← Commit list

48e196426b

Daniel Stenberg

2026-09-03

SUCCESS Duration: 402s

Subject

escape: polish encoding and decoding for speed

Binary

No binary (build failed)

Architecture & Compiler Flags

Architecture: x86_64-linux-gnu Compiler: gcc (Ubuntu 11.4.0-1ubuntu1~22.04.3) 11.4.0

—

Build Log

(log yok)

Commit Diff GitHub →

lib/escape.c +113 -30 tests/data/test558 +1 -2 tests/libtest/lib1396.c +4 -0

lib/escape.c+113 -30

@@ -44,13 +44,55 @@ char curl_unescape(const char string, int length)
	return curl_easy_unescape(NULL, string, length, NULL);
	}

+	#define NOPE 0xff
+	#define _OK_ 0x1f /* octet not needing encoding but not a hexadecimal
+	character */
+
+	static const unsigned char hextable[256] = {
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, /* 00 */
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, /* 08 */
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, /* 10 */
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, /* 18 */
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, /* 20 */
+	NOPE, NOPE, NOPE, NOPE, NOPE, _OK_, _OK_, NOPE, /* 28 */
+	0, 1, 2, 3, 4, 5, 6, 7, /* 30 */
+	8, 9, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, /* 38 */
+	NOPE, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, _OK_, /* 40 */
+	_OK_, _OK_, _OK_, _OK_, _OK_, _OK_, _OK_, _OK_, /* 48 */
+	_OK_, _OK_, _OK_, _OK_, _OK_, _OK_, _OK_, _OK_, /* 50 */
+	_OK_, _OK_, _OK_, NOPE, NOPE, NOPE, NOPE, _OK_, /* 58 */
+	NOPE, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, _OK_, /* 60 */
+	_OK_, _OK_, _OK_, _OK_, _OK_, _OK_, _OK_, _OK_, /* 68 */
+	_OK_, _OK_, _OK_, _OK_, _OK_, _OK_, _OK_, _OK_, /* 70 */
+	_OK_, _OK_, _OK_, NOPE, NOPE, NOPE, _OK_, NOPE, /* 78 */
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, /* 80 - not ASCII */
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE,
+	NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE, NOPE
+	};
+
	/* Escapes for URL the given unescaped string of given length.
	* 'data' is ignored since 7.82.0.
	*/
	char curl_easy_escape(CURL curl, const char *string, int length)
	{
	size_t len;
-	struct dynbuf d;
+	const char *f = string;
+	size_t nappend = 0;
+	char *target;
+	char *encoded;
	(void)curl;

	if(!string \|\| (length < 0))
@@ -63,27 +105,37 @@ char curl_easy_escape(CURL curl, const char *string, int length)
	if(len > SIZE_MAX / 16)
	return NULL;

-	curlx_dyn_init(&d, (len * 3) + 1);
+	encoded = target = curlx_malloc((len * 3) + 1);
+	if(!target)
+	return NULL;

	while(len--) {
-	/* treat the characters unsigned */
-	unsigned char in = (unsigned char)*string++;
+	uint8_t in = (uint8_t)*string++;

-	if(ISUNRESERVED(in)) {
+	if(!(hextable[in] & 0x80))
	/* append this */
-	if(curlx_dyn_addn(&d, &in, 1))
-	return NULL;
-	}
+	nappend++;
	else {
+	if(nappend) {
+	memcpy(target, f, nappend);
+	target += nappend;
+	nappend = 0;
+	}
+	f = string;
	/* encode it */
-	unsigned char out[3] = { '%' };
-	Curl_hexbyte(&out[1], in);
-	if(curlx_dyn_addn(&d, out, 3))
-	return NULL;
+	target[0] = '%';
+	target[1] = Curl_udigits[in >> 4];
+	target[2] = Curl_udigits[in & 0x0F];
+	target += 3;
	}
	}
+	if(nappend) {
+	memcpy(target, f, nappend);
+	target += nappend;
+	}
+	target = '\0'; / null terminate */

-	return curlx_dyn_ptr(&d);
+	return encoded;
	}

	/*
@@ -108,6 +160,7 @@ CURLcode Curl_urldecode(const char *string, size_t length,
	{
	size_t alloc;
	char *ns;
+	uint8_t reject_limit;

	DEBUGASSERT(string);
	DEBUGASSERT(ctrl >= REJECT_NADA); /* crash on TRUE/FALSE */
@@ -121,36 +174,66 @@ CURLcode Curl_urldecode(const char *string, size_t length,
	/* store output string */
	*ostring = ns;

+	reject_limit = (ctrl == REJECT_CTRL) ? 0x20 :
+	(ctrl == REJECT_ZERO) ? 1 : 0;
+
	while(alloc) {
-	unsigned char in = (unsigned char)*string;
-	if(('%' == in) && (alloc > 2) &&
-	ISXDIGIT(string[1]) && ISXDIGIT(string[2])) {
-	/* this is two hexadecimal digits following a '%' */
-	in = (unsigned char)((curlx_hexval(string[1]) << 4) \|
-	curlx_hexval(string[2]));
-	string += 3;
-	alloc -= 3;
-	}
-	else {
+	uint8_t in;
+
+	if(*string == '%') {
+	if(alloc > 2) {
+	uint8_t h1 = hextable[(uint8_t)string[1]];
+	uint8_t h2 = hextable[(uint8_t)string[2]];
+	if(!((h1 \| h2) & 0xf0)) {
+	in = (uint8_t)((h1 << 4) \| h2);
+	string += 3;
+	alloc -= 3;
+	if(in < reject_limit)
+	goto error;
+	*ns++ = (char)in;
+	continue;
+	}
+	}
+	in = '%';
	string++;
	alloc--;
+	*ns++ = (char)in;
+	continue;
	}
+	else {
+	const char *p = memchr(string + 1, '%', alloc - 1);
+	size_t n = p ? (size_t)(p - string) : alloc;

-	if(((ctrl == REJECT_CTRL) && (in < 0x20)) \|\|
-	((ctrl == REJECT_ZERO) && (in == 0))) {
-	curlx_safefree(*ostring);
-	return CURLE_URL_MALFORMAT;
-	}
+	if(reject_limit) {
+	if(reject_limit == 1) {
+	if(memchr(string, 0, n))
+	goto error;
+	}
+	else {
+	size_t i;
+	for(i = 0; i < n; i++) {
+	if((uint8_t)string[i] < 0x20)
+	goto error;
+	}
+	}
+	}

-	*ns++ = (char)in;
+	memcpy(ns, string, n);
+	ns += n;
+	string += n;
+	alloc -= n;
+	}
	}
	ns = 0; / terminate it */

	if(olen)
	/* store output size */
-	olen = ns - ostring;
+	olen = (size_t)(ns - ostring);

	return CURLE_OK;
+	error:
+	curlx_safefree(*ostring);
+	return CURLE_URL_MALFORMAT;
	}

	/*

tests/data/test558+1 -2

@@ -32,8 +32,7 @@ libtest memory tracking operational
	MEM easy.c: malloc()
	MEM lib%TESTNUMBER.c: malloc()
	MEM lib%TESTNUMBER.c: free()
-	MEM dynbuf.c: realloc()
-	MEM dynbuf.c: realloc()
+	MEM escape.c: malloc()
	MEM escape.c: free()
	MEM easy.c: free()
	</file>

tests/libtest/lib1396.c+4 -0

@@ -69,6 +69,10 @@ static CURLcode test_lib1396(const char *arg)
	/* escape, this => that */
	static const struct test list2[] = {
	{ "a", 1, "a", 1 },
+	{ "-", 1, "-", 1 },
+	{ ".", 1, ".", 1 },
+	{ "_", 1, "_", 1 },
+	{ "~", 1, "~", 1 },
	{ "/", 1, "%2F", 3 },
	{ "a=b", 3, "a%3Db", 5 },
	{ "a=b", 0, "a%3Db", 5 },