num:
numstackptr->contidional_second_val_initialized = 0;
numstackptr++;
lasttok = TOK_NUM;
continue;
} else if (arithval >= '0' && arithval <= '9') {
numstackptr->val = strtoll(expr, (char **) &expr, 0);
while(*expr == 'l' || *expr == 'L' || *expr == 'u' ||
*expr == 'U')
expr++;
goto num;
}
for(p = op_tokens; ; p++) {
const char *o;
if(*p == 0) {
/* strange operator not found */
goto err;
}
for(o = expr; *p && *o == *p; p++)
o++;
if(! *p) {
/* found */
expr = o - 1;
break;
}
/* skip tail uncompared token */
while(*p)
p++;
/* skip zero delim */
p++;
}
op = p[1];
/* Plus and minus are binary (not unary) _only_ if the last
* token was as number, or a right paren (which pretends to be
* a number, since it evaluates to one). Think about it.
* It makes sense. */
if (lasttok != TOK_NUM) {
if(op == TOK_ADD)
op = TOK_UPLUS;
else if(op == TOK_SUB)
op = TOK_UMINUS;
}
/* We don't want a unary operator to cause recursive descent on the
* stack, because there can be many in a row and it could cause an
* operator to be evaluated before its argument is pushed onto the
* integer stack. */
/* But for binary operators, "apply" everything on the operator
* stack until we find an operator with a lesser priority than the
* one we have just extracted. */
/* Left paren is given the lowest priority so it will never be
* "applied" in this way.
* if associativity is right and priority eq, applied also skip
*/
prec = PREC(op);
if ((prec > 0 && prec < UNARYPREC) || prec == SPEC_PREC) {
/* not left paren or unary */
if (lasttok != TOK_NUM) {
/* binary op must be preceded by a num */
goto err;
}
while (stackptr != stack) {
if (op == TOK_RPAREN) {
/* The algorithm employed here is simple: while we don't
* hit an open paren nor the bottom of the stack, pop
* tokens and apply them */
if (stackptr[-1] == TOK_LPAREN) {
--stackptr;
/* Any operator directly after a */
lasttok = TOK_NUM;
/* close paren should consider itself binary */
goto prologue;
}
} else {
operator prev_prec = PREC(stackptr[-1]);
if (prev_prec < prec)
break;
/* check right assoc */
if(prev_prec == prec && prec == PREC(TOK_CONDITIONAL))
break;
}
errcode = arith_apply(*--stackptr, numstack, &numstackptr);
if(errcode) goto ret;
}
if (op == TOK_RPAREN) {
goto err;
}
}
/* Push this operator to the stack and remember it. */
*stackptr++ = lasttok = op;
prologue:
++expr;
}
}
}
/* stupid C lexical analyser for configs.h */
static void c_lex_config(const char *fname, long fsize)
{
int c;
int state;
int line;
char *id = id_s;
size_t id_len = 0; /* stupid initialization */
unsigned char *optr, *oend;
int mode = CONFIG_MODE;
int fd;
char *map;
int mapsize;
if(fsize == 0) {
fprintf(stderr, "Warning: %s is empty\n", fname);
return;
}
fd = open(fname, O_RDONLY);
if(fd < 0) {
perror(fname);
return;
}
mapsize = (fsize+pagesizem1) & ~pagesizem1;
map = mmap(NULL, mapsize, PROT_READ, MAP_PRIVATE, fd, 0);
if ((long) map == -1)
bb_error_d("%s: mmap: %m", fname);
optr = (unsigned char *)map;
oend = optr + fsize;
line = 1;
state = S;
for(;;) {
getc1();
for(;;) {
/* [ \t]+ eat first space */
while(c == ' ' || c == '\t')
getc1();
if(state == S) {
while(first_chars[c] == ANY) {
/* <S>unparsed */
if(c == '\n')
line++;
getc1();
}
if(c == L_EOF) {
/* <S><<EOF>> */
munmap(map, mapsize);
close(fd);
if(mode != CONFIG_MODE)
yy_error_d("expected #endif");
return;
}
if(c == REM) {
/* <S>/ */
getc0();
if(c == REM) {
/* <S>"//"[^\n]* */
do getc0(); while(c != '\n' && c != L_EOF);
} else if(c == '*') {
/* <S>[/][*] goto parse block comments */
break;
}
} else if(c == POUND) {
/* <S># */
state = c;
getc1();
} else if(c == STR || c == CHR) {
/* <S>\"|\' */
int qc = c;
for(;;) {
/* <STR,CHR>. */
getc1();
if(c == qc) {
/* <STR>\" or <CHR>\' */
break;
}
if(c == BS) {
/* <STR,CHR>\\ but is not <STR,CHR>\\\n */
getc0();
}
if(c == '\n' || c == L_EOF)
yy_error_d("unterminated");
}
getc1();
} else {
/* <S>[A-Z_a-z0-9] */
/* trick for fast drop id
if key with this first char undefined */
if(first_chars[c] == 0 || (mode & FALSE_MODES) != 0) {
/* skip <S>[A-Z_a-z0-9]+ */
do getc1(); while(isalnums[c]);
} else {
id_len = 0;
do {
/* <S>[A-Z_a-z0-9]+ */
put_id(c);
getc1();
} while(isalnums[c]);
check_key(key_top, id, id_len);
}
}
continue;
}
/* begin preprocessor states */
if(c == L_EOF)
yy_error_d("unexpected EOF");
if(c == REM) {
/* <#.*>/ */
getc0();
if(c == REM)
yy_error_d("detected // in preprocessor line");
if(c == '*') {
/* <#.*>[/][*] goto parse block comments */
break;
}
/* hmm, #.*[/] */
yy_error_d("strange preprocessor line");
}
if(state == POUND) {
/* tricks */
int diu = (int)first_chars_deiu[c]; /* preproc ptr */
state = S;
if(diu != S) {
int p_num_str, p_num_max;
getc1();
id_len = 0;
while(isalnums[c]) {
put_id(c);
getc1();
}
put_id(0);
p_num_str = diu & 0xf;
p_num_max = diu >> 4;
for(diu = p_num_str; diu <= p_num_max; diu++)
if(!strcmp(id, preproc[diu])) {
state = (diu + '0');
/* common */
id_len = 0;
break;
}
} else {
while(isalnums[c]) getc1();
}
} else if(state == EF) {
/* #endif */
pop_mode();
state = S;
} else if(state == I) {
if(c == STR && (mode & FALSE_MODES) == 0) {
/* <I>\" */
for(;;) {
getc1();
if(c == STR)
break;
if(c == L_EOF)
yy_error_d("unexpected EOF");
put_id(c);
}
put_id(0);
/* store "include.h" */
parse_inc(id, fname);
getc1();
}
/* else another (may be wrong) #include ... */
state = S;
} else if(state == F) {
arith_t t;
int errcode;
while(c != '\n' && c != L_EOF) {
put_id(c);
getc1();
}
put_id(0);
t = arith(id, &errcode);
if (errcode < 0) {
if (errcode == -2)
yy_error_d("divide by zero");
else if (errcode == -4)
yy_error_d("undefined");
else if (errcode == -5)
yy_error_d("expression recursion loop detected");
else
yy_error_d("syntax error");
}
push_mode();
mode = t != 0 ? IF1_MODE : IF0_MODE;
state = S;
} else if(state == IFD || state == IFND) {
/* save KEY from #if(n)def KEY ... */
const char *v;
push_mode();
while(isalnums[c]) {
put_id(c);
getc1();
}
if(!id_len)
yy_error_d("expected identifier");
v = lookup_key(id, id_len);
mode = IF1_MODE;
if(state == IFD && v == NULL)
mode = IF0_MODE;
else if(state == IFND && v != NULL)
mode = IF0_MODE;
state = S;
} else if(state == EI) {
/* #elif */
if(mode == CONFIG_MODE || mode == ELSE0_MODE || mode == ELSE1_MODE)
yy_error_d("unexpected #elif");
if(mode == IF0_MODE) {
pop_mode();
state = F;
} else {
mode = ELIF1_MODE;
state = S;
}
} else if(state == E) {
if(mode == CONFIG_MODE || mode == ELSE0_MODE || mode == ELSE1_MODE)
yy_error_d("unexpected #else");
if(mode == IF0_MODE)
mode = ELSE0_MODE;
else if(mode == IF1_MODE)
mode = ELSE1_MODE;
state = S;
} else if(state == D || state == U) {
/* save KEY from #"define"|"undef" ... */
while(isalnums[c]) {
put_id(c);
getc1();
}
if(!id_len)
yy_error_d("expected identifier");
if(state == U) {
if((mode & FALSE_MODES) == 0)
parse_conf_opt(id, NULL, id_len);
state = S;
} else {
/* D -> DK */
state = DK;
}
} else {
/* state==<DK> #define KEY[ ] */
size_t opt_len = id_len;
char *val = id + opt_len;
char *sp;
for(;;) {
if(c == L_EOF || c == '\n')
break;
put_id(c);
getc1();
}
sp = id + id_len;
put_id(0);
/* trim tail spaces */
while(--sp >= val && (*sp == ' ' || *sp == '\t'
|| *sp == '\f' || *sp == '\v'))
*sp = '\0';
if((mode & FALSE_MODES) == 0)
parse_conf_opt(id, val, opt_len);
state = S;
}
}
/* <REM> */
getc0();
for(;;) {
/* <REM>[^*]+ */
while(c != '*') {
if(c == '\n') {
/* <REM>\n */
if(state != S)
yy_error_d("unexpected newline");
line++;
} else if(c == L_EOF)
yy_error_d("unexpected EOF");
getc0();
}
/* <REM>[*] */
getc0();
if(c == REM) {
/* <REM>[*][/] */
break;
}
}
}
too_long:
yy_error_d("phrase too long");
}
/* trick for fast find "define", "include", "undef" */
static const char first_chars_diu[UCHAR_MAX] = {
[(int)'d'] = (char)5, /* strlen("define") - 1; */
[(int)'i'] = (char)6, /* strlen("include") - 1; */
[(int)'u'] = (char)4, /* strlen("undef") - 1; */
};
#undef D
#undef I
#undef U
#define D '5' /* #define preprocessor's directive */
#define I '6' /* #include preprocessor's directive */
#define U '4' /* #undef preprocessor's directive */
/* stupid C lexical analyser for sources */
static void c_lex_src(const char *fname, long fsize)
{
int c;
int state;
int line;
char *id = id_s;
size_t id_len = 0; /* stupid initialization */
unsigned char *optr, *oend;
int fd;
char *map;
int mapsize;
if(fsize == 0) {
fprintf(stderr, "Warning: %s is empty\n", fname);
return;
}
fd = open(fname, O_RDONLY);
if(fd < 0) {
perror(fname);
return;
}
mapsize = (fsize+pagesizem1) & ~pagesizem1;
map = mmap(NULL, mapsize, PROT_READ, MAP_PRIVATE, fd, 0);
if ((long) map == -1)
bb_error_d("%s: mmap: %m", fname);
optr = (unsigned char *)map;
oend = optr + fsize;
line = 1;
state = S;
for(;;) {
getc1();
for(;;) {
/* [ \t]+ eat first space */
while(c == ' ' || c == '\t')
getc1();
if(state == S) {
while(first_chars[c] == ANY) {
/* <S>unparsed */
if(c == '\n')
line++;
getc1();
}
if(c == L_EOF) {
/* <S><<EOF>> */
munmap(map, mapsize);
close(fd);
return;
}
if(c == REM) {
/* <S>/ */
getc0();
if(c == REM) {
/* <S>"//"[^\n]* */
do getc0(); while(c != '\n' && c != L_EOF);
} else if(c == '*') {
/* <S>[/][*] goto parse block comments */
break;
}
} else if(c == POUND) {
/* <S># */
state = c;
getc1();
} else if(c == STR || c == CHR) {
/* <S>\"|\' */
int qc = c;
for(;;) {
/* <STR,CHR>. */
getc1();
if(c == qc) {
/* <STR>\" or <CHR>\' */
break;
}
if(c == BS) {
/* <STR,CHR>\\ but is not <STR,CHR>\\\n */
getc0();
}
if(c == '\n' || c == L_EOF)
yy_error_d("unterminated");
}
getc1();
} else {
/* <S>[A-Z_a-z0-9] */
/* trick for fast drop id
if key with this first char undefined */
if(first_chars[c] == 0) {
/* skip <S>[A-Z_a-z0-9]+ */
do getc1(); while(isalnums[c]);
} else {
id_len = 0;
do {
/* <S>[A-Z_a-z0-9]+ */
put_id(c);
getc1();
} while(isalnums[c]);
check_key(key_top, id, id_len);
}
}
continue;
}
/* begin preprocessor states */
if(c == L_EOF)
yy_error_d("unexpected EOF");
if(c == REM) {
/* <#.*>/ */
getc0();
if(c == REM)
yy_error_d("detected // in preprocessor line");
if(c == '*') {
/* <#.*>[/][*] goto parse block comments */
break;
}
/* hmm, #.*[/] */
yy_error_d("strange preprocessor line");
}
if(state == POUND) {
/* tricks */
static const char * const p_preproc[] = {
/* 0 1 2 3 4 5 6 */
"", "", "", "", "ndef", "efine", "nclude"
};
size_t diu = first_chars_diu[c]; /* strlen and p_preproc ptr */
state = S;
if(diu != S) {
getc1();
id_len = 0;
while(isalnums[c]) {
put_id(c);
getc1();
}
/* str begins with c, read == strlen key and compared */
if(diu == id_len && !memcmp(id, p_preproc[diu], diu)) {
state = diu + '0';
id_len = 0; /* common for save */
}
} else {
while(isalnums[c]) getc1();
}
} else if(state == I) {
if(c == STR) {
/* <I>\" */
for(;;) {
getc1();
if(c == STR)
break;
if(c == L_EOF)
yy_error_d("unexpected EOF");
put_id(c);
}
put_id(0);
/* store "include.h" */
parse_inc(id, fname);
getc1();
}
/* else another (may be wrong) #include ... */
state = S;
} else /* if(state == D || state == U) */ {
/* ignore depend with #define or #undef KEY */
while(isalnums[c]) getc1();
state = S;
}
}
/* <REM> */
getc0();
for(;;) {
/* <REM>[^*]+ */
while(c != '*') {
if(c == '\n') {
/* <REM>\n */
if(state != S)
yy_error_d("unexpected newline");
line++;
} else if(c == L_EOF)
yy_error_d("unexpected EOF");
getc0();
}
/* <REM>[*] */
getc0();
if(c == REM) {
/* <REM>[*][/] */
break;
}
}
}
too_long:
yy_error_d("phrase too long");
}
/* bb_simplify_path special variant for absolute pathname */
static size_t bb_qa_simplify_path(char *path)
{
char *s, *p;
p = s = path;
do {
if (*p == '/') {
if (*s == '/') { /* skip duplicate (or initial) slash */
continue;
} else if (*s == '.') {
if (s[1] == '/' || s[1] == 0) { /* remove extra '.' */
continue;
} else if ((s[1] == '.') && (s[2] == '/' || s[2] == 0)) {
++s;
if (p > path) {
while (*--p != '/'); /* omit previous dir */
}
continue;
}
}
}
*++p = *s;
} while (*++s);
if ((p == path) || (*p != '/')) { /* not a trailing slash */
++p; /* so keep last character */
}
*p = 0;
return (p - path);
}
static void parse_inc(const char *include, const char *fname)
{
bb_key_t *cur;
const char *p_i;
llist_t *lo;
char *ap;
size_t key_sz;
struct stat st;
if(*include == '/') {
lo = NULL;
ap = bb_xstrdup(include);
} else {
int w;
const char *p;
lo = Iop;
p = strrchr(fname, '/'); /* fname has absolute pathname */
w = (p-fname);
/* find from current directory of source file */
ap = bb_asprint("%.*s/%s", w, fname, include);
}
for(;;) {
key_sz = bb_qa_simplify_path(ap);
cur = check_key(Ifound, ap, key_sz);
if(cur) {
cur->checked = cur->value;
free(ap);
return;
}
if(stat(ap, &st) == 0) {
/* found */
llist_t *cfl;
for(cfl = configs; cfl; cfl = cfl->link) {
struct stat *config = (struct stat *)cfl->data;
if (st.st_dev == config->st_dev && st.st_ino == config->st_ino) {
/* skip depend with bb_configs.h */
return;
}
}
p_i = ap;
break;
} else if(lo == NULL) {
p_i = NULL;
break;
}
/* find from "-I include" specified directories */
free(ap);
/* lo->data has absolute pathname */
ap = bb_asprint("%s/%s", lo->data, include);
lo = lo->link;
}
cur = xmalloc(sizeof(bb_key_t));
cur->keyname = ap;
cur->key_sz = key_sz;
cur->stored_path = ap;
cur->value = cur->checked = p_i;
if(p_i == NULL && noiwarning)
fprintf(stderr, "%s: Warning: #include \"%s\" not found\n", fname, include);
cur->next = Ifound;
Ifound = cur;
}
static size_t max_rec_sz;
static void parse_conf_opt(const char *opt, const char *val, size_t key_sz)
{
bb_key_t *cur;
char *k;
char *p;
size_t val_sz = 0;
cur = check_key(key_top, opt, key_sz);
if(cur != NULL) {
/* already present */
cur->checked = NULL; /* store only */
if(cur->value == NULL && val == NULL)
return;
if(cur->value != NULL && val != NULL && !strcmp(cur->value, val))
return;
k = cur->keyname;
fprintf(stderr, "Warning: redefined %s\n", k);
} else {
size_t recordsz;
if(val && *val)
val_sz = strlen(val) + 1;
recordsz = key_sz + val_sz + 1;
if(max_rec_sz < recordsz)
max_rec_sz = recordsz;
cur = xmalloc(sizeof(bb_key_t) + recordsz);
k = cur->keyname = memcpy(cur + 1, opt, key_sz);
cur->keyname[key_sz] = '\0';
cur->key_sz = key_sz;
cur->checked = NULL;
cur->src_have_this_key = NULL;
cur->next = key_top;
key_top = cur;
}
/* save VAL */
if(val) {
if(*val == '\0') {
cur->value = "";
} else {
cur->value = p = cur->keyname + key_sz + 1;
memcpy(p, val, val_sz);
}
} else {
cur->value = NULL;
}
/* trick, save first char KEY for do fast identify id */
first_chars[(int)*k] = *k;
cur->stored_path = k = bb_asprint("%s/%s.h", kp, k);
/* key conversion [A-Z_] -> [a-z/] */
for(p = k + kp_len + 1; *p; p++) {
if(*p >= 'A' && *p <= 'Z')
*p = *p - 'A' + 'a';
else if(*p == '_' && p[1] > '9') /* do not change A_1 to A/1 */
*p = '/';
}
}
static void store_keys(void)
{
bb_key_t *cur;
char *k;
struct stat st;
int cmp_ok;
ssize_t rw_ret;
size_t recordsz = max_rec_sz * 2 + 10 * 2 + 16;
/* buffer for double "#define KEY VAL\n" */
char *record_buf = xmalloc(recordsz);
for(cur = key_top; cur; cur = cur->next) {
if(cur->src_have_this_key) {
/* do generate record */
k = cur->keyname;
if(cur->value == NULL) {
recordsz = sprintf(record_buf, "#undef %s\n", k);
} else {
const char *val = cur->value;
if(*val == '\0') {
recordsz = sprintf(record_buf, "#define %s\n", k);
} else {
if(val[0] != '(')
recordsz = sprintf(record_buf, "#define %s %s\n", k, val);
else
recordsz = sprintf(record_buf, "#define %s%s\n", k, val);
}
}
/* size_t -> ssize_t :( */
rw_ret = (ssize_t)recordsz;
/* check kp/key.h, compare after previous use */
cmp_ok = 0;
k = cur->stored_path;
if(stat(k, &st)) {
char *p;
for(p = k + kp_len + 1; *p; p++) {
/* Auto-create directories. */
if (*p == '/') {
*p = '\0';
if (access(k, F_OK) != 0 && mkdir(k, 0755) != 0)
bb_error_d("mkdir(%s): %m", k);
*p = '/';
}
}
} else {
/* found */
if(st.st_size == (off_t)recordsz) {
char *r_cmp;
int fd;
size_t padded = recordsz;
/* 16-byte padding for read(2) and memcmp(3) */
padded = (padded+15) & ~15;
r_cmp = record_buf + padded;
fd = open(k, O_RDONLY);
if(fd < 0 || read(fd, r_cmp, recordsz) < rw_ret)
bb_error_d("%s: %m", k);
close(fd);
cmp_ok = memcmp(record_buf, r_cmp, recordsz) == 0;
}
}
if(!cmp_ok) {
int fd = open(k, O_WRONLY|O_CREAT|O_TRUNC, 0644);
if(fd < 0 || write(fd, record_buf, recordsz) < rw_ret)
bb_error_d("%s: %m", k);
close(fd);
}
}
}
}
static int show_dep(int first, bb_key_t *k, const char *name, const char *f)
{
bb_key_t *cur;
for(cur = k; cur; cur = cur->next) {
if(cur->checked) {
if(first >= 0) {
if(first) {
if(f == NULL)
printf("\n%s:", name);
else
printf("\n%s/%s:", pwd, name);
first = 0;
} else {
printf(" \\\n ");
}
printf(" %s", cur->checked);
}
cur->src_have_this_key = cur->checked;
cur->checked = NULL;
}
}
return first;
}
static char *
parse_chd(const char *fe, const char *p, size_t dirlen)
{
struct stat st;
char *fp;
size_t df_sz;
static char dir_and_entry[4096];
size_t fe_sz = strlen(fe) + 1;
df_sz = dirlen + fe_sz + 1; /* dir/file\0 */
if(df_sz > sizeof(dir_and_entry))
bb_error_d("%s: file name too long", fe);
fp = dir_and_entry;
/* sprintf(fp, "%s/%s", p, fe); */
memcpy(fp, p, dirlen);
fp[dirlen] = '/';
memcpy(fp + dirlen + 1, fe, fe_sz);
if(stat(fp, &st)) {
fprintf(stderr, "Warning: stat(%s): %m\n", fp);
return NULL;
}
if(S_ISREG(st.st_mode)) {
llist_t *cfl;
char *e = fp + df_sz - 3;
if(*e++ != '.' || (*e != 'c' && *e != 'h')) {
/* direntry is regular file, but is not *.[ch] */
return NULL;
}
for(cfl = configs; cfl; cfl = cfl->link) {
struct stat *config = (struct stat *)cfl->data;
if (st.st_dev == config->st_dev && st.st_ino == config->st_ino) {
/* skip already parsed bb_configs.h */
return NULL;
}
}
/* direntry is *.[ch] regular file and is not configs */
c_lex_src(fp, st.st_size);
if(!dontgenerate_dep) {
int first;
if(*e == 'c') {
/* *.c -> *.o */
*e = 'o';
/* /src_dir/path/file.o to path/file.o */
fp += replace;
if(*fp == '/')
fp++;
} else {
e = NULL;
}
first = show_dep(1, Ifound, fp, e);
first = show_dep(first, key_top, fp, e);
if(first == 0)
putchar('\n');
} else {
show_dep(-1, key_top, NULL, NULL);
}
return NULL;
} else if(S_ISDIR(st.st_mode)) {
if (st.st_dev == st_kp.st_dev && st.st_ino == st_kp.st_ino)
return NULL; /* drop scan kp/ directory */
/* direntry is directory. buff is returned */
return bb_xstrdup(fp);
}
/* hmm, direntry is device! */
return NULL;
}
/* from libbb but inlined for speed considerations */
static inline llist_t *llist_add_to(llist_t *old_head, char *new_item)
{
llist_t *new_head;
new_head = xmalloc(sizeof(llist_t));
new_head->data = new_item;
new_head->link = old_head;
return(new_head);
}
static void scan_dir_find_ch_files(const char *p)
{
llist_t *dirs;
llist_t *d_add;
llist_t *d;
struct dirent *de;
DIR *dir;
size_t dirlen;
dirs = llist_add_to(NULL, bb_simplify_path(p));
replace = strlen(dirs->data);
/* emulate recursion */
while(dirs) {
d_add = NULL;
while(dirs) {
dir = opendir(dirs->data);
if (dir == NULL)
fprintf(stderr, "Warning: opendir(%s): %m\n", dirs->data);
dirlen = strlen(dirs->data);
while ((de = readdir(dir)) != NULL) {
char *found_dir;
if (de->d_name[0] == '.')
continue;
found_dir = parse_chd(de->d_name, dirs->data, dirlen);
if(found_dir)
d_add = llist_add_to(d_add, found_dir);
}
closedir(dir);
free(dirs->data);
d = dirs;
dirs = dirs->link;
free(d);
}
dirs = d_add;
}
}
static void show_usage(void) ATTRIBUTE ((noreturn));
static void show_usage(void)
{
bb_error_d("%s\n%s\n", bb_mkdep_terse_options, bb_mkdep_full_options);
}
int main(int argc, char **argv)
{
char *s;
int i;
llist_t *fl;
{
/* for bb_simplify_path, this program has no chdir() */
/* libbb-like my xgetcwd() */
unsigned path_max = 512;
s = xmalloc (path_max);
while (getcwd (s, path_max) == NULL) {
if(errno != ERANGE)
bb_error_d("getcwd: %m");
free(s);
s = xmalloc(path_max *= 2);
}
pwd = s;
}
while ((i = getopt(argc, argv, "I:c:dk:w")) > 0) {
switch(i) {
case 'I':
s = bb_simplify_path(optarg);
Iop = llist_add_to(Iop, s);
break;
case 'c':
s = bb_simplify_path(optarg);
configs = llist_add_to(configs, s);
break;
case 'd':
dontgenerate_dep = 1;
break;
case 'k':
if(kp)
bb_error_d("Hmm, why multiple -k?");
kp = bb_simplify_path(optarg);
break;
case 'w':
noiwarning = 1;
break;
default:
show_usage();
}
}
/* default kp */
if(kp == NULL)
kp = bb_simplify_path(INCLUDE_CONFIG_PATH);
/* globals initialize */
kp_len = strlen(kp);
if(stat(kp, &st_kp))
bb_error_d("stat(%s): %m", kp);
if(!S_ISDIR(st_kp.st_mode))
bb_error_d("%s is not directory", kp);
/* defaults */
if(Iop == NULL)
Iop = llist_add_to(Iop, bb_simplify_path(LOCAL_INCLUDE_PATH));
if(configs == NULL) {
s = bb_simplify_path(INCLUDE_CONFIG_KEYS_PATH);
configs = llist_add_to(configs, s);
}
/* for c_lex */
pagesizem1 = getpagesize() - 1;
for(i = 0; i < UCHAR_MAX; i++) {
if(ISALNUM(i))
isalnums[i] = i;
/* set unparsed chars to speed up the parser */
else if(i != CHR && i != STR && i != POUND && i != REM)
first_chars[i] = ANY;
}
first_chars[i] = '-'; /* L_EOF */
/* parse configs */
for(fl = configs; fl; fl = fl->link) {
struct stat st;
if(stat(fl->data, &st))
bb_error_d("stat(%s): %m", fl->data);
c_lex_config(fl->data, st.st_size);
free(fl->data);
/* trick for fast comparing found files with configs */
fl->data = xmalloc(sizeof(struct stat));
memcpy(fl->data, &st, sizeof(struct stat));
}
/* main loop */
argv += optind;
if(*argv) {
while(*argv)
scan_dir_find_ch_files(*argv++);
} else {
scan_dir_find_ch_files(".");
}
store_keys();
return 0;
}
/* partial and simplified libbb routine */
static void bb_error_d(const char *s, ...)
{
va_list p;
va_start(p, s);
vfprintf(stderr, s, p);
va_end(p);
putc('\n', stderr);
exit(1);
}
static char *bb_asprint(const char *format, ...)
{
va_list p;
int r;
char *out;
#ifdef __USE_GNU
va_start(p, format);
r = vasprintf(&out, format, p);
va_end(p);
#else
out = xmalloc(BUFSIZ);
va_start(p, format);
r = vsprintf(out, format, p);
va_end(p);
#endif
if (r < 0)
bb_error_d("bb_asprint: %m");
return out;
}
/* partial libbb routine as is */
static char *bb_simplify_path(const char *path)
{
char *s, *start, *p;
if (path[0] == '/')
start = bb_xstrdup(path);
else {
/* is not libbb, but this program has no chdir() */
start = bb_asprint("%s/%s", pwd, path);
}
p = s = start;
do {
if (*p == '/') {
if (*s == '/') { /* skip duplicate (or initial) slash */
continue;
} else if (*s == '.') {
if (s[1] == '/' || s[1] == 0) { /* remove extra '.' */
continue;
} else if ((s[1] == '.') && (s[2] == '/' || s[2] == 0)) {
++s;
if (p > start) {
while (*--p != '/'); /* omit previous dir */
}
continue;
}
}
}
*++p = *s;
} while (*++s);
if ((p == start) || (*p != '/')) { /* not a trailing slash */
++p; /* so keep last character */
}
*p = 0;
return start;
}