busybox/editors/diff.c

1061 lines
30 KiB
C
Raw Normal View History

/* vi: set sw=4 ts=4: */
/*
* Mini diff implementation for busybox, adapted from OpenBSD diff.
*
* Copyright (C) 2010 by Matheus Izvekov <mizvekov@gmail.com>
* Copyright (C) 2006 by Robert Sullivan <cogito.ergo.cogito@hotmail.com>
* Copyright (c) 2003 Todd C. Miller <Todd.Miller@courtesan.com>
*
* Sponsored in part by the Defense Advanced Research Projects
* Agency (DARPA) and Air Force Research Laboratory, Air Force
* Materiel Command, USAF, under agreement number F39502-99-1-0512.
*
* Licensed under GPLv2 or later, see file LICENSE in this source tree.
*/
/*
* The following code uses an algorithm due to Harold Stone,
* which finds a pair of longest identical subsequences in
* the two files.
*
* The major goal is to generate the match vector J.
* J[i] is the index of the line in file1 corresponding
* to line i in file0. J[i] = 0 if there is no
* such line in file1.
*
* Lines are hashed so as to work in core. All potential
* matches are located by sorting the lines of each file
* on the hash (called "value"). In particular, this
* collects the equivalence classes in file1 together.
* Subroutine equiv replaces the value of each line in
* file0 by the index of the first element of its
* matching equivalence in (the reordered) file1.
* To save space equiv squeezes file1 into a single
* array member in which the equivalence classes
* are simply concatenated, except that their first
* members are flagged by changing sign.
*
* Next the indices that point into member are unsorted into
* array class according to the original order of file0.
*
* The cleverness lies in routine stone. This marches
* through the lines of file0, developing a vector klist
* of "k-candidates". At step i a k-candidate is a matched
* pair of lines x,y (x in file0, y in file1) such that
* there is a common subsequence of length k
* between the first i lines of file0 and the first y
* lines of file1, but there is no such subsequence for
* any smaller y. x is the earliest possible mate to y
* that occurs in such a subsequence.
*
* Whenever any of the members of the equivalence class of
* lines in file1 matable to a line in file0 has serial number
* less than the y of some k-candidate, that k-candidate
* with the smallest such y is replaced. The new
* k-candidate is chained (via pred) to the current
* k-1 candidate so that the actual subsequence can
* be recovered. When a member has serial number greater
* that the y of all k-candidates, the klist is extended.
* At the end, the longest subsequence is pulled out
* and placed in the array J by unravel
*
* With J in hand, the matches there recorded are
* checked against reality to assure that no spurious
* matches have crept in due to hashing. If they have,
* they are broken, and "jackpot" is recorded--a harmless
* matter except that a true match for a spuriously
* mated line may now be unnecessarily reported as a change.
*
* Much of the complexity of the program comes simply
* from trying to minimize core utilization and
* maximize the range of doable problems by dynamically
* allocating what is needed and reusing what is not.
* The core requirements for problems larger than somewhat
* are (in words) 2*length(file0) + length(file1) +
* 3*(number of k-candidates installed), typically about
* 6n words for files of length n.
*/
//config:config DIFF
//config: bool "diff (13 kb)"
//config: default y
//config: help
//config: diff compares two files or directories and outputs the
//config: differences between them in a form that can be given to
//config: the patch command.
//config:
//config:config FEATURE_DIFF_LONG_OPTIONS
//config: bool "Enable long options"
//config: default y
//config: depends on DIFF && LONG_OPTS
//config:
//config:config FEATURE_DIFF_DIR
//config: bool "Enable directory support"
//config: default y
//config: depends on DIFF
//config: help
//config: This option enables support for directory and subdirectory
//config: comparison.
//kbuild:lib-$(CONFIG_DIFF) += diff.o
//applet:IF_DIFF(APPLET(diff, BB_DIR_USR_BIN, BB_SUID_DROP))
//usage:#define diff_trivial_usage
//usage: "[-abBdiNqrTstw] [-L LABEL] [-S FILE] [-U LINES] FILE1 FILE2"
//usage:#define diff_full_usage "\n\n"
//usage: "Compare files line by line and output the differences between them.\n"
//usage: "This implementation supports unified diffs only.\n"
//usage: "\n -a Treat all files as text"
//usage: "\n -b Ignore changes in the amount of whitespace"
//usage: "\n -B Ignore changes whose lines are all blank"
//usage: "\n -d Try hard to find a smaller set of changes"
//usage: "\n -i Ignore case differences"
//usage: "\n -L Use LABEL instead of the filename in the unified header"
//usage: "\n -N Treat absent files as empty"
//usage: "\n -q Output only whether files differ"
//usage: "\n -r Recurse"
//usage: "\n -S Start with FILE when comparing directories"
//usage: "\n -T Make tabs line up by prefixing a tab when necessary"
//usage: "\n -s Report when two files are the same"
//usage: "\n -t Expand tabs to spaces in output"
//usage: "\n -U Output LINES lines of context"
//usage: "\n -w Ignore all whitespace"
#include "libbb.h"
#include "common_bufsiz.h"
#if 0
# define dbg_error_msg(...) bb_error_msg(__VA_ARGS__)
#else
# define dbg_error_msg(...) ((void)0)
#endif
enum { /* print_status() and diffreg() return values */
STATUS_SAME, /* files are the same */
STATUS_DIFFER, /* files differ */
STATUS_BINARY, /* binary files differ */
};
enum { /* Commandline flags */
FLAG_a,
FLAG_b,
FLAG_d,
FLAG_i,
FLAG_L, /* never used, handled by getopt32 */
FLAG_N,
FLAG_q,
FLAG_r,
FLAG_s,
FLAG_S, /* never used, handled by getopt32 */
FLAG_t,
FLAG_T,
FLAG_U, /* never used, handled by getopt32 */
FLAG_w,
FLAG_u, /* ignored, this is the default */
FLAG_p, /* not implemented */
FLAG_B,
FLAG_E, /* not implemented */
};
#define FLAG(x) (1 << FLAG_##x)
/* We cache file position to avoid excessive seeking */
typedef struct FILE_and_pos_t {
FILE *ft_fp;
off_t ft_pos;
} FILE_and_pos_t;
struct globals {
smallint exit_status;
int opt_U_context;
const char *other_dir;
char *label[2];
struct stat stb[2];
};
#define G (*ptr_to_globals)
#define exit_status (G.exit_status )
#define opt_U_context (G.opt_U_context )
#define label (G.label )
#define stb (G.stb )
#define INIT_G() do { \
SET_PTR_TO_GLOBALS(xzalloc(sizeof(G))); \
opt_U_context = 3; \
} while (0)
typedef int token_t;
enum {
/* Public */
TOK_EMPTY = 1 << 9, /* Line fully processed, you can proceed to the next */
TOK_EOF = 1 << 10, /* File ended */
/* Private (Only to be used by read_token() */
TOK_EOL = 1 << 11, /* we saw EOL (sticky) */
TOK_SPACE = 1 << 12, /* used -b code, means we are skipping spaces */
SHIFT_EOF = (sizeof(token_t)*8 - 8) - 1,
CHAR_MASK = 0x1ff, /* 8th bit is used to distinguish EOF from 0xff */
};
/* Restores full EOF from one 8th bit: */
//#define TOK2CHAR(t) (((t) << SHIFT_EOF) >> SHIFT_EOF)
/* We don't really need the above, we only need to have EOF != any_real_char: */
#define TOK2CHAR(t) ((t) & CHAR_MASK)
static void seek_ft(FILE_and_pos_t *ft, off_t pos)
{
if (ft->ft_pos != pos) {
ft->ft_pos = pos;
fseeko(ft->ft_fp, pos, SEEK_SET);
2006-05-29 17:42:45 +05:30
}
}
/* Reads tokens from given fp, handling -b and -w flags
* The user must reset tok every line start
*/
static int read_token(FILE_and_pos_t *ft, token_t tok)
{
tok |= TOK_EMPTY;
while (!(tok & TOK_EOL)) {
bool is_space;
int t;
t = fgetc(ft->ft_fp);
if (t != EOF)
ft->ft_pos++;
is_space = (t == EOF || isspace(t));
/* If t == EOF (-1), set both TOK_EOF and TOK_EOL */
tok |= (t & (TOK_EOF + TOK_EOL));
/* Only EOL? */
if (t == '\n')
tok |= TOK_EOL;
if (option_mask32 & FLAG(i)) /* Handcoded tolower() */
t = (t >= 'A' && t <= 'Z') ? t - ('A' - 'a') : t;
if ((option_mask32 & FLAG(w)) && is_space)
continue;
/* Trim char value to low 9 bits */
t &= CHAR_MASK;
if (option_mask32 & FLAG(b)) {
/* Was prev char whitespace? */
if (tok & TOK_SPACE) { /* yes */
if (is_space) /* this one too, ignore it */
continue;
tok &= ~TOK_SPACE;
} else if (is_space) {
/* 1st whitespace char.
* Set TOK_SPACE and replace char by ' ' */
t = TOK_SPACE + ' ';
2006-05-29 17:42:45 +05:30
}
}
/* Clear EMPTY */
tok &= ~(TOK_EMPTY + CHAR_MASK);
/* Assign char value (low 9 bits) and maybe set TOK_SPACE */
tok |= t;
break;
2006-05-29 17:42:45 +05:30
}
#if 0
bb_error_msg("fp:%p tok:%x '%c'%s%s%s%s", fp, tok, tok & 0xff
, tok & TOK_EOF ? " EOF" : ""
, tok & TOK_EOL ? " EOL" : ""
, tok & TOK_EMPTY ? " EMPTY" : ""
, tok & TOK_SPACE ? " SPACE" : ""
);
#endif
return tok;
}
struct cand {
int x;
int y;
int pred;
};
static int search(const int *c, int k, int y, const struct cand *list)
{
int i, j;
if (list[c[k]].y < y) /* quick look for typical case */
return k + 1;
for (i = 0, j = k + 1;;) {
const int l = (i + j) >> 1;
if (l > i) {
const int t = list[c[l]].y;
if (t > y)
j = l;
else if (t < y)
i = l;
else
return l;
} else
return l + 1;
2006-05-29 17:42:45 +05:30
}
}
static void stone(const int *a, int n, const int *b, int *J, int pref)
{
const unsigned isq = isqrt(n);
const unsigned bound =
(option_mask32 & FLAG(d)) ? UINT_MAX : MAX(256, isq);
int clen = 1;
int clistlen = 100;
int k = 0;
struct cand *clist = xzalloc(clistlen * sizeof(clist[0]));
struct cand cand;
struct cand *q;
int *klist = xzalloc((n + 2) * sizeof(klist[0]));
/*clist[0] = (struct cand){0}; - xzalloc did it */
/*klist[0] = 0; */
for (cand.x = 1; cand.x <= n; cand.x++) {
int j = a[cand.x], oldl = 0;
unsigned numtries = 0;
if (j == 0)
continue;
cand.y = -b[j];
cand.pred = klist[0];
do {
int l, tc;
if (cand.y <= clist[cand.pred].y)
continue;
l = search(klist, k, cand.y, clist);
if (l != oldl + 1)
cand.pred = klist[l - 1];
if (l <= k && clist[klist[l]].y <= cand.y)
continue;
if (clen == clistlen) {
clistlen = clistlen * 11 / 10;
clist = xrealloc(clist, clistlen * sizeof(clist[0]));
}
clist[clen] = cand;
tc = klist[l];
klist[l] = clen++;
if (l <= k) {
cand.pred = tc;
oldl = l;
numtries++;
} else {
k++;
break;
}
} while ((cand.y = b[++j]) > 0 && numtries < bound);
2006-05-29 17:42:45 +05:30
}
/* Unravel */
for (q = clist + klist[k]; q->y; q = clist + q->pred)
J[q->x + pref] = q->y + pref;
free(klist);
free(clist);
}
struct line {
/* 'serial' is not used in the beginning, so we reuse it
* to store line offsets, thus reducing memory pressure
*/
union {
unsigned serial;
off_t offset;
};
unsigned value;
};
static void equiv(struct line *a, int n, struct line *b, int m, int *c)
{
int i = 1, j = 1;
2006-05-29 17:42:45 +05:30
while (i <= n && j <= m) {
if (a[i].value < b[j].value)
a[i++].value = 0;
else if (a[i].value == b[j].value)
a[i++].value = j;
else
j++;
}
while (i <= n)
a[i++].value = 0;
b[m + 1].value = 0;
j = 0;
while (++j <= m) {
c[j] = -b[j].serial;
while (b[j + 1].value == b[j].value) {
j++;
c[j] = b[j].serial;
}
}
c[j] = -1;
}
static void unsort(const struct line *f, int l, int *b)
{
int i;
int *a = xmalloc((l + 1) * sizeof(a[0]));
for (i = 1; i <= l; i++)
2006-05-29 17:42:45 +05:30
a[f[i].serial] = f[i].value;
for (i = 1; i <= l; i++)
2006-05-29 17:42:45 +05:30
b[i] = a[i];
free(a);
}
static int line_compar(const void *a, const void *b)
{
#define l0 ((const struct line*)a)
#define l1 ((const struct line*)b)
int r = l0->value - l1->value;
if (r)
return r;
return l0->serial - l1->serial;
#undef l0
#undef l1
}
static void fetch(FILE_and_pos_t *ft, const off_t *ix, int a, int b, int ch)
{
int i, j, col;
for (i = a; i <= b; i++) {
seek_ft(ft, ix[i - 1]);
putchar(ch);
if (option_mask32 & FLAG(T))
putchar('\t');
for (j = 0, col = 0; j < ix[i] - ix[i - 1]; j++) {
int c = fgetc(ft->ft_fp);
if (c == EOF) {
puts("\n\\ No newline at end of file");
return;
2006-05-29 17:42:45 +05:30
}
ft->ft_pos++;
if (c == '\t' && (option_mask32 & FLAG(t)))
do putchar(' '); while (++col & 7);
else {
2006-05-29 17:42:45 +05:30
putchar(c);
col++;
}
}
}
}
/* Creates the match vector J, where J[i] is the index
* of the line in the new file corresponding to the line i
* in the old file. Lines start at 1 instead of 0, that value
* being used instead to denote no corresponding line.
* This vector is dynamically allocated and must be freed by the caller.
*
* * fp is an input parameter, where fp[0] and fp[1] are the open
* old file and new file respectively.
* * nlen is an output variable, where nlen[0] and nlen[1]
* gets the number of lines in the old and new file respectively.
* * ix is an output variable, where ix[0] and ix[1] gets
* assigned dynamically allocated vectors of the offsets of the lines
* of the old and new file respectively. These must be freed by the caller.
*/
static NOINLINE int *create_J(FILE_and_pos_t ft[2], int nlen[2], off_t *ix[2])
{
int *J, slen[2], *class, *member;
struct line *nfile[2], *sfile[2];
int pref = 0, suff = 0, i, j, delta;
/* Lines of both files are hashed, and in the process
* their offsets are stored in the array ix[fileno]
* where fileno == 0 points to the old file, and
* fileno == 1 points to the new one.
*/
for (i = 0; i < 2; i++) {
unsigned hash;
token_t tok;
size_t sz = 100;
nfile[i] = xmalloc((sz + 3) * sizeof(nfile[i][0]));
/* ft gets here without the correct position, cant use seek_ft */
ft[i].ft_pos = 0;
fseeko(ft[i].ft_fp, 0, SEEK_SET);
nlen[i] = 0;
/* We could zalloc nfile, but then zalloc starts showing in gprof at ~1% */
nfile[i][0].offset = 0;
goto start; /* saves code */
while (1) {
tok = read_token(&ft[i], tok);
if (!(tok & TOK_EMPTY)) {
/* Hash algorithm taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578. */
/*hash = hash * 128 - hash + TOK2CHAR(tok);
* gcc insists on optimizing above to "hash * 127 + ...", thus... */
unsigned o = hash - TOK2CHAR(tok);
hash = hash * 128 - o; /* we want SPEED here */
continue;
}
if (nlen[i]++ == sz) {
sz = sz * 3 / 2;
nfile[i] = xrealloc(nfile[i], (sz + 3) * sizeof(nfile[i][0]));
}
/* line_compar needs hashes fit into positive int */
nfile[i][nlen[i]].value = hash & INT_MAX;
/* like ftello(ft[i].ft_fp) but faster (avoids lseek syscall) */
nfile[i][nlen[i]].offset = ft[i].ft_pos;
if (tok & TOK_EOF) {
/* EOF counts as a token, so we have to adjust it here */
nfile[i][nlen[i]].offset++;
break;
}
start:
hash = tok = 0;
}
/* Exclude lone EOF line from the end of the file, to make fetch()'s job easier */
if (nfile[i][nlen[i]].offset - nfile[i][nlen[i] - 1].offset == 1)
nlen[i]--;
/* Now we copy the line offsets into ix */
ix[i] = xmalloc((nlen[i] + 2) * sizeof(ix[i][0]));
for (j = 0; j < nlen[i] + 1; j++)
ix[i][j] = nfile[i][j].offset;
}
/* length of prefix and suffix is calculated */
for (; pref < nlen[0] && pref < nlen[1] &&
nfile[0][pref + 1].value == nfile[1][pref + 1].value;
pref++);
for (; suff < nlen[0] - pref && suff < nlen[1] - pref &&
nfile[0][nlen[0] - suff].value == nfile[1][nlen[1] - suff].value;
suff++);
/* Arrays are pruned by the suffix and prefix length,
* the result being sorted and stored in sfile[fileno],
* and their sizes are stored in slen[fileno]
*/
for (j = 0; j < 2; j++) {
sfile[j] = nfile[j] + pref;
slen[j] = nlen[j] - pref - suff;
for (i = 0; i <= slen[j]; i++)
sfile[j][i].serial = i;
qsort(sfile[j] + 1, slen[j], sizeof(*sfile[j]), line_compar);
}
/* nfile arrays are reused to reduce memory pressure
* The #if zeroed out section performs the same task as the
* one in the #else section.
* Peak memory usage is higher, but one array copy is avoided
* by not using unsort()
2006-05-29 17:42:45 +05:30
*/
#if 0
member = xmalloc((slen[1] + 2) * sizeof(member[0]));
equiv(sfile[0], slen[0], sfile[1], slen[1], member);
free(nfile[1]);
class = xmalloc((slen[0] + 1) * sizeof(class[0]));
for (i = 1; i <= slen[0]; i++) /* Unsorting */
class[sfile[0][i].serial] = sfile[0][i].value;
free(nfile[0]);
#else
member = (int *)nfile[1];
equiv(sfile[0], slen[0], sfile[1], slen[1], member);
member = xrealloc(member, (slen[1] + 2) * sizeof(member[0]));
class = (int *)nfile[0];
unsort(sfile[0], slen[0], (int *)nfile[0]);
class = xrealloc(class, (slen[0] + 2) * sizeof(class[0]));
#endif
J = xmalloc((nlen[0] + 2) * sizeof(J[0]));
/* The elements of J which fall inside the prefix and suffix regions
* are marked as unchanged, while the ones which fall outside
* are initialized with 0 (no matches), so that function stone can
* then assign them their right values
*/
for (i = 0, delta = nlen[1] - nlen[0]; i <= nlen[0]; i++)
J[i] = i <= pref ? i :
i > (nlen[0] - suff) ? (i + delta) : 0;
/* Here the magic is performed */
stone(class, slen[0], member, J, pref);
J[nlen[0] + 1] = nlen[1] + 1;
free(class);
free(member);
/* Both files are rescanned, in an effort to find any lines
* which, due to limitations intrinsic to any hashing algorithm,
* are different but ended up confounded as the same
*/
for (i = 1; i <= nlen[0]; i++) {
if (!J[i])
continue;
seek_ft(&ft[0], ix[0][i - 1]);
seek_ft(&ft[1], ix[1][J[i] - 1]);
for (j = J[i]; i <= nlen[0] && J[i] == j; i++, j++) {
token_t tok0 = 0, tok1 = 0;
do {
tok0 = read_token(&ft[0], tok0);
tok1 = read_token(&ft[1], tok1);
2006-05-29 17:42:45 +05:30
if (((tok0 ^ tok1) & TOK_EMPTY) != 0 /* one is empty (not both) */
|| (!(tok0 & TOK_EMPTY) && TOK2CHAR(tok0) != TOK2CHAR(tok1))
) {
J[i] = 0; /* Break the correspondence */
}
} while (!(tok0 & tok1 & TOK_EMPTY));
}
2006-05-29 17:42:45 +05:30
}
return J;
}
static bool diff(FILE* fp[2], char *file[2])
{
int nlen[2];
off_t *ix[2];
FILE_and_pos_t ft[2];
typedef struct { int a, b; } vec_t[2];
vec_t *vec = NULL;
int i = 1, j, k, idx = -1;
bool anychange = false;
int *J;
ft[0].ft_fp = fp[0];
ft[1].ft_fp = fp[1];
/* note that ft[i].ft_pos is unintitalized, create_J()
* must not assume otherwise */
J = create_J(ft, nlen, ix);
2006-05-29 17:42:45 +05:30
do {
bool nonempty = false;
while (1) {
vec_t v;
2006-05-29 17:42:45 +05:30
for (v[0].a = i; v[0].a <= nlen[0] && J[v[0].a] == J[v[0].a - 1] + 1; v[0].a++)
continue;
v[1].a = J[v[0].a - 1] + 1;
2006-05-29 17:42:45 +05:30
for (v[0].b = v[0].a - 1; v[0].b < nlen[0] && !J[v[0].b + 1]; v[0].b++)
continue;
v[1].b = J[v[0].b + 1] - 1;
/*
* Indicate that there is a difference between lines a and b of the 'from' file
* to get to lines c to d of the 'to' file. If a is greater than b then there
* are no lines in the 'from' file involved and this means that there were
* lines appended (beginning at b). If c is greater than d then there are
* lines missing from the 'to' file.
*/
if (v[0].a <= v[0].b || v[1].a <= v[1].b) {
/*
* If this change is more than 'context' lines from the
* previous change, dump the record and reset it.
*/
int ct = (2 * opt_U_context) + 1;
if (idx >= 0
&& v[0].a > vec[idx][0].b + ct
&& v[1].a > vec[idx][1].b + ct
) {
break;
}
for (j = 0; j < 2; j++)
for (k = v[j].a; k <= v[j].b; k++)
nonempty |= (ix[j][k] - ix[j][k - 1] != 1);
vec = xrealloc_vector(vec, 6, ++idx);
memcpy(vec[idx], v, sizeof(v));
}
i = v[0].b + 1;
if (i > nlen[0])
break;
J[v[0].b] = v[1].b;
}
if (idx < 0 || ((option_mask32 & FLAG(B)) && !nonempty))
goto cont;
if (!(option_mask32 & FLAG(q))) {
int lowa;
vec_t span, *cvp = vec;
if (!anychange) {
/* Print the context/unidiff header first time through */
printf("--- %s\n", label[0] ? label[0] : file[0]);
printf("+++ %s\n", label[1] ? label[1] : file[1]);
}
2006-05-29 17:42:45 +05:30
printf("@@");
for (j = 0; j < 2; j++) {
int a = span[j].a = MAX(1, (*cvp)[j].a - opt_U_context);
int b = span[j].b = MIN(nlen[j], vec[idx][j].b + opt_U_context);
2006-05-29 17:42:45 +05:30
printf(" %c%d", j ? '+' : '-', MIN(a, b));
if (a == b)
continue;
printf(",%d", (a < b) ? b - a + 1 : 0);
}
puts(" @@");
/*
* Output changes in "unified" diff format--the old and new lines
* are printed together.
*/
for (lowa = span[0].a; ; lowa = (*cvp++)[0].b + 1) {
bool end = cvp > &vec[idx];
fetch(&ft[0], ix[0], lowa, end ? span[0].b : (*cvp)[0].a - 1, ' ');
if (end)
break;
for (j = 0; j < 2; j++)
fetch(&ft[j], ix[j], (*cvp)[j].a, (*cvp)[j].b, j ? '+' : '-');
}
}
anychange = true;
cont:
idx = -1;
} while (i <= nlen[0]);
free(vec);
free(ix[0]);
free(ix[1]);
free(J);
return anychange;
}
static int diffreg(char *file[2])
2006-05-29 17:42:45 +05:30
{
FILE *fp[2];
bool binary = false, differ = false;
int status = STATUS_SAME, i;
fp[0] = stdin;
fp[1] = stdin;
for (i = 0; i < 2; i++) {
int fd = STDIN_FILENO;
if (!LONE_DASH(file[i])) {
if (!(option_mask32 & FLAG(N))) {
fd = open_or_warn(file[i], O_RDONLY);
if (fd == -1)
goto out;
} else {
/* -N: if some file does not exist compare it like empty */
fd = open(file[i], O_RDONLY);
if (fd == -1)
fd = xopen("/dev/null", O_RDONLY);
}
}
/* Our diff implementation is using seek.
* When we meet non-seekable file, we must make a temp copy.
*/
if (lseek(fd, 0, SEEK_SET) == -1 && errno == ESPIPE) {
char name[] = "/tmp/difXXXXXX";
int fd_tmp = xmkstemp(name);
unlink(name);
if (bb_copyfd_eof(fd, fd_tmp) < 0)
xfunc_die();
if (fd != STDIN_FILENO)
close(fd);
fd = fd_tmp;
xlseek(fd, 0, SEEK_SET);
2007-03-27 04:28:21 +05:30
}
fp[i] = fdopen(fd, "r");
}
setup_common_bufsiz();
while (1) {
const size_t sz = COMMON_BUFSIZE / 2;
char *const buf0 = bb_common_bufsiz1;
char *const buf1 = buf0 + sz;
int j, k;
i = fread(buf0, 1, sz, fp[0]);
j = fread(buf1, 1, sz, fp[1]);
if (i != j) {
differ = true;
i = MIN(i, j);
}
if (i == 0)
break;
for (k = 0; k < i; k++) {
if (!buf0[k] || !buf1[k])
binary = true;
if (buf0[k] != buf1[k])
differ = true;
2007-03-27 04:28:21 +05:30
}
}
if (differ) {
if (binary && !(option_mask32 & FLAG(a)))
status = STATUS_BINARY;
else if (diff(fp, file))
status = STATUS_DIFFER;
}
if (status != STATUS_SAME)
exit_status |= 1;
out:
fclose_if_not_stdin(fp[0]);
fclose_if_not_stdin(fp[1]);
2006-05-29 17:42:45 +05:30
return status;
}
static void print_status(int status, char *path[2])
{
switch (status) {
case STATUS_BINARY:
case STATUS_DIFFER:
if ((option_mask32 & FLAG(q)) || status == STATUS_BINARY)
printf("Files %s and %s differ\n", path[0], path[1]);
break;
case STATUS_SAME:
if (option_mask32 & FLAG(s))
printf("Files %s and %s are identical\n", path[0], path[1]);
break;
}
}
#if ENABLE_FEATURE_DIFF_DIR
struct dlist {
size_t len;
int s, e;
char **dl;
};
/* This function adds a filename to dl, the directory listing. */
static int FAST_FUNC add_to_dirlist(const char *filename,
2008-07-05 14:48:54 +05:30
struct stat *sb UNUSED_PARAM,
void *userdata, int depth UNUSED_PARAM)
2006-05-29 17:42:45 +05:30
{
struct dlist *const l = userdata;
const char *file = filename + l->len;
while (*file == '/')
file++;
l->dl = xrealloc_vector(l->dl, 6, l->e);
l->dl[l->e] = xstrdup(file);
l->e++;
return TRUE;
}
/* If recursion is not set, this function adds the directory
* to the list and prevents recursive_action from recursing into it.
*/
static int FAST_FUNC skip_dir(const char *filename,
struct stat *sb, void *userdata,
int depth)
2006-05-29 17:42:45 +05:30
{
if (!(option_mask32 & FLAG(r)) && depth) {
add_to_dirlist(filename, sb, userdata, depth);
return SKIP;
}
if (!(option_mask32 & FLAG(N))) {
/* -r without -N: no need to recurse into dirs
* which do not exist on the "other side".
* Testcase: diff -r /tmp /
* (it would recurse deep into /proc without this code) */
struct dlist *const l = userdata;
filename += l->len;
if (filename[0]) {
struct stat osb;
char *othername = concat_path_file(G.other_dir, filename);
int r = stat(othername, &osb);
free(othername);
if (r != 0 || !S_ISDIR(osb.st_mode)) {
/* other dir doesn't have similarly named
* directory, don't recurse; return 1 upon
* exit, just like diffutils' diff */
exit_status |= 1;
return SKIP;
}
}
}
return TRUE;
}
static void diffdir(char *p[2], const char *s_start)
2006-05-29 17:42:45 +05:30
{
struct dlist list[2];
int i;
memset(&list, 0, sizeof(list));
for (i = 0; i < 2; i++) {
/*list[i].s = list[i].e = 0; - memset did it */
/*list[i].dl = NULL; */
G.other_dir = p[1 - i];
/* We need to trim root directory prefix.
* Using list.len to specify its length,
* add_to_dirlist will remove it. */
list[i].len = strlen(p[i]);
recursive_action(p[i], ACTION_RECURSE | ACTION_FOLLOWLINKS,
add_to_dirlist, skip_dir, &list[i], 0);
/* Sort dl alphabetically.
* GNU diff does this ignoring any number of trailing dots.
* We don't, so for us dotted files almost always are
* first on the list.
*/
qsort_string_vector(list[i].dl, list[i].e);
/* If -S was set, find the starting point. */
if (!s_start)
continue;
while (list[i].s < list[i].e && strcmp(list[i].dl[list[i].s], s_start) < 0)
list[i].s++;
}
/* Now that both dirlist1 and dirlist2 contain sorted directory
* listings, we can start to go through dirlist1. If both listings
* contain the same file, then do a normal diff. Otherwise, behaviour
2006-05-29 17:42:45 +05:30
* is determined by whether the -N flag is set. */
while (1) {
char *dp[2];
int pos;
int k;
dp[0] = list[0].s < list[0].e ? list[0].dl[list[0].s] : NULL;
dp[1] = list[1].s < list[1].e ? list[1].dl[list[1].s] : NULL;
if (!dp[0] && !dp[1])
break;
pos = !dp[0] ? 1 : (!dp[1] ? -1 : strcmp(dp[0], dp[1]));
k = pos > 0;
if (pos && !(option_mask32 & FLAG(N))) {
printf("Only in %s: %s\n", p[k], dp[k]);
exit_status |= 1;
} else {
char *fullpath[2], *path[2]; /* if -N */
for (i = 0; i < 2; i++) {
if (pos == 0 || i == k) {
path[i] = fullpath[i] = concat_path_file(p[i], dp[i]);
stat(fullpath[i], &stb[i]);
} else {
fullpath[i] = concat_path_file(p[i], dp[1 - i]);
path[i] = (char *)bb_dev_null;
}
}
if (pos)
stat(fullpath[k], &stb[1 - k]);
if (S_ISDIR(stb[0].st_mode) && S_ISDIR(stb[1].st_mode))
printf("Common subdirectories: %s and %s\n", fullpath[0], fullpath[1]);
else if (!S_ISREG(stb[0].st_mode) && !S_ISDIR(stb[0].st_mode))
printf("File %s is not a regular file or directory and was skipped\n", fullpath[0]);
else if (!S_ISREG(stb[1].st_mode) && !S_ISDIR(stb[1].st_mode))
printf("File %s is not a regular file or directory and was skipped\n", fullpath[1]);
else if (S_ISDIR(stb[0].st_mode) != S_ISDIR(stb[1].st_mode)) {
if (S_ISDIR(stb[0].st_mode))
printf("File %s is a %s while file %s is a %s\n", fullpath[0], "directory", fullpath[1], "regular file");
else
printf("File %s is a %s while file %s is a %s\n", fullpath[0], "regular file", fullpath[1], "directory");
} else
print_status(diffreg(path), fullpath);
free(fullpath[0]);
free(fullpath[1]);
}
free(dp[k]);
list[k].s++;
if (pos == 0) {
free(dp[1 - k]);
list[1 - k].s++;
}
}
if (ENABLE_FEATURE_CLEAN_UP) {
free(list[0].dl);
free(list[1].dl);
}
}
#endif
#if ENABLE_FEATURE_DIFF_LONG_OPTIONS
static const char diff_longopts[] ALIGN1 =
"ignore-case\0" No_argument "i"
"ignore-tab-expansion\0" No_argument "E"
"ignore-space-change\0" No_argument "b"
"ignore-all-space\0" No_argument "w"
"ignore-blank-lines\0" No_argument "B"
"text\0" No_argument "a"
"unified\0" Required_argument "U"
"label\0" Required_argument "L"
"show-c-function\0" No_argument "p"
"brief\0" No_argument "q"
"expand-tabs\0" No_argument "t"
"initial-tab\0" No_argument "T"
"recursive\0" No_argument "r"
"new-file\0" No_argument "N"
"report-identical-files\0" No_argument "s"
"starting-file\0" Required_argument "S"
"minimal\0" No_argument "d"
;
getopt32: remove applet_long_options FEATURE_GETOPT_LONG made dependent on LONG_OPTS. The folloving options are removed, now LONG_OPTS enables long options for affected applets: FEATURE_ENV_LONG_OPTIONS FEATURE_EXPAND_LONG_OPTIONS FEATURE_UNEXPAND_LONG_OPTIONS FEATURE_MKDIR_LONG_OPTIONS FEATURE_MV_LONG_OPTIONS FEATURE_RMDIR_LONG_OPTIONS FEATURE_ADDGROUP_LONG_OPTIONS FEATURE_ADDUSER_LONG_OPTIONS FEATURE_HWCLOCK_LONG_OPTIONS FEATURE_NSENTER_LONG_OPTS FEATURE_CHCON_LONG_OPTIONS FEATURE_RUNCON_LONG_OPTIONS They either had a small number of long options, or their long options are essential. Example: upstream addgroup and adduser have ONLY longopts, we should probably go further and get rid of non-standard short options. To this end, make addgroup and adduser "select LONG_OPTS". We had this breakage caused by us even in our own package! #if ENABLE_LONG_OPTS || !ENABLE_ADDGROUP /* We try to use --gid, not -g, because "standard" addgroup * has no short option -g, it has only long --gid. */ argv[1] = (char*)"--gid"; #else /* Breaks if system in fact does NOT use busybox addgroup */ argv[1] = (char*)"-g"; #endif xargs: its lone longopt no longer depends on DESKTOP, only on LONG_OPTS. hwclock TODO: get rid of incompatible -t, -l aliases to --systz, --localtime Shorten help texts by omitting long option when short opt alternative exists. Reduction of size comes from the fact that store of an immediate (an address of longopts) to a fixed address (global variable) is a longer insn than pushing that immediate or passing it in a register. This effect is CPU-agnostic. function old new delta getopt32 1350 22 -1328 vgetopt32 - 1318 +1318 getopt32long - 24 +24 tftpd_main 562 567 +5 scan_recursive 376 380 +4 collect_cpu 545 546 +1 date_main 1096 1095 -1 hostname_main 262 259 -3 uname_main 259 255 -4 setpriv_main 362 358 -4 rmdir_main 191 187 -4 mv_main 562 558 -4 ipcalc_main 548 544 -4 ifenslave_main 641 637 -4 gzip_main 192 188 -4 gunzip_main 77 73 -4 fsfreeze_main 81 77 -4 flock_main 318 314 -4 deluser_main 337 333 -4 cp_main 374 370 -4 chown_main 175 171 -4 applet_long_options 4 - -4 xargs_main 894 889 -5 wget_main 2540 2535 -5 udhcpc_main 2767 2762 -5 touch_main 436 431 -5 tar_main 1014 1009 -5 start_stop_daemon_main 1033 1028 -5 sed_main 682 677 -5 script_main 1082 1077 -5 run_parts_main 330 325 -5 rtcwake_main 459 454 -5 od_main 2169 2164 -5 nl_main 201 196 -5 modprobe_main 773 768 -5 mkdir_main 160 155 -5 ls_main 568 563 -5 install_main 773 768 -5 hwclock_main 411 406 -5 getopt_main 622 617 -5 fstrim_main 256 251 -5 env_main 198 193 -5 dumpleases_main 635 630 -5 dpkg_main 3991 3986 -5 diff_main 1355 1350 -5 cryptpw_main 233 228 -5 cpio_main 593 588 -5 conspy_main 1135 1130 -5 chpasswd_main 313 308 -5 adduser_main 887 882 -5 addgroup_main 416 411 -5 ftpgetput_main 351 345 -6 get_terminal_width_height 242 234 -8 expand_main 690 680 -10 static.expand_longopts 18 - -18 static.unexpand_longopts 27 - -27 mkdir_longopts 28 - -28 env_longopts 30 - -30 static.ifenslave_longopts 34 - -34 mv_longopts 46 - -46 static.rmdir_longopts 48 - -48 packed_usage 31739 31687 -52 ------------------------------------------------------------------------------ (add/remove: 2/8 grow/shrink: 3/49 up/down: 1352/-1840) Total: -488 bytes text data bss dec hex filename 915681 485 6880 923046 e15a6 busybox_old 915428 485 6876 922789 e14a5 busybox_unstripped Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2017-08-08 20:08:18 +05:30
# define GETOPT32 getopt32long
# define LONGOPTS ,diff_longopts
#else
# define GETOPT32 getopt32
# define LONGOPTS
#endif
int diff_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
2008-07-05 14:48:54 +05:30
int diff_main(int argc UNUSED_PARAM, char **argv)
2006-05-29 17:42:45 +05:30
{
int gotstdin = 0, i;
char *file[2], *s_start = NULL;
llist_t *L_arg = NULL;
2006-05-29 17:42:45 +05:30
INIT_G();
/* exactly 2 params; collect multiple -L <label>; -U N */
getopt32: remove opt_complementary function old new delta vgetopt32 1318 1392 +74 runsvdir_main 703 713 +10 bb_make_directory 423 425 +2 collect_cpu 546 545 -1 opt_chars 3 - -3 opt_complementary 4 - -4 tftpd_main 567 562 -5 ntp_init 476 471 -5 zcip_main 1266 1256 -10 xxd_main 428 418 -10 whois_main 140 130 -10 who_main 463 453 -10 which_main 212 202 -10 wget_main 2535 2525 -10 watchdog_main 291 281 -10 watch_main 222 212 -10 vlock_main 399 389 -10 uuencode_main 332 322 -10 uudecode_main 316 306 -10 unlink_main 45 35 -10 udhcpd_main 1482 1472 -10 udhcpc_main 2762 2752 -10 tune2fs_main 290 280 -10 tunctl_main 366 356 -10 truncate_main 218 208 -10 tr_main 518 508 -10 time_main 1134 1124 -10 tftp_main 286 276 -10 telnetd_main 1873 1863 -10 tcpudpsvd_main 1785 1775 -10 taskset_main 521 511 -10 tar_main 1009 999 -10 tail_main 1644 1634 -10 syslogd_main 1967 1957 -10 switch_root_main 368 358 -10 svlogd_main 1454 1444 -10 sv 1296 1286 -10 stat_main 104 94 -10 start_stop_daemon_main 1028 1018 -10 split_main 542 532 -10 sort_main 796 786 -10 slattach_main 624 614 -10 shuf_main 504 494 -10 setsid_main 96 86 -10 setserial_main 1132 1122 -10 setfont_main 388 378 -10 setconsole_main 78 68 -10 sendmail_main 1209 1199 -10 sed_main 677 667 -10 script_main 1077 1067 -10 run_parts_main 325 315 -10 rtcwake_main 454 444 -10 rm_main 175 165 -10 reformime_main 119 109 -10 readlink_main 123 113 -10 rdate_main 246 236 -10 pwdx_main 189 179 -10 pstree_main 317 307 -10 pscan_main 663 653 -10 popmaildir_main 818 808 -10 pmap_main 80 70 -10 nc_main 1042 1032 -10 mv_main 558 548 -10 mountpoint_main 477 467 -10 mount_main 1264 1254 -10 modprobe_main 768 758 -10 modinfo_main 333 323 -10 mktemp_main 200 190 -10 mkswap_main 324 314 -10 mkfs_vfat_main 1489 1479 -10 microcom_main 715 705 -10 md5_sha1_sum_main 521 511 -10 man_main 867 857 -10 makedevs_main 1052 1042 -10 ls_main 563 553 -10 losetup_main 432 422 -10 loadfont_main 89 79 -10 ln_main 524 514 -10 link_main 75 65 -10 ipcalc_main 544 534 -10 iostat_main 2397 2387 -10 install_main 768 758 -10 id_main 480 470 -10 i2cset_main 1239 1229 -10 i2cget_main 380 370 -10 i2cdump_main 1482 1472 -10 i2cdetect_main 682 672 -10 hwclock_main 406 396 -10 httpd_main 741 731 -10 grep_main 837 827 -10 getty_main 1559 1549 -10 fuser_main 297 287 -10 ftpgetput_main 345 335 -10 ftpd_main 2232 2222 -10 fstrim_main 251 241 -10 fsfreeze_main 77 67 -10 fsck_minix_main 2921 2911 -10 flock_main 314 304 -10 flashcp_main 740 730 -10 flash_eraseall_main 833 823 -10 fdformat_main 532 522 -10 expand_main 680 670 -10 eject_main 335 325 -10 dumpleases_main 630 620 -10 du_main 314 304 -10 dos2unix_main 441 431 -10 diff_main 1350 1340 -10 df_main 1064 1054 -10 date_main 1095 1085 -10 cut_main 961 951 -10 cryptpw_main 228 218 -10 crontab_main 575 565 -10 crond_main 1149 1139 -10 cp_main 370 360 -10 common_traceroute_main 3834 3824 -10 common_ping_main 1767 1757 -10 comm_main 239 229 -10 cmp_main 655 645 -10 chrt_main 379 369 -10 chpst_main 704 694 -10 chpasswd_main 308 298 -10 chown_main 171 161 -10 chmod_main 158 148 -10 cat_main 428 418 -10 bzip2_main 120 110 -10 blkdiscard_main 264 254 -10 base64_main 221 211 -10 arping_main 1665 1655 -10 ar_main 556 546 -10 adjtimex_main 406 396 -10 adduser_main 882 872 -10 addgroup_main 411 401 -10 acpid_main 1198 1188 -10 optstring 11 - -11 opt_string 18 - -18 OPT_STR 25 - -25 ubi_tools_main 1288 1258 -30 ls_options 31 - -31 ------------------------------------------------------------------------------ (add/remove: 0/6 grow/shrink: 3/129 up/down: 86/-1383) Total: -1297 bytes text data bss dec hex filename 915428 485 6876 922789 e14a5 busybox_old 914629 485 6872 921986 e1182 busybox_unstripped Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2017-08-09 01:25:02 +05:30
GETOPT32(argv, "^" "abdiL:*NqrsS:tTU:+wupBE" "\0" "=2"
LONGOPTS,
&L_arg, &s_start, &opt_U_context);
argv += optind;
while (L_arg)
label[!!label[0]] = llist_pop(&L_arg);
/* Compat: "diff file name_which_doesnt_exist" exits with 2 */
xfunc_error_retval = 2;
for (i = 0; i < 2; i++) {
file[i] = argv[i];
if (LONE_DASH(file[i])) {
fstat(STDIN_FILENO, &stb[i]);
gotstdin++;
} else if (option_mask32 & FLAG(N)) {
if (stat(file[i], &stb[i]))
xstat("/dev/null", &stb[i]);
} else {
xstat(file[i], &stb[i]);
}
}
xfunc_error_retval = 1;
if (gotstdin && (S_ISDIR(stb[0].st_mode) || S_ISDIR(stb[1].st_mode)))
bb_error_msg_and_die("can't compare stdin to a directory");
/* Compare metadata to check if the files are the same physical file.
*
* Comment from diffutils source says:
* POSIX says that two files are identical if st_ino and st_dev are
* the same, but many file systems incorrectly assign the same (device,
* inode) pair to two distinct files, including:
* GNU/Linux NFS servers that export all local file systems as a
* single NFS file system, if a local device number (st_dev) exceeds
* 255, or if a local inode number (st_ino) exceeds 16777215.
*/
if (ENABLE_DESKTOP
&& stb[0].st_ino == stb[1].st_ino
&& stb[0].st_dev == stb[1].st_dev
&& stb[0].st_size == stb[1].st_size
&& stb[0].st_mtime == stb[1].st_mtime
&& stb[0].st_ctime == stb[1].st_ctime
&& stb[0].st_mode == stb[1].st_mode
&& stb[0].st_nlink == stb[1].st_nlink
&& stb[0].st_uid == stb[1].st_uid
&& stb[0].st_gid == stb[1].st_gid
) {
/* files are physically the same; no need to compare them */
return STATUS_SAME;
}
if (S_ISDIR(stb[0].st_mode) && S_ISDIR(stb[1].st_mode)) {
#if ENABLE_FEATURE_DIFF_DIR
diffdir(file, s_start);
#else
bb_error_msg_and_die("no support for directory comparison");
#endif
} else {
bool dirfile = S_ISDIR(stb[0].st_mode) || S_ISDIR(stb[1].st_mode);
bool dir = S_ISDIR(stb[1].st_mode);
if (dirfile) {
const char *slash = strrchr(file[!dir], '/');
file[dir] = concat_path_file(file[dir], slash ? slash + 1 : file[!dir]);
xstat(file[dir], &stb[dir]);
}
/* diffreg can get non-regular files here */
print_status(gotstdin > 1 ? STATUS_SAME : diffreg(file), file);
if (dirfile)
free(file[dir]);
}
return exit_status;
}