summaryrefslogtreecommitdiff
path: root/libs/libcurl/src/base64.c
diff options
context:
space:
mode:
authorGeorge Hazan <ghazan@miranda.im>2017-12-06 21:44:09 +0300
committerGeorge Hazan <ghazan@miranda.im>2017-12-06 21:44:17 +0300
commit4b6980f68d25901133519bc1ad1c5376819a3876 (patch)
tree0d919622bfc2659f34a7bed303fefb99ecab052a /libs/libcurl/src/base64.c
parent0112d2767268037cf63e44c4464cf9eed237d06d (diff)
libcurl: update to 7.57
Diffstat (limited to 'libs/libcurl/src/base64.c')
-rw-r--r--libs/libcurl/src/base64.c240
1 files changed, 156 insertions, 84 deletions
diff --git a/libs/libcurl/src/base64.c b/libs/libcurl/src/base64.c
index 3f3f0f9b8b..204a2273d1 100644
--- a/libs/libcurl/src/base64.c
+++ b/libs/libcurl/src/base64.c
@@ -5,11 +5,11 @@
* | (__| |_| | _ <| |___
* \___|\___/|_| \_\_____|
*
- * Copyright (C) 1998 - 2012, Daniel Stenberg, <daniel@haxx.se>, et al.
+ * Copyright (C) 1998 - 2016, Daniel Stenberg, <daniel@haxx.se>, et al.
*
* This software is licensed as described in the file COPYING, which
* you should have received as part of this distribution. The terms
- * are also available at http://curl.haxx.se/docs/copyright.html.
+ * are also available at https://curl.haxx.se/docs/copyright.html.
*
* You may opt to use, copy, modify, merge, publish, distribute and/or sell
* copies of the Software, and permit persons to whom the Software is
@@ -23,46 +23,64 @@
/* Base64 encoding/decoding */
#include "curl_setup.h"
-
-#define _MPRINTF_REPLACE /* use our functions only */
-#include <curl/mprintf.h>
-
-#include "urldata.h" /* for the SessionHandle definition */
+#include "urldata.h" /* for the Curl_easy definition */
#include "warnless.h"
#include "curl_base64.h"
-#include "curl_memory.h"
#include "non-ascii.h"
-/* include memdebug.h last */
+/* The last 3 #include files should be in this order */
+#include "curl_printf.h"
+#include "curl_memory.h"
#include "memdebug.h"
/* ---- Base64 Encoding/Decoding Table --- */
-static const char table64[]=
+static const char base64[]=
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
-static void decodeQuantum(unsigned char *dest, const char *src)
+/* The Base 64 encoding with an URL and filename safe alphabet, RFC 4648
+ section 5 */
+static const char base64url[]=
+ "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_";
+
+static size_t decodeQuantum(unsigned char *dest, const char *src)
{
+ size_t padding = 0;
const char *s, *p;
- unsigned long i, v, x = 0;
+ unsigned long i, x = 0;
for(i = 0, s = src; i < 4; i++, s++) {
- v = 0;
- p = table64;
- while(*p && (*p != *s)) {
- v++;
- p++;
- }
- if(*p == *s)
- x = (x << 6) + v;
- else if(*s == '=')
+ unsigned long v = 0;
+
+ if(*s == '=') {
x = (x << 6);
+ padding++;
+ }
+ else {
+ p = base64;
+
+ while(*p && (*p != *s)) {
+ v++;
+ p++;
+ }
+
+ if(*p == *s)
+ x = (x << 6) + v;
+ else
+ return 0;
+ }
}
- dest[2] = curlx_ultouc(x & 0xFFUL);
+ if(padding < 1)
+ dest[2] = curlx_ultouc(x & 0xFFUL);
+
x >>= 8;
- dest[1] = curlx_ultouc(x & 0xFFUL);
+ if(padding < 2)
+ dest[1] = curlx_ultouc(x & 0xFFUL);
+
x >>= 8;
dest[0] = curlx_ultouc(x & 0xFFUL);
+
+ return 3 - padding;
}
/*
@@ -82,84 +100,80 @@ static void decodeQuantum(unsigned char *dest, const char *src)
CURLcode Curl_base64_decode(const char *src,
unsigned char **outptr, size_t *outlen)
{
+ size_t srclen = 0;
size_t length = 0;
- size_t equalsTerm = 0;
+ size_t padding = 0;
size_t i;
size_t numQuantums;
- unsigned char lastQuantum[3];
size_t rawlen = 0;
+ unsigned char *pos;
unsigned char *newstr;
*outptr = NULL;
*outlen = 0;
+ srclen = strlen(src);
+
+ /* Check the length of the input string is valid */
+ if(!srclen || srclen % 4)
+ return CURLE_BAD_CONTENT_ENCODING;
+ /* Find the position of any = padding characters */
while((src[length] != '=') && src[length])
length++;
+
/* A maximum of two = padding characters is allowed */
if(src[length] == '=') {
- equalsTerm++;
- if(src[length+equalsTerm] == '=')
- equalsTerm++;
+ padding++;
+ if(src[length + 1] == '=')
+ padding++;
}
- numQuantums = (length + equalsTerm) / 4;
- /* Don't allocate a buffer if the decoded length is 0 */
- if(numQuantums == 0)
- return CURLE_OK;
+ /* Check the = padding characters weren't part way through the input */
+ if(length + padding != srclen)
+ return CURLE_BAD_CONTENT_ENCODING;
- rawlen = (numQuantums * 3) - equalsTerm;
+ /* Calculate the number of quantums */
+ numQuantums = srclen / 4;
- /* The buffer must be large enough to make room for the last quantum
- (which may be partially thrown out) and the zero terminator. */
- newstr = malloc(rawlen+4);
+ /* Calculate the size of the decoded string */
+ rawlen = (numQuantums * 3) - padding;
+
+ /* Allocate our buffer including room for a zero terminator */
+ newstr = malloc(rawlen + 1);
if(!newstr)
return CURLE_OUT_OF_MEMORY;
- *outptr = newstr;
+ pos = newstr;
- /* Decode all but the last quantum (which may not decode to a
- multiple of 3 bytes) */
- for(i = 0; i < numQuantums - 1; i++) {
- decodeQuantum(newstr, src);
- newstr += 3; src += 4;
- }
+ /* Decode the quantums */
+ for(i = 0; i < numQuantums; i++) {
+ size_t result = decodeQuantum(pos, src);
+ if(!result) {
+ free(newstr);
+
+ return CURLE_BAD_CONTENT_ENCODING;
+ }
- /* This final decode may actually read slightly past the end of the buffer
- if the input string is missing pad bytes. This will almost always be
- harmless. */
- decodeQuantum(lastQuantum, src);
- for(i = 0; i < 3 - equalsTerm; i++)
- newstr[i] = lastQuantum[i];
+ pos += result;
+ src += 4;
+ }
- newstr[i] = '\0'; /* zero terminate */
+ /* Zero terminate */
+ *pos = '\0';
- *outlen = rawlen; /* return size of decoded data */
+ /* Return the decoded data */
+ *outptr = newstr;
+ *outlen = rawlen;
return CURLE_OK;
}
-/*
- * Curl_base64_encode()
- *
- * Given a pointer to an input buffer and an input size, encode it and
- * return a pointer in *outptr to a newly allocated memory area holding
- * encoded data. Size of encoded data is returned in variable pointed by
- * outlen.
- *
- * Input length of 0 indicates input buffer holds a NUL-terminated string.
- *
- * Returns CURLE_OK on success, otherwise specific error code. Function
- * output shall not be considered valid unless CURLE_OK is returned.
- *
- * When encoded data length is 0, returns NULL in *outptr.
- *
- * @unittest: 1302
- */
-CURLcode Curl_base64_encode(struct SessionHandle *data,
- const char *inputbuff, size_t insize,
- char **outptr, size_t *outlen)
+static CURLcode base64_encode(const char *table64,
+ struct Curl_easy *data,
+ const char *inputbuff, size_t insize,
+ char **outptr, size_t *outlen)
{
- CURLcode error;
+ CURLcode result;
unsigned char ibuf[3];
unsigned char obuf[4];
int i;
@@ -173,11 +187,16 @@ CURLcode Curl_base64_encode(struct SessionHandle *data,
*outptr = NULL;
*outlen = 0;
- if(0 == insize)
+ if(!insize)
insize = strlen(indata);
- base64data = output = malloc(insize*4/3+4);
- if(NULL == output)
+#if SIZEOF_SIZE_T == 4
+ if(insize > UINT_MAX/4)
+ return CURLE_OUT_OF_MEMORY;
+#endif
+
+ base64data = output = malloc(insize * 4 / 3 + 4);
+ if(!output)
return CURLE_OUT_OF_MEMORY;
/*
@@ -185,10 +204,10 @@ CURLcode Curl_base64_encode(struct SessionHandle *data,
* not the host encoding. And we can't change the actual input
* so we copy it to a buffer, translate it, and use that instead.
*/
- error = Curl_convert_clone(data, indata, insize, &convbuf);
- if(error) {
+ result = Curl_convert_clone(data, indata, insize, &convbuf);
+ if(result) {
free(output);
- return error;
+ return result;
}
if(convbuf)
@@ -219,30 +238,83 @@ CURLcode Curl_base64_encode(struct SessionHandle *data,
table64[obuf[0]],
table64[obuf[1]]);
break;
+
case 2: /* two bytes read */
snprintf(output, 5, "%c%c%c=",
table64[obuf[0]],
table64[obuf[1]],
table64[obuf[2]]);
break;
+
default:
snprintf(output, 5, "%c%c%c%c",
table64[obuf[0]],
table64[obuf[1]],
table64[obuf[2]],
- table64[obuf[3]] );
+ table64[obuf[3]]);
break;
}
output += 4;
}
+
+ /* Zero terminate */
*output = '\0';
- *outptr = base64data; /* return pointer to new data, allocated memory */
- if(convbuf)
- free(convbuf);
+ /* Return the pointer to the new data (allocated memory) */
+ *outptr = base64data;
+
+ free(convbuf);
- *outlen = strlen(base64data); /* return the length of the new data */
+ /* Return the length of the new data */
+ *outlen = strlen(base64data);
return CURLE_OK;
}
-/* ---- End of Base64 Encoding ---- */
+
+/*
+ * Curl_base64_encode()
+ *
+ * Given a pointer to an input buffer and an input size, encode it and
+ * return a pointer in *outptr to a newly allocated memory area holding
+ * encoded data. Size of encoded data is returned in variable pointed by
+ * outlen.
+ *
+ * Input length of 0 indicates input buffer holds a NUL-terminated string.
+ *
+ * Returns CURLE_OK on success, otherwise specific error code. Function
+ * output shall not be considered valid unless CURLE_OK is returned.
+ *
+ * When encoded data length is 0, returns NULL in *outptr.
+ *
+ * @unittest: 1302
+ */
+CURLcode Curl_base64_encode(struct Curl_easy *data,
+ const char *inputbuff, size_t insize,
+ char **outptr, size_t *outlen)
+{
+ return base64_encode(base64, data, inputbuff, insize, outptr, outlen);
+}
+
+/*
+ * Curl_base64url_encode()
+ *
+ * Given a pointer to an input buffer and an input size, encode it and
+ * return a pointer in *outptr to a newly allocated memory area holding
+ * encoded data. Size of encoded data is returned in variable pointed by
+ * outlen.
+ *
+ * Input length of 0 indicates input buffer holds a NUL-terminated string.
+ *
+ * Returns CURLE_OK on success, otherwise specific error code. Function
+ * output shall not be considered valid unless CURLE_OK is returned.
+ *
+ * When encoded data length is 0, returns NULL in *outptr.
+ *
+ * @unittest: 1302
+ */
+CURLcode Curl_base64url_encode(struct Curl_easy *data,
+ const char *inputbuff, size_t insize,
+ char **outptr, size_t *outlen)
+{
+ return base64_encode(base64url, data, inputbuff, insize, outptr, outlen);
+}