Skip to content
Permalink
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
547 lines (460 sloc) 10.3 KB
/*
* uhub - A tiny ADC p2p connection hub
* Copyright (C) 2007-2014, Jan Vidar Krey
*
* This program 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; either version 3 of the License, or
* (at your option) any later version.
*
* This program 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 <https://www.gnu.org/licenses/>.
*
*/
#include "uhub.h"
static const char* BASE32_ALPHABET = "ABCDEFGHIJKLMNOPQRSTUVWXYZ234567";
int is_space(char c)
{
if (c == ' ') return 1;
return 0;
}
int is_white_space(char c)
{
if (c == ' ' || c == '\t' || c == '\r') return 1;
return 0;
}
static int is_printable(unsigned char c)
{
if (c >= 32)
return 1;
if (c == '\t' || c == '\r' || c == '\n')
return 1;
return 0;
}
char* strip_white_space(char* string)
{
char* pos;
if (!string)
return "";
while (string[0] && is_white_space(string[0])) string++;
if (!*string)
return string;
/* Strip appending whitespace */
pos = &string[strlen(string)-1];
while (&string[0] < &pos[0] && is_white_space(pos[0])) { pos[0] = 0; pos--; }
return string;
}
static int is_valid_utf8_str(const unsigned char* string, size_t length)
{
int expect = 0;
char div = 0;
size_t pos = 0;
if (length == 0) return 1;
for (pos = 0; pos < length; pos++)
{
if (expect)
{
if ((string[pos] & 0xC0) == 0x80) expect--;
else return 0;
}
else
{
if (string[pos] & 0x80)
{
for (div = 0x40; div > 0x08; div /= 2)
{
if (string[pos] & div) expect++;
else break;
}
if ((string[pos] & div) || (pos+expect >= length)) return 0;
switch (expect) {
case 0:
return 0;
case 1:
/* Out of range */
if (string[pos] < 0xC2) return 0;
break;
case 2:
/* Out of range */
if ((string[pos] == 0xE0) && (string[pos+1] < 0xA0 )) return 0;
/* Surrogates */
if ((string[pos] == 0xED) && (string[pos+1] > 0x9F )) return 0;
break;
case 3:
/* Out of range */
if ((string[pos] == 0xF0) && (string[pos+1] < 0x90 )) return 0;
if (string[pos] > 0xF4) return 0;
if ((string[pos] == 0xF4) && (string[pos+1] > 0x8F )) return 0;
break;
}
}
}
}
return 1;
}
int is_valid_utf8(const char* string)
{
return is_valid_utf8_str(string, strlen(string));
}
int is_printable_utf8(const char* string, size_t length)
{
size_t pos = 0;
for (pos = 0; pos < length; pos++)
{
if (!is_printable(string[pos]))
return 0;
}
return is_valid_utf8_str(string, length);
}
int is_valid_base32_char(char c)
{
if ((c >= 'A' && c <= 'Z') || (c >= '2' && c <= '7')) return 1;
return 0;
}
int is_num(char c)
{
if (c >= '0' && c <= '9') return 1;
return 0;
}
void base32_encode(const unsigned char* buffer, size_t len, char* result) {
unsigned char word = 0;
size_t n = 0;
size_t i = 0;
size_t index = 0;
for (; i < len;) {
if (index > 3) {
word = (buffer[i] & (0xFF >> index));
index = (index + 5) % 8;
word <<= index;
if (i < len - 1)
word |= buffer[i + 1] >> (8 - index);
i++;
} else {
word = (buffer[i] >> (8 - (index + 5))) & 0x1F;
index = (index + 5) % 8;
if (index == 0) i++;
}
result[n++] = BASE32_ALPHABET[word];
}
result[n] = '\0';
}
void base32_decode(const char* src, unsigned char* dst, size_t len) {
size_t index = 0;
size_t offset = 0;
size_t i = 0;
memset(dst, 0, len);
for (i = 0; src[i]; i++) {
unsigned char n = 0;
for (; n < 32; n++) if (src[i] == BASE32_ALPHABET[n]) break;
if (n == 32) continue;
if (index <= 3) {
index = (index + 5) % 8;
if (index == 0) {
dst[offset++] |= n;
if (offset == len) break;
} else {
dst[offset] |= n << (8 - index);
}
} else {
index = (index + 5) % 8;
dst[offset++] |= (n >> index);
if (offset == len) break;
dst[offset] |= n << (8 - index);
}
}
}
int string_split(const char* string, const char* split, void* data, string_split_handler_t handler)
{
char* buf = strdup(string);
char* start;
char* pos;
int count = 0;
start = buf;
while ((pos = strstr(start, split)))
{
pos[0] = '\0';
start = strip_white_space(start);
if (*start)
{
if (handler(start, count, data) < 0)
{
hub_free(buf);
return -1;
}
}
start = &pos[1];
count++;
}
start = strip_white_space(start);
if (*start)
{
if (handler(start, count, data) < 0)
{
hub_free(buf);
return -1;
}
}
hub_free(buf);
return count+1;
}
struct file_read_line_data
{
file_line_handler_t handler;
void* data;
};
static int file_read_line_handler(char* line, int count, void* ptr)
{
struct file_read_line_data* data = (struct file_read_line_data*) ptr;
LOG_DUMP("Line: %s", line);
if (data->handler(line, count+1, data->data) < 0)
return -1;
return 0;
}
int file_read_lines(const char* file, void* data, file_line_handler_t handler)
{
int fd;
ssize_t ret;
char buf[MAX_RECV_BUF];
struct file_read_line_data split_data;
memset(buf, 0, MAX_RECV_BUF);
LOG_TRACE("Opening file %s for line reading.", file);
fd = open(file, 0);
if (fd == -1)
{
LOG_ERROR("Unable to open file %s: %s", file, strerror(errno));
return -2;
}
ret = read(fd, buf, MAX_RECV_BUF-1);
close(fd);
if (ret < 0)
{
LOG_ERROR("Unable to read from file %s: %s", file, strerror(errno));
return -1;
}
else if (ret == 0)
{
LOG_WARN("File is empty.");
return 0;
}
buf[ret] = 0;
/* Parse configuration */
split_data.handler = handler;
split_data.data = data;
return string_split(buf, "\n", &split_data, file_read_line_handler);
}
int uhub_atoi(const char* value) {
int len = strlen(value);
int offset = 0;
int val = 0;
int i = 0;
for (; i < len; i++)
if (value[i] > '9' || value[i] < '0')
offset++;
for (i = offset; i< len; i++)
val = val*10 + (value[i] - '0');
return value[0] == '-' ? -val : val;
}
int is_number(const char* value, int* num)
{
int len = strlen(value);
int offset = (value[0] == '-') ? 1 : 0;
int val = 0;
int i = offset;
if (!*(value + offset))
return 0;
for (; i < len; i++)
if (value[i] > '9' || value[i] < '0')
return 0;
for (i = offset; i< len; i++)
val = val*10 + (value[i] - '0');
*num = value[0] == '-' ? -val : val;
return 1;
}
const char* format_size(size_t bytes, char* buf, size_t bufsize)
{
static const char* quant[] = { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB", "ZiB", "YiB" };
size_t b = bytes;
size_t factor = 0;
size_t divisor = 1;
while (b > 1024)
{
factor++;
b = (b >> 10);
divisor = (divisor << 10);
}
uhub_assert(factor < (sizeof(quant) / sizeof(const char*)));
if (factor >= 2)
snprintf(buf, bufsize, "%.1f %s", (double) bytes / (double) divisor, quant[factor]);
else
snprintf(buf, bufsize, PRINTF_SIZE_T " %s", bytes / divisor, quant[factor]);
return buf;
}
const char* uhub_itoa(int val)
{
static char buf[22];
return snprintf(buf, sizeof(buf), "%d", val) < 0 ? NULL : buf;
}
const char* uhub_ulltoa(uint64_t val)
{
static char buf[22];
return snprintf(buf, sizeof(buf), PRINTF_UINT64_T, val) < 0 ? NULL : buf;
}
#ifndef HAVE_STRNDUP
char* strndup(const char* string, size_t n)
{
size_t max = MIN(strlen(string), n);
char* tmp = hub_malloc(max+1);
memcpy(tmp, string, max);
tmp[max] = 0;
return tmp;
}
#endif
#ifndef HAVE_MEMMEM
void* memmem(const void *haystack, size_t haystacklen, const void *needle, size_t needlelen)
{
char* c_buf = (char*) haystack;
char* c_pat = (char*) needle;
char* ptr = memchr(c_buf, c_pat[0], haystacklen);
while (ptr && ((size_t) (&ptr[0] - &c_buf[0]) < haystacklen))
{
if (!memcmp(ptr, c_pat, needlelen))
return ptr;
ptr = memchr(&ptr[1], c_pat[0], &c_buf[haystacklen] - &ptr[0]);
}
return 0;
}
#endif
int split_string(const char* string, const char* split, struct linked_list* list, int allow_empty)
{
char* tmp1, *tmp2;
int n = 0;
if (!string || !*string || !split || !*split || !list)
return -1;
for (;;)
{
tmp1 = strstr(string, split);
if (tmp1) tmp2 = hub_strndup(string, tmp1 - string);
else tmp2 = hub_strdup(string);
if (!tmp2)
{
list_clear(list, &hub_free);
return -1;
}
if (*tmp2 || allow_empty)
{
/* store in list */
list_append(list, tmp2);
n++;
}
else
{
/* ignore element */
hub_free(tmp2);
}
if (!tmp1) break; /* last element found */
string = tmp1;
string += strlen(split);
}
return n;
}
const char* get_timestamp(time_t now)
{
static char ts[32] = {0, };
struct tm* t = localtime(&now);
snprintf(ts, sizeof(ts), "[%02d:%02d]", t->tm_hour, t->tm_min);
return ts;
}
void strip_off_ini_line_comments(char* line, int line_count)
{
char* p = line;
char* out = line;
int backslash = 0;
if (!*line)
return;
for (; *p; p++)
{
if (!backslash)
{
if (*p == '\\')
{
backslash = 1;
}
else if (*p == '#')
{
*out = '\0';
out++;
break;
}
else
{
*out = *p;
out++;
}
}
else
{
if (*p == '\\' || *p == '#' || *p == '\"')
{
*out = *p;
out++;
}
else
{
LOG_WARN("Invalid backslash escape on line %d", line_count);
*out = *p;
out++;
}
backslash = 0;
}
}
*out = '\0';
}
int string_to_boolean(const char* str, int* boolean)
{
if (!str || !*str || !boolean)
return 0;
switch (strlen(str))
{
case 1:
if (str[0] == '1') { *boolean = 1; return 1; }
else if (str[0] == '0') { *boolean = 0; return 1; }
return 0;
case 2:
if (!strcasecmp(str, "on")) { *boolean = 1; return 1; }
if (!strcasecmp(str, "no")) { *boolean = 0; return 1; }
return 0;
case 3:
if (!strcasecmp(str, "yes")) { *boolean = 1; return 1; }
if (!strcasecmp(str, "off")) { *boolean = 0; return 1; }
return 0;
case 4:
if (!strcasecmp(str, "true")) { *boolean = 1; return 1; }
return 0;
case 5:
if (!strcasecmp(str, "false")) { *boolean = 0; return 1; }
return 0;
default:
return 0;
}
}
char* strip_off_quotes(char* line)
{
size_t len;
if (!*line)
return line;
len = strlen(line);
if (len < 2)
return line;
if ((line[0] == '"' && line[len - 1] == '"') ||
(line[0] == '\'' && line[len - 1] == '\''))
{
line[len-1] = '\0';
return line + 1;
}
return line;
}
You can’t perform that action at this time.