mirror of
https://github.com/netsurf-browser/netsurf
synced 2024-11-23 14:59:47 +03:00
c105738fa3
This changes the LOG macro to be varadic removing the need for all callsites to have double bracketing and allows for future improvement on how we use the logging macros. The callsites were changed with coccinelle and the changes checked by hand. Compile tested for several frontends but not all. A formatting annotation has also been added which allows the compiler to check the parameters and types passed to the logging.
631 lines
13 KiB
C
631 lines
13 KiB
C
/*
|
|
* Copyright 2007 Rob Kendrick <rjek@netsurf-browser.org>
|
|
* Copyright 2004-2007 James Bursa <bursa@users.sourceforge.net>
|
|
* Copyright 2003 Phil Mellor <monkeyson@users.sourceforge.net>
|
|
* Copyright 2003 John M Bell <jmb202@ecs.soton.ac.uk>
|
|
* Copyright 2004 John Tytgat <joty@netsurf-browser.org>
|
|
*
|
|
* This file is part of NetSurf, http://www.netsurf-browser.org/
|
|
*
|
|
* NetSurf is free software; you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation; version 2 of the License.
|
|
*
|
|
* NetSurf is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
*/
|
|
|
|
#include <assert.h>
|
|
#include <ctype.h>
|
|
#include <stdlib.h>
|
|
#include <stdio.h>
|
|
#include <string.h>
|
|
#include <strings.h>
|
|
#include <sys/stat.h>
|
|
#include <sys/types.h>
|
|
#include <sys/time.h>
|
|
#include <regex.h>
|
|
#include <time.h>
|
|
#include <errno.h>
|
|
|
|
#include "utils/config.h"
|
|
#include "utils/log.h"
|
|
#include "utils/messages.h"
|
|
#include "utils/utf8.h"
|
|
#include "utils/time.h"
|
|
#include "utils/utils.h"
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
char *remove_underscores(const char *s, bool replacespace)
|
|
{
|
|
size_t i, ii, len;
|
|
char *ret;
|
|
len = strlen(s);
|
|
ret = malloc(len + 1);
|
|
if (ret == NULL)
|
|
return NULL;
|
|
for (i = 0, ii = 0; i < len; i++) {
|
|
if (s[i] != '_')
|
|
ret[ii++] = s[i];
|
|
else if (replacespace)
|
|
ret[ii++] = ' ';
|
|
}
|
|
ret[ii] = '\0';
|
|
return ret;
|
|
}
|
|
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
char *squash_whitespace(const char *s)
|
|
{
|
|
char *c;
|
|
int i = 0, j = 0;
|
|
|
|
c = malloc(strlen(s) + 1);
|
|
if (c != NULL) {
|
|
do {
|
|
if (s[i] == ' ' ||
|
|
s[i] == '\n' ||
|
|
s[i] == '\r' ||
|
|
s[i] == '\t') {
|
|
c[j++] = ' ';
|
|
while (s[i] == ' ' ||
|
|
s[i] == '\n' ||
|
|
s[i] == '\r' ||
|
|
s[i] == '\t')
|
|
i++;
|
|
}
|
|
c[j++] = s[i++];
|
|
} while (s[i - 1] != 0);
|
|
}
|
|
return c;
|
|
}
|
|
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
char *cnv_space2nbsp(const char *s)
|
|
{
|
|
const char *srcP;
|
|
char *d, *d0;
|
|
unsigned int numNBS;
|
|
/* Convert space & TAB into non breaking space character (0xA0) */
|
|
for (numNBS = 0, srcP = (const char *)s; *srcP != '\0'; ++srcP)
|
|
if (*srcP == ' ' || *srcP == '\t')
|
|
++numNBS;
|
|
if ((d = (char *)malloc((srcP - s) + numNBS + 1)) == NULL)
|
|
return NULL;
|
|
for (d0 = d, srcP = (const char *)s; *srcP != '\0'; ++srcP) {
|
|
if (*srcP == ' ' || *srcP == '\t') {
|
|
*d0++ = 0xC2;
|
|
*d0++ = 0xA0;
|
|
} else
|
|
*d0++ = *srcP;
|
|
}
|
|
*d0 = '\0';
|
|
return d;
|
|
}
|
|
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
bool is_dir(const char *path)
|
|
{
|
|
struct stat s;
|
|
|
|
if (stat(path, &s))
|
|
return false;
|
|
|
|
return S_ISDIR(s.st_mode) ? true : false;
|
|
}
|
|
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
nserror vsnstrjoin(char **str, size_t *size, char sep, size_t nelm, va_list ap)
|
|
{
|
|
const char *elm[16];
|
|
size_t elm_len[16];
|
|
size_t elm_idx;
|
|
char *fname;
|
|
size_t fname_len = 0;
|
|
char *curp;
|
|
|
|
/* check the parameters are all sensible */
|
|
if ((nelm == 0) || (nelm > 16)) {
|
|
return NSERROR_BAD_PARAMETER;
|
|
}
|
|
if ((*str != NULL) && (size == NULL)) {
|
|
/* if the caller is providing the buffer they must say
|
|
* how much space is available.
|
|
*/
|
|
return NSERROR_BAD_PARAMETER;
|
|
}
|
|
|
|
/* calculate how much storage we need for the complete path
|
|
* with all the elements.
|
|
*/
|
|
for (elm_idx = 0; elm_idx < nelm; elm_idx++) {
|
|
elm[elm_idx] = va_arg(ap, const char *);
|
|
/* check the argument is not NULL */
|
|
if (elm[elm_idx] == NULL) {
|
|
return NSERROR_BAD_PARAMETER;
|
|
}
|
|
elm_len[elm_idx] = strlen(elm[elm_idx]);
|
|
fname_len += elm_len[elm_idx];
|
|
}
|
|
fname_len += nelm; /* allow for separators and terminator */
|
|
|
|
/* ensure there is enough space */
|
|
fname = *str;
|
|
if (fname != NULL) {
|
|
if (fname_len > *size) {
|
|
return NSERROR_NOSPACE;
|
|
}
|
|
} else {
|
|
fname = malloc(fname_len);
|
|
if (fname == NULL) {
|
|
return NSERROR_NOMEM;
|
|
}
|
|
}
|
|
|
|
/* copy the elements in with apropriate separator */
|
|
curp = fname;
|
|
for (elm_idx = 0; elm_idx < nelm; elm_idx++) {
|
|
memmove(curp, elm[elm_idx], elm_len[elm_idx]);
|
|
curp += elm_len[elm_idx];
|
|
/* ensure string are separated */
|
|
if (curp[-1] != sep) {
|
|
*curp = sep;
|
|
curp++;
|
|
}
|
|
}
|
|
curp[-1] = 0; /* NULL terminate */
|
|
|
|
assert((curp - fname) <= (int)fname_len);
|
|
|
|
*str = fname;
|
|
if (size != NULL) {
|
|
*size = fname_len;
|
|
}
|
|
|
|
return NSERROR_OK;
|
|
}
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
nserror snstrjoin(char **str, size_t *size, char sep, size_t nelm, ...)
|
|
{
|
|
va_list ap;
|
|
nserror ret;
|
|
|
|
va_start(ap, nelm);
|
|
ret = vsnstrjoin(str, size, sep, nelm, ap);
|
|
va_end(ap);
|
|
|
|
return ret;
|
|
}
|
|
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
nserror regcomp_wrapper(regex_t *preg, const char *regex, int cflags)
|
|
{
|
|
int r;
|
|
r = regcomp(preg, regex, cflags);
|
|
if (r) {
|
|
char errbuf[200];
|
|
regerror(r, preg, errbuf, sizeof errbuf);
|
|
LOG("Failed to compile regexp '%s': %s\n", regex, errbuf);
|
|
return NSERROR_INIT_FAILED;
|
|
}
|
|
return NSERROR_OK;
|
|
}
|
|
|
|
|
|
/**
|
|
* The size of buffers within human_friendly_bytesize.
|
|
*
|
|
* We can have a fairly good estimate of how long the buffer needs to
|
|
* be. The unsigned long can store a value representing a maximum
|
|
* size of around 4 GB. Therefore the greatest space required is to
|
|
* represent 1023MB. Currently that would be represented as "1023MB"
|
|
* so 12 including a null terminator. Ideally we would be able to
|
|
* know this value for sure, in the mean time the following should
|
|
* suffice.
|
|
*/
|
|
#define BYTESIZE_BUFFER_SIZE 20
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
char *human_friendly_bytesize(unsigned long bsize) {
|
|
static char buffer1[BYTESIZE_BUFFER_SIZE];
|
|
static char buffer2[BYTESIZE_BUFFER_SIZE];
|
|
static char buffer3[BYTESIZE_BUFFER_SIZE];
|
|
static char *curbuffer = buffer3;
|
|
enum {bytes, kilobytes, megabytes, gigabytes} unit = bytes;
|
|
static char units[][7] = {"Bytes", "kBytes", "MBytes", "GBytes"};
|
|
|
|
float bytesize = (float)bsize;
|
|
|
|
if (curbuffer == buffer1)
|
|
curbuffer = buffer2;
|
|
else if (curbuffer == buffer2)
|
|
curbuffer = buffer3;
|
|
else
|
|
curbuffer = buffer1;
|
|
|
|
if (bytesize > 1024) {
|
|
bytesize /= 1024;
|
|
unit = kilobytes;
|
|
}
|
|
|
|
if (bytesize > 1024) {
|
|
bytesize /= 1024;
|
|
unit = megabytes;
|
|
}
|
|
|
|
if (bytesize > 1024) {
|
|
bytesize /= 1024;
|
|
unit = gigabytes;
|
|
}
|
|
|
|
snprintf(curbuffer, BYTESIZE_BUFFER_SIZE, "%3.2f%s", bytesize, messages_get(units[unit]));
|
|
|
|
return curbuffer;
|
|
}
|
|
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
const char *rfc1123_date(time_t t)
|
|
{
|
|
static char ret[30];
|
|
|
|
struct tm *tm = gmtime(&t);
|
|
const char *days[] = { "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat" },
|
|
*months[] = { "Jan", "Feb", "Mar", "Apr", "May", "Jun",
|
|
"Jul", "Aug", "Sep", "Oct", "Nov", "Dec" };
|
|
|
|
snprintf(ret, sizeof ret, "%s, %02d %s %d %02d:%02d:%02d GMT",
|
|
days[tm->tm_wday], tm->tm_mday, months[tm->tm_mon],
|
|
tm->tm_year + 1900, tm->tm_hour, tm->tm_min,
|
|
tm->tm_sec);
|
|
|
|
return ret;
|
|
}
|
|
|
|
|
|
/* exported interface documented in utils/utils.h */
|
|
unsigned int wallclock(void)
|
|
{
|
|
struct timeval tv;
|
|
|
|
if (gettimeofday(&tv, NULL) == -1)
|
|
return 0;
|
|
|
|
return ((tv.tv_sec * 100) + (tv.tv_usec / 10000));
|
|
}
|
|
|
|
#ifndef HAVE_STRCASESTR
|
|
|
|
/**
|
|
* Case insensitive strstr implementation
|
|
*
|
|
* \param haystack String to search in
|
|
* \param needle String to look for
|
|
* \return Pointer to start of found substring, or NULL if not found
|
|
*/
|
|
char *strcasestr(const char *haystack, const char *needle)
|
|
{
|
|
size_t needle_len = strlen(needle);
|
|
const char * last_start = haystack + (strlen(haystack) - needle_len);
|
|
|
|
while (haystack <= last_start) {
|
|
if (strncasecmp(haystack, needle, needle_len) == 0)
|
|
return (char *)haystack;
|
|
haystack++;
|
|
}
|
|
|
|
return NULL;
|
|
}
|
|
|
|
#endif
|
|
|
|
#ifndef HAVE_STRNDUP
|
|
|
|
/**
|
|
* Duplicate up to n characters of a string.
|
|
*/
|
|
|
|
char *strndup(const char *s, size_t n)
|
|
{
|
|
size_t len;
|
|
char *s2;
|
|
|
|
for (len = 0; len != n && s[len]; len++)
|
|
continue;
|
|
|
|
s2 = malloc(len + 1);
|
|
if (!s2)
|
|
return 0;
|
|
|
|
memcpy(s2, s, len);
|
|
s2[len] = 0;
|
|
return s2;
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
/* Exported interface, documented in utils.h */
|
|
int dir_sort_alpha(const struct dirent **d1, const struct dirent **d2)
|
|
{
|
|
const char *s1 = (*d1)->d_name;
|
|
const char *s2 = (*d2)->d_name;
|
|
|
|
while (*s1 != '\0' && *s2 != '\0') {
|
|
if ((*s1 >= '0' && *s1 <= '9') &&
|
|
(*s2 >= '0' && *s2 <= '9')) {
|
|
int n1 = 0, n2 = 0;
|
|
while (*s1 >= '0' && *s1 <= '9') {
|
|
n1 = n1 * 10 + (*s1) - '0';
|
|
s1++;
|
|
}
|
|
while (*s2 >= '0' && *s2 <= '9') {
|
|
n2 = n2 * 10 + (*s2) - '0';
|
|
s2++;
|
|
}
|
|
if (n1 != n2) {
|
|
return n1 - n2;
|
|
}
|
|
if (*s1 == '\0' || *s2 == '\0')
|
|
break;
|
|
}
|
|
if (tolower(*s1) != tolower(*s2))
|
|
break;
|
|
|
|
s1++;
|
|
s2++;
|
|
}
|
|
|
|
return tolower(*s1) - tolower(*s2);
|
|
}
|
|
|
|
|
|
#ifndef HAVE_SCANDIR
|
|
int alphasort(const struct dirent **d1, const struct dirent **d2)
|
|
{
|
|
return strcasecmp((*d1)->d_name, (*d2)->d_name);
|
|
}
|
|
|
|
int scandir(const char *dir, struct dirent ***namelist,
|
|
int (*sel)(const struct dirent *),
|
|
int (*compar)(const struct dirent **, const struct dirent **))
|
|
{
|
|
struct dirent **entlist = NULL;
|
|
struct dirent **entlist_temp = NULL;
|
|
struct dirent *ent = NULL, *new_ent;
|
|
int alloc_n = 0;
|
|
int n = 0;
|
|
DIR *d;
|
|
|
|
d = opendir(dir);
|
|
if (d == NULL) {
|
|
goto error;
|
|
}
|
|
|
|
while ((ent = readdir(d)) != NULL) {
|
|
/* Avoid entries that caller doesn't want */
|
|
if (sel && (*sel)(ent) == 0)
|
|
continue;
|
|
|
|
/* Ensure buffer is big enough to list this entry */
|
|
if (n == alloc_n) {
|
|
alloc_n *= 4;
|
|
if (alloc_n == 0) {
|
|
alloc_n = 32;
|
|
}
|
|
entlist_temp = realloc(entlist,
|
|
sizeof(*entlist) * alloc_n);
|
|
if (entlist_temp == NULL) {
|
|
goto error;
|
|
}
|
|
entlist = entlist_temp;
|
|
}
|
|
|
|
/* Make copy of ent */
|
|
new_ent = malloc(sizeof(*new_ent));
|
|
if (new_ent == NULL) {
|
|
goto error;
|
|
}
|
|
memcpy(new_ent, ent, sizeof(struct dirent));
|
|
|
|
/* Make list entry point to this copy of ent */
|
|
entlist[n] = new_ent;
|
|
|
|
n++;
|
|
}
|
|
|
|
closedir(d);
|
|
|
|
/* Sort */
|
|
if (compar != NULL && n > 1)
|
|
qsort(entlist, n, sizeof(*entlist),
|
|
(int (*)(const void *, const void *))compar);
|
|
*namelist = entlist;
|
|
return n;
|
|
|
|
error:
|
|
|
|
if (entlist != NULL) {
|
|
int i;
|
|
for (i = 0; i < n; i++) {
|
|
free(entlist[i]);
|
|
}
|
|
free(entlist);
|
|
}
|
|
|
|
if (d != NULL) {
|
|
closedir(d);
|
|
}
|
|
|
|
return -1;
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
#ifndef HAVE_STRCHRNUL
|
|
|
|
/**
|
|
* Find the first occurrence of C in S or the final NUL byte.
|
|
*/
|
|
char *strchrnul (const char *s, int c_in)
|
|
{
|
|
const unsigned char *us = (const unsigned char *) s;
|
|
|
|
while (*us != c_in && *us != '\0')
|
|
us++;
|
|
|
|
return (void *) us;
|
|
}
|
|
|
|
#endif
|
|
|
|
#ifndef HAVE_UTSNAME
|
|
#include "utils/utsname.h"
|
|
|
|
int uname(struct utsname *buf) {
|
|
strcpy(buf->sysname,"windows");
|
|
strcpy(buf->nodename,"nodename");
|
|
strcpy(buf->release,"release");
|
|
strcpy(buf->version,"version");
|
|
strcpy(buf->machine,"pc");
|
|
|
|
return 0;
|
|
}
|
|
#endif
|
|
|
|
#ifndef HAVE_REALPATH
|
|
char *realpath(const char *path, char *resolved_path)
|
|
{
|
|
char *ret;
|
|
if (resolved_path == NULL) {
|
|
ret=strdup(path);
|
|
} else {
|
|
ret = resolved_path;
|
|
strcpy(resolved_path, path);
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
#ifndef HAVE_INETATON
|
|
|
|
|
|
int inet_aton(const char *cp, struct in_addr *inp)
|
|
{
|
|
unsigned int b1, b2, b3, b4;
|
|
unsigned char c;
|
|
|
|
if (strspn(cp, "0123456789.") < strlen(cp))
|
|
return 0;
|
|
|
|
if (sscanf(cp, "%3u.%3u.%3u.%3u%c", &b1, &b2, &b3, &b4, &c) != 4)
|
|
return 0;
|
|
|
|
if ((b1 > 255) || (b2 > 255) || (b3 > 255) || (b4 > 255))
|
|
return 0;
|
|
|
|
inp->s_addr = b4 << 24 | b3 << 16 | b2 << 8 | b1;
|
|
|
|
return 1;
|
|
}
|
|
|
|
#endif
|
|
|
|
#ifndef HAVE_INETPTON
|
|
|
|
int inet_pton(int af, const char *src, void *dst)
|
|
{
|
|
int ret;
|
|
|
|
if (af == AF_INET) {
|
|
ret = inet_aton(src, dst);
|
|
}
|
|
#if !defined(NO_IPV6)
|
|
else if (af == AF_INET6) {
|
|
/* TODO: implement v6 address support */
|
|
ret = -1;
|
|
errno = EAFNOSUPPORT;
|
|
}
|
|
#endif
|
|
else {
|
|
ret = -1;
|
|
errno = EAFNOSUPPORT;
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
#endif
|
|
|
|
/* exported function documented in utils/time.h */
|
|
int nsc_sntimet(char *str, size_t size, time_t *timep)
|
|
{
|
|
#ifndef HAVE_STRFTIME
|
|
long long val;
|
|
val = (long long)*timep;
|
|
|
|
return snprintf(str, size, "%lld", val);
|
|
#else
|
|
struct tm *ltm;
|
|
|
|
ltm = localtime(timep);
|
|
if (ltm == NULL) {
|
|
return -1;
|
|
}
|
|
|
|
return strftime(str, size, "%s", ltm);
|
|
#endif
|
|
}
|
|
|
|
nserror nsc_snptimet(char *str, size_t size, time_t *timep)
|
|
{
|
|
time_t time_out;
|
|
|
|
#ifndef HAVE_STRPTIME
|
|
char *rstr;
|
|
|
|
if (size < 1) {
|
|
return NSERROR_BAD_PARAMETER;
|
|
}
|
|
|
|
errno = 0;
|
|
time_out = (time_t)strtoll(str, &rstr, 10);
|
|
|
|
/* The conversion may have a range faliure or no digits were found */
|
|
if ((errno != 0) || (rstr == str)) {
|
|
return NSERROR_BAD_PARAMETER;
|
|
}
|
|
|
|
#else
|
|
struct tm ltm;
|
|
|
|
if (size < 1) {
|
|
return NSERROR_BAD_PARAMETER;
|
|
}
|
|
|
|
if (strptime(str, "%s", <m) == NULL) {
|
|
return NSERROR_BAD_PARAMETER;
|
|
}
|
|
|
|
time_out = mktime(<m);
|
|
|
|
#endif
|
|
*timep = time_out;
|
|
|
|
return NSERROR_OK;
|
|
}
|