/src/curl/lib/urlapi.c

Source
/***************************************************************************
 *                                  _   _ ____  _
 *  Project                     ___| | | |  _ \| |
 *                             / __| | | | |_) | |
 *                            | (__| |_| |  _ <| |___
 *                             \___|\___/|_| \_\_____|
 *
 * Copyright (C) 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 https://curl.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
 * furnished to do so, under the terms of the COPYING file.
 *
 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
 * KIND, either express or implied.
 *
 * SPDX-License-Identifier: curl
 *
 ***************************************************************************/
#include "curl_setup.h"

#include "urldata.h"
#include "urlapi-int.h"
#include "strcase.h"
#include "url.h"
#include "escape.h"
#include "curlx/inet_pton.h"
#include "curlx/inet_ntop.h"
#include "curlx/strdup.h"
#include "idn.h"
#include "curlx/strparse.h"
#include "curl_memrchr.h"

#ifdef _WIN32
/* MS-DOS/Windows style drive prefix, eg c: in c:foo */
#define STARTS_WITH_DRIVE_PREFIX(str)        \
  ((('a' <= (str)[0] && (str)[0] <= 'z') ||  \
    ('A' <= (str)[0] && (str)[0] <= 'Z')) && \
   ((str)[1] == ':'))
#endif

/* MS-DOS/Windows style drive prefix, optionally with
 * a '|' instead of ':', followed by a slash or NUL */
#define STARTS_WITH_URL_DRIVE_PREFIX(str)                  \
  ((('a' <= (str)[0] && (str)[0] <= 'z') ||                \
    ('A' <= (str)[0] && (str)[0] <= 'Z')) &&               \
   ((str)[1] == ':' || (str)[1] == '|') &&                 \
   ((str)[2] == '/' || (str)[2] == '\\' || (str)[2] == 0))

/* scheme is not URL encoded, the longest libcurl supported ones are... */
#define MAX_SCHEME_LEN 40
#define MAX_ZONEID_LEN 16

/*
 * If USE_IPV6 is disabled, we still want to parse IPv6 addresses, so make
 * sure we have _some_ value for AF_INET6 without polluting our fake value
 * everywhere.
 */
#if !defined(USE_IPV6) && !defined(AF_INET6)
#define AF_INET6 (AF_INET + 1)
#endif

#define DEFAULT_SCHEME "https"

static void free_urlhandle(struct Curl_URL *u)
{
  curlx_free(u->scheme);
  curlx_free(u->user);
  curlx_strzero(u->password);
  curlx_free(u->password);
  curlx_free(u->options);
  curlx_free(u->host);
  curlx_free(u->zoneid);
  curlx_free(u->path);
  curlx_free(u->query);
  curlx_free(u->fragment);
}

/*
 * Find the separator at the end of the hostname, or the '?' in cases like
 * http://www.example.com?id=2380
 */
static const char *find_host_sep(const char *url)
{
  /* Find the start of the hostname */
  const char *sep = strstr(url, "//");
  if(!sep)
    sep = url;
  else
    sep += 2;

  /* Find first / or ? */
  while(*sep && *sep != '/' && *sep != '?')
    sep++;

  return sep;
}

/* convert CURLcode to CURLUcode */
#define cc2cu(x) \
  ((x) == CURLE_TOO_LARGE ? CURLUE_TOO_LARGE : CURLUE_OUT_OF_MEMORY)

/* urlencode_str() writes data into an output dynbuf and URL-encodes the
 * spaces in the source URL accordingly.
 *
 * This function re-encodes the string, meaning that it leaves already encoded
 * bytes as-is and works by encoding only what *has* to be encoded - unless it
 * has to uppercase the hex to normalize.
 *
 * Illegal percent-encoding sequences are left as-is.
 *
 * URL encoding should be skipped for hostnames, otherwise IDN resolution
 * will fail.
 *
 * 'query' tells if it is a query part or not, or if it is allowed to
 * "transition" into a query part with a question mark.
 *
 * @unittest 1675
 */
UNITTEST CURLUcode urlencode_str(struct dynbuf *o, const char *url,
                                 size_t len, bool relative,
                                 unsigned int query);
UNITTEST CURLUcode urlencode_str(struct dynbuf *o, const char *url,
                                 size_t len, bool relative,
                                 unsigned int query)
{
  /* we must add this with whitespace-replacing */
  const unsigned char *iptr;
  const unsigned char *host_sep = (const unsigned char *)url;
  CURLcode result = CURLE_OK;

  DEBUGASSERT((query >= QUERY_NO) && (query <= QUERY_YES));

  if(!relative) {
    size_t n;
    host_sep = (const unsigned char *)find_host_sep(url);

    /* output the first piece as-is */
    n = (const char *)host_sep - url;
    result = curlx_dyn_addn(o, url, n);
    len -= n;
  }

  for(iptr = host_sep; len && !result;) {
    if(*iptr == ' ') {
      if(query != QUERY_YES)
        result = curlx_dyn_addn(o, "%20", 3);
      else
        result = curlx_dyn_addn(o, "+", 1);
      iptr++;
      len--;
    }
    else if((*iptr < ' ') || (*iptr >= 0x7f)) {
      unsigned char out[3] = { '%' };
      Curl_hexbyte(&out[1], *iptr);
      result = curlx_dyn_addn(o, out, 3);
      iptr++;
      len--;
    }
    else if(*iptr == '%' && (len >= 3) &&
            ISXDIGIT(iptr[1]) && ISXDIGIT(iptr[2]) &&
            (ISLOWER(iptr[1]) || ISLOWER(iptr[2]))) {
      /* uppercase it */
      unsigned char hex = (unsigned char)((curlx_hexval(iptr[1]) << 4) |
                                          curlx_hexval(iptr[2]));
      unsigned char out[3] = { '%' };
      Curl_hexbyte(&out[1], hex);
      result = curlx_dyn_addn(o, out, 3);
      iptr += 3;
      len -= 3;
    }
    else {
      const unsigned char *start = iptr;
      while(len) {
        if(*iptr == ' ' || *iptr < ' ' || *iptr >= 0x7f)
          break;
        if(*iptr == '%' && (len >= 3) &&
           ISXDIGIT(iptr[1]) && ISXDIGIT(iptr[2]) &&
           (ISLOWER(iptr[1]) || ISLOWER(iptr[2])))
          break;
        if(*iptr == '?') {
          if(query == QUERY_NOT_YET) {
            iptr++;
            len--;
            query = QUERY_YES;
            break;
          }
        }
        iptr++;
        len--;
      }
      result = curlx_dyn_addn(o, (const char *)start, (size_t)(iptr - start));
    }
  }

  if(result)
    return cc2cu(result);
  return CURLUE_OK;
}

/*
 * Returns the length of the scheme if the given URL is absolute (as opposed
 * to relative). Stores the scheme in the buffer if TRUE and 'buf' is
 * non-NULL. The buflen must be larger than MAX_SCHEME_LEN if buf is set.
 *
 * If 'guess_scheme' is TRUE, it means the URL might be provided without
 * scheme.
 */
size_t Curl_is_absolute_url(const char *url, char *buf, size_t buflen,
                            bool guess_scheme)
{
  size_t i = 0;
  DEBUGASSERT(!buf || (buflen > MAX_SCHEME_LEN));
  (void)buflen; /* only used in debug-builds */
  if(buf)
    buf[0] = 0; /* always leave a defined value in buf */
#ifdef _WIN32
  if(guess_scheme && STARTS_WITH_DRIVE_PREFIX(url))
    return 0;
#endif
  if(ISALPHA(url[0])) {
    if(buf)
      buf[0] = Curl_raw_tolower(url[0]);
    for(i = 1; i < MAX_SCHEME_LEN; ++i) {
      char s = url[i];
      if(s && (ISALNUM(s) || (s == '+') || (s == '-') || (s == '.'))) {
        if(buf)
          buf[i] = Curl_raw_tolower(s);
      }
      else {
        break;
      }
    }
  }
  if(i && (url[i] == ':') && ((url[i + 1] == '/') || !guess_scheme)) {
    /* If this does not guess scheme, the scheme always ends with the colon so
       that this also detects data: URLs etc. In guessing mode, data: could
       be the hostname "data" with a specified port number. */

    /* the length of the scheme is the name part only */
    size_t len = i;
    if(buf)
      buf[i] = 0;
    return len;
  }
  if(buf)
    buf[0] = 0;
  return 0;
}

/* scan for byte values <= 31, 127 and maybe space */
static bool badoctets(const char *input, size_t n, int flags)
{
  const uint8_t *p = (const unsigned char *)input;
  const uint8_t control = flags & CURLU_ALLOW_SPACE ? 0x1f : 0x20;
  while(n--) {
    if(*p <= control || *p == 127)
      return TRUE;
    p++;
  }
  return FALSE;
}

/*
 * parse_hostname_login()
 *
 * Parse the login details (username, password and options) from the URL and
 * strip them out of the hostname
 *
 * @unittest 1675
 */
UNITTEST CURLUcode parse_hostname_login(struct Curl_URL *u,
                                        const char *login,
                                        size_t len,
                                        unsigned int flags,
                                        size_t *hostname_offset);
UNITTEST CURLUcode parse_hostname_login(struct Curl_URL *u,
                                        const char *login,
                                        size_t len,
                                        unsigned int flags,
                                        size_t *hostname_offset)
{
  CURLUcode ures = CURLUE_OK;
  CURLcode result;
  char *userp = NULL;
  char *passwdp = NULL;
  char *optionsp = NULL;
  const struct Curl_scheme *h = NULL;

  /* At this point, we assume all the other special cases have been taken
   * care of, so the host is at most
   *
   *   [user[:password][;options]]@]hostname
   *
   * We need somewhere to put the embedded details, so do that first.
   */
  const char *ptr;

  DEBUGASSERT(login);

  *hostname_offset = 0;
  ptr = memchr(login, '@', len);
  if(!ptr)
    goto out;

  /* We will now try to extract the
   * possible login information in a string like:
   * ftp://user:password@ftp.site.example:8021/README */
  ptr++;

  /* if this is a known scheme, get some details */
  if(u->scheme)
    h = Curl_get_scheme(u->scheme);

  /* We could use the login information in the URL so extract it. Only parse
     options if the handler says we should. Note that 'h' might be NULL! */
  result = Curl_parse_login_details(login, ptr - login - 1,
                                    &userp, &passwdp,
                                    (h && (h->flags & PROTOPT_URLOPTIONS)) ?
                                    &optionsp : NULL);
  if(result) {
    /* the only possible error from Curl_parse_login_details is out of
       memory: */
    ures = CURLUE_OUT_OF_MEMORY;
    goto out;
  }

  if(userp) {
    if(flags & CURLU_DISALLOW_USER) {
      /* Option DISALLOW_USER is set and URL contains username. */
      ures = CURLUE_USER_NOT_ALLOWED;
      goto out;
    }
    curlx_free(u->user);
    u->user = userp;
  }

  if(passwdp) {
    curlx_strzero(u->password);
    curlx_free(u->password);
    u->password = passwdp;
  }

  if(optionsp) {
    curlx_free(u->options);
    u->options = optionsp;
  }

  if(userp && badoctets(userp, strlen(userp), flags))
    ures = CURLUE_BAD_USER;
  else if(passwdp && badoctets(passwdp, strlen(passwdp), flags))
    ures = CURLUE_BAD_PASSWORD;
  else if(optionsp && badoctets(optionsp, strlen(optionsp), flags))
    ures = CURLUE_MALFORMED_INPUT;

  userp = passwdp = optionsp = NULL;

  if(!ures) {
    /* the hostname starts at this offset */
    *hostname_offset = ptr - login;
    return CURLUE_OK;
  }

out:

  curlx_free(userp);
  curlx_strzero(passwdp);
  curlx_free(passwdp);
  curlx_free(optionsp);
  curlx_safefree(u->user);
  curlx_strzero(u->password);
  curlx_safefree(u->password);
  curlx_safefree(u->options);

  return ures;
}

/* @unittest 1653 */
UNITTEST CURLUcode parse_port(struct Curl_URL *u, struct dynbuf *host,
                              bool has_scheme);
UNITTEST CURLUcode parse_port(struct Curl_URL *u, struct dynbuf *host,
                              bool has_scheme)
{
  const char *portptr;
  const char *hostname = curlx_dyn_ptr(host);
  /*
   * Find the end of an IPv6 address on the ']' ending bracket.
   */
  u->portnum = 0;
  u->port_present = FALSE;
  if(hostname[0] == '[') {
    portptr = memchr(hostname + 1, ']', curlx_dyn_len(host) - 1);
    if(!portptr)
      return CURLUE_BAD_IPV6;
    portptr++;
    /* this is a RFC2732-style specified IP-address */
    if(*portptr) {
      if(*portptr != ':')
        return CURLUE_BAD_PORT_NUMBER;
    }
    else
      portptr = NULL;
  }
  else
    portptr = memchr(hostname, ':', curlx_dyn_len(host));

  if(portptr) {
    curl_off_t port;
    size_t keep = portptr - hostname;
    int rc;

    /* Browser behavior adaptation. If there is a colon with no digits after,
       cut off the name there which makes us ignore the colon and use the
       default port. Firefox, Chrome and Safari all do that.

       Do not do it if the URL has no scheme, to make something that looks like
       a scheme not work! */
    curlx_dyn_setlen(host, keep);
    portptr++;
    if(!*portptr)
      return has_scheme ? CURLUE_OK : CURLUE_BAD_PORT_NUMBER;
    if(*portptr == '\\')
      return CURLUE_BACKSLASH;
    rc = curlx_str_number(&portptr, &port, 0xffff);
    if(rc)
      return CURLUE_BAD_PORT_NUMBER;
    else if(*portptr == '\\')
      return CURLUE_BACKSLASH;
    else if(*portptr)
      return CURLUE_BAD_PORT_NUMBER;

    u->portnum = (uint16_t)port;
    u->port_present = TRUE;
  }

  return CURLUE_OK;
}

/* This function assumes 'hostname' now starts with [. It trims 'hostname' in
 * place and it sets u->zoneid if present.
 *
 * @unittest 1675
 */
UNITTEST CURLUcode ipv6_parse(struct Curl_URL *u, char *hostname,
                              size_t hlen);
UNITTEST CURLUcode ipv6_parse(struct Curl_URL *u, char *hostname,
                              size_t hlen) /* length of hostname */
{
  size_t len;
  DEBUGASSERT(*hostname == '[');
  if(hlen < 4) /* '[::]' is the shortest possible valid string */
    return CURLUE_BAD_IPV6;
  hostname++;
  hlen -= 2;

  /* only valid IPv6 letters are ok */
  len = strspn(hostname, "0123456789abcdefABCDEF:.");

  if(hlen != len) {
    hlen = len;
    if(hostname[len] == '%') {
      /* this could now be '%[zone id]' */
      char zoneid[MAX_ZONEID_LEN];
      int i = 0;
      char *h = &hostname[len + 1];
      /* pass '25' if present and is a URL encoded percent sign */
      if(!strncmp(h, "25", 2) && h[2] && (h[2] != ']'))
        h += 2;
      while(*h && (*h != ']') && (i < (MAX_ZONEID_LEN - 1)) &&
            (*h != ' '))
        zoneid[i++] = *h++;
      if(!i || (']' != *h))
        return CURLUE_BAD_IPV6;
      zoneid[i] = 0;
      u->zoneid = curlx_strdup(zoneid);
      if(!u->zoneid)
        return CURLUE_OUT_OF_MEMORY;
      hostname[len] = ']'; /* insert end bracket */
      hostname[len + 1] = 0; /* terminate the hostname */
    }
    else
      return CURLUE_BAD_IPV6;
    /* hostname is fine */
  }

  /* Normalize the IPv6 address */
  {
    char dest[16]; /* fits a binary IPv6 address */
    hostname[hlen] = 0; /* end the address there */
    if(curlx_inet_pton(AF_INET6, hostname, dest) != 1)
      return CURLUE_BAD_IPV6;
    if(!curlx_inet_ntop(AF_INET6, dest, hostname, hlen + 1)) {
      hlen = strlen(hostname); /* might be shorter now */
      hostname[hlen + 1] = 0;
    }
    hostname[hlen] = ']'; /* restore ending bracket */
  }
  return CURLUE_OK;
}

/* characters not allowed in hostnames:
   " \r\n\t/:#?!@{}[]\\$\'\"^`*<>=;,+&()%|" */

static const bool invalid_host_char[256] = {
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 0x00-0x0F */
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 0x10-0x1F */
  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, /* 0x20-0x2F */
  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, /* 0x30-0x3F */
  1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x40-0x4F */
  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, /* 0x50-0x5F */
  1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x60-0x6F */
  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 1  /* 0x70-0x7F */
};

/* the input is a confirmed hostname, never an IPv6 address */
static CURLUcode hostname_check(char *hostname, size_t hlen)
{
  size_t i;
  for(i = 0; i < hlen; i++) {
    if(invalid_host_char[(unsigned char)hostname[i]])
      return CURLUE_BAD_HOSTNAME;
  }
  if((hlen >= 2) &&
     (hostname[hlen - 1] == '.') && (hostname[hlen - 2] == '.'))
    /* more than one trailing dot is not allowed */
    return CURLUE_BAD_HOSTNAME;
  else if((hlen == 1) && (hostname[0] == '.'))
    /* a single dot alone is not allowed */
    return CURLUE_BAD_HOSTNAME;
  return CURLUE_OK;
}

/* the input is a hostname or perhaps an IPv6 address */
static CURLUcode hostname_check6(struct Curl_URL *u, char *hostname,
                                size_t hlen) /* length of hostname */
{
  DEBUGASSERT(hostname);

  if(!hlen)
    return CURLUE_NO_HOST;
  else if(hostname[0] == '[')
    return ipv6_parse(u, hostname, hlen);

  return hostname_check(hostname, hlen);
}

/*
 * Handle partial IPv4 numerical addresses and different bases, like
 * '16843009', '0x7f', '0x7f.1' '0177.1.1.1' etc.
 *
 * If the given input string is syntactically wrong IPv4 or any part for
 * example is too big, this function returns HOST_NAME.
 *
 * Output the "normalized" version of that input string in plain quad decimal
 * integers.
 *
 * A single dot following the numerical address is accepted and "swallowed" as
 * if it was never there.
 *
 * Returns the host type.
 *
 * @unittest 1675
 */
UNITTEST int ipv4_normalize(struct dynbuf *host);
UNITTEST int ipv4_normalize(struct dynbuf *host)
{
  bool done = FALSE;
  int n = 0;
  const char *c = curlx_dyn_ptr(host);
  unsigned int parts[4] = { 0, 0, 0, 0 };
  CURLcode result = CURLE_OK;

  if(!ISDIGIT(*c))
    return HOST_NAME;

  while(!done) {
    int rc;
    curl_off_t l;
    if(*c == '0') {
      if((c[1] | 0x20) == 'x') {
        c += 2; /* skip the prefix */
        rc = curlx_str_hex(&c, &l, UINT_MAX);
        if(rc)
          return HOST_NAME;
      }
      else
        rc = curlx_str_octal(&c, &l, UINT_MAX);
    }
    else
      rc = curlx_str_number(&c, &l, UINT_MAX);

    if(rc) {
      if(!n || (rc != STRE_NO_NUM) || *c)
        return HOST_NAME;
      n--;
    }
    else
      parts[n] = (unsigned int)l;

    switch(*c) {
    case '.':
      if(n == 3) {
        if(c[1])
          /* something follows this dot */
          return HOST_NAME;
        done = TRUE;
      }
      else {
        n++;
        c++;
      }
      break;

    case '\0':
      done = TRUE;
      break;

    default:
      return HOST_NAME;
    }
  }

  switch(n) {
  case 0: /* a -- 32 bits */
    curlx_dyn_reset(host);

    result = curlx_dyn_addf(host, "%u.%u.%u.%u",
                            (parts[0] >> 24),
                            ((parts[0] >> 16) & 0xff),
                            ((parts[0] >> 8) & 0xff),
                            (parts[0] & 0xff));
    break;
  case 1: /* a.b -- 8.24 bits */
    if((parts[0] > 0xff) || (parts[1] > 0xffffff))
      return HOST_NAME;
    curlx_dyn_reset(host);
    result = curlx_dyn_addf(host, "%u.%u.%u.%u",
                            parts[0],
                            ((parts[1] >> 16) & 0xff),
                            ((parts[1] >> 8) & 0xff),
                            (parts[1] & 0xff));
    break;
  case 2: /* a.b.c -- 8.8.16 bits */
    if((parts[0] > 0xff) || (parts[1] > 0xff) || (parts[2] > 0xffff))
      return HOST_NAME;
    curlx_dyn_reset(host);
    result = curlx_dyn_addf(host, "%u.%u.%u.%u",
                            parts[0],
                            parts[1],
                            ((parts[2] >> 8) & 0xff),
                            (parts[2] & 0xff));
    break;
  case 3: /* a.b.c.d -- 8.8.8.8 bits */
    if((parts[0] > 0xff) || (parts[1] > 0xff) || (parts[2] > 0xff) ||
       (parts[3] > 0xff))
      return HOST_NAME;
    curlx_dyn_reset(host);
    result = curlx_dyn_addf(host, "%u.%u.%u.%u",
                            parts[0],
                            parts[1],
                            parts[2],
                            parts[3]);
    break;
  }
  if(result)
    return HOST_ERROR;
  return HOST_IPV4;
}

/* if necessary, replace the host content with a URL decoded version */
static CURLUcode urldecode_host(struct dynbuf *host)
{
  const char *per;
  const char *hostname = curlx_dyn_ptr(host);
  per = memchr(hostname, '%', curlx_dyn_len(host));
  if(!per)
    /* nothing to decode */
    return CURLUE_OK;
  else {
    /* encoded */
    size_t dlen;
    char *decoded;
    CURLcode result = Curl_urldecode(hostname, 0, &decoded, &dlen,
                                     REJECT_CTRL);
    if(result)
      return CURLUE_BAD_HOSTNAME;
    curlx_dyn_reset(host);
    result = curlx_dyn_addn(host, decoded, dlen);
    curlx_free(decoded);
    if(result)
      return cc2cu(result);
  }

  return CURLUE_OK;
}

static CURLUcode parse_authority(struct Curl_URL *u,
                                 const char *auth, size_t authlen,
                                 unsigned int flags,
                                 struct dynbuf *host,
                                 bool has_scheme)
{
  size_t offset;
  CURLUcode uc;
  CURLcode result;

  /*
   * Parse the login details and strip them out of the hostname.
   */
  uc = parse_hostname_login(u, auth, authlen, flags, &offset);
  if(uc)
    return uc;

  result = curlx_dyn_addn(host, auth + offset, authlen - offset);
  if(result) {
    uc = cc2cu(result);
    return uc;
  }

  /* parse_port() also sets the hostname length correctly */
  uc = parse_port(u, host, has_scheme);

  if(!curlx_dyn_len(host))
    /* this makes no-host errors override port number problems */
    uc = CURLUE_NO_HOST;
  if(!uc)
    uc = urldecode_host(host);
  if(uc)
    ;
  else if(auth[offset] == '[')
    uc = ipv6_parse(u, curlx_dyn_ptr(host), curlx_dyn_len(host));
  else {
    /* ipv4_normalize() returns *NAME, *IPV4 or *ERROR */
    int type = ipv4_normalize(host);

    if(type == HOST_NAME)
      uc = hostname_check(curlx_dyn_ptr(host), curlx_dyn_len(host));
    else if(type == HOST_ERROR)
      uc = CURLUE_OUT_OF_MEMORY;
  }

  return uc;
}

/* used for HTTP/2 server push */
CURLUcode Curl_url_set_authority(CURLU *u, const char *authority)
{
  CURLUcode ures;
  struct dynbuf host;

  DEBUGASSERT(authority);
  curlx_dyn_init(&host, CURL_MAX_INPUT_LENGTH);

  ures = parse_authority(u, authority, strlen(authority),
                         CURLU_DISALLOW_USER, &host, !!u->scheme);
  if(ures)
    curlx_dyn_free(&host);
  else {
    curlx_free(u->host);
    u->host = curlx_dyn_ptr(&host);
  }
  return ures;
}

/*
 * "Remove Dot Segments"
 * https://datatracker.ietf.org/doc/html/rfc3986#section-5.2.4
 */

static bool is_dot(const char **str, size_t *clen)
{
  const char *p = *str;
  if(*p == '.') {
    (*str)++;
    (*clen)--;
    return TRUE;
  }
  else if((*clen >= 3) &&
          (p[0] == '%') && (p[1] == '2') && ((p[2] | 0x20) == 'e')) {
    *str += 3;
    *clen -= 3;
    return TRUE;
  }
  return FALSE;
}

#define ISSLASH(x) ((x) == '/')

/* prescan the string to see if it needs work */
static bool needs_dedotdot(const char *p, size_t pn)
{
  /* a single byte path cannot be cleaned up */
  if(pn < 2)
    return FALSE;
  if(!memchr(p, '.', pn) && !memchr(p, '%', pn))
    return FALSE;
  while(pn) {
    if(is_dot(&p, &pn)) {
      /* "./" or dot before end of string */
      if(!pn || ISSLASH(*p))
        return TRUE;
      /* "../" or ".." before end of string */
      else if(is_dot(&p, &pn) && (!pn || ISSLASH(*p)))
        return TRUE;
    }
    else {
      p++;
      pn--;
    }
  }
  return FALSE;
}

/*
 * dedotdotify()
 *
 * This function gets a null-terminated path with dot and dotdot sequences
 * passed in and strips them off according to the rules in RFC 3986 section
 * 5.2.4.
 *
 * The function handles a path. It should not contain the query nor fragment.
 *
 * RETURNS
 *
 * Zero for success and 'out' set to an allocated string (or NULL if there's
 * nothing to do).
 *
 * @unittest 1395
 */
UNITTEST int dedotdotify(const char *input, size_t clen, char **outp);
UNITTEST int dedotdotify(const char *input, size_t clen, char **outp)
{
  struct dynbuf out;
  CURLcode result = CURLE_OK;

  /* variables for leading dot checks */
  const char *dinput = input;
  size_t dlen = clen;

  *outp = NULL;
  if(!needs_dedotdot(input, clen))
    return 0;

  curlx_dyn_init(&out, clen + 1);

  /* if the input buffer begins with a prefix of "../" or "./", then remove
     that prefix from the input buffer; otherwise, */
  if(is_dot(&dinput, &dlen)) {
    if(ISSLASH(*dinput)) {
      /* one dot followed by a slash */
      input = dinput + 1;
      clen = dlen - 1;
    }

    /* if the input buffer consists only of "." or "..", then remove
       that from the input buffer; otherwise, */
    else if(is_dot(&dinput, &dlen)) {
      if(!dlen)
        /* .. [end] */
        goto end;
      else if(ISSLASH(*dinput)) {
        /* ../ */
        input = dinput + 1;
        clen = dlen - 1;
      }
    }
  }

  while(clen && !result) { /* until end of path content */
    if(ISSLASH(*input)) {
      const char *p = &input[1];
      size_t blen = clen - 1;
      /* if the input buffer begins with a prefix of "/./" or "/.", where "."
         is a complete path segment, then replace that prefix with "/" in the
         input buffer; otherwise, */
      if(is_dot(&p, &blen)) {
        if(!blen) { /* /. */
          result = curlx_dyn_addn(&out, "/", 1);
          break;
        }
        else if(ISSLASH(*p)) { /* /./ */
          input = p;
          clen = blen;
          continue;
        }

        /* if the input buffer begins with a prefix of "/../" or "/..", where
           ".." is a complete path segment, then replace that prefix with "/"
           in the input buffer and remove the last segment and its preceding
           "/" (if any) from the output buffer; otherwise, */
        else if(is_dot(&p, &blen) && (ISSLASH(*p) || !blen)) {
          /* remove the last segment from the output buffer */
          size_t len = curlx_dyn_len(&out);
          if(len) {
            const char *ptr = curlx_dyn_ptr(&out);
            const char *last = memrchr(ptr, '/', len);
            if(last)
              /* trim the output at the slash */
              curlx_dyn_setlen(&out, last - ptr);
          }

          if(blen) { /* /../ */
            input = p;
            clen = blen;
            continue;
          }
          result = curlx_dyn_addn(&out, "/", 1);
          break;
        }
      }
    }

    /* move the first path segment in the input buffer to the end of the
       output buffer, including the initial "/" character (if any) and any
       subsequent characters up to, but not including, the next "/" character
       or the end of the input buffer. */

    result = curlx_dyn_addn(&out, input, 1);
    input++;
    clen--;
  }
end:
  if(!result) {
    if(curlx_dyn_len(&out))
      *outp = curlx_dyn_ptr(&out);
    else {
      *outp = curlx_strdup("");
      if(!*outp)
        return 1;
    }
  }
  return result ? 1 : 0; /* success */
}

/*
 * @unittest 1675
 */
UNITTEST CURLUcode parse_file(const char *url, size_t urllen, CURLU *u,
                              const char **pathp, size_t *pathlenp);
UNITTEST CURLUcode parse_file(const char *url, size_t urllen, CURLU *u,
                              const char **pathp, size_t *pathlenp)
{
  const char *path;
  size_t pathlen;

  *pathp = NULL;
  *pathlenp = 0;
  if(urllen <= 6)
    /* file:/ is not enough to actually be a complete file: URL */
    return CURLUE_BAD_FILE_URL;

  /* path has been allocated large enough to hold this */
  path = &url[5];
  pathlen = urllen - 5;

  /* RFC 8089: file-hier-part = ( "//" auth-path ) / local-path, where
     local-path also starts with a "/". So reject anything that does not
     start with at least one "/" */
  if(path[0] != '/')
    return CURLUE_BAD_FILE_URL;

  /* Extra handling URLs with an authority component (i.e. that start with
   * "file://")
   *
   * We allow omitted hostname (e.g. file:/<path>) -- valid according to
   * RFC 8089, but not the (current) WHAT-WG URL spec.
   */
  if(path[1] == '/') {
    /* swallow the two slashes */
    const char *ptr = &path[2];

    /*
     * According to RFC 8089, a file: URL can be reliably dereferenced if:
     *
     *  o it has no/blank hostname, or
     *
     *  o the hostname matches "localhost" (case-insensitively), or
     *
     *  o the hostname is a FQDN that resolves to this machine, or
     *
     * For brevity, we only consider URLs with empty, "localhost", or
     * "127.0.0.1" hostnames as local, otherwise as an UNC String.
     *
     * Additionally, there is an exception for URLs with a Windows drive
     * letter in the authority (which was accidentally omitted from RFC 8089
     * Appendix E, but believe me, it was meant to be there. --MK)
     */
    if(ptr[0] != '/' && !STARTS_WITH_URL_DRIVE_PREFIX(ptr)) {
      /* the URL includes a hostname, it must match "localhost" or
         "127.0.0.1" to be valid */
      if(checkprefix("localhost/", ptr) ||
         checkprefix("127.0.0.1/", ptr)) {
        ptr += 9; /* now points to the slash after the host */
      }
      else
        /* Invalid file://hostname/, expected localhost or 127.0.0.1 or
           none */
        return CURLUE_BAD_FILE_URL;
    }

    path = ptr;
    pathlen = urllen - (ptr - url);
  }

#if !defined(_WIN32) && !defined(MSDOS) && !defined(__CYGWIN__)
  /* Do not allow Windows drive letters when not in Windows.
   * This catches both "file:/c:" and "file:c:" */
  if(('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) ||
     STARTS_WITH_URL_DRIVE_PREFIX(path)) {
    /* File drive letters are only accepted in MS-DOS/Windows */
    return CURLUE_BAD_FILE_URL;
  }
#else
  /* If the path starts with a slash and a drive letter, ditch the slash */
  if('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) {
    /* This cannot be done with strcpy, as the memory chunks overlap! */
    path++;
    pathlen--;
  }
#endif
  u->scheme = curlx_strdup("file");
  if(!u->scheme)
    return CURLUE_OUT_OF_MEMORY;

  *pathp = path;
  *pathlenp = pathlen;
  return CURLUE_OK;
}

static CURLUcode parse_scheme(const char *url, CURLU *u, char *schemebuf,
                              size_t schemelen, unsigned int flags,
                              const char **hostpp)
{
  /* clear path */
  const char *schemep = NULL;

  if(schemelen) {
    int num_slashes = 0;
    const char *p = &url[schemelen + 1];
    if(!Curl_getn_scheme(schemebuf, schemelen) &&
       !(flags & CURLU_NON_SUPPORT_SCHEME))
      return CURLUE_UNSUPPORTED_SCHEME;

    if(!ISSLASH(*p))
      /* less than one */
      return CURLUE_BAD_SLASHES;
    if((flags & CURLU_NO_AUTHORITY)) {
      while(ISSLASH(*p) && (num_slashes < 2)) {
        p++;
        num_slashes++;
      }
    }
    else {
      while(ISSLASH(*p) && (num_slashes < 4)) {
        p++;
        num_slashes++;
      }
      if(num_slashes > 3)
        return CURLUE_BAD_SLASHES;
    }

    schemep = schemebuf;
    *hostpp = p; /* hostname starts here */
  }
  else {
    /* no scheme! */

    if(!(flags & (CURLU_DEFAULT_SCHEME | CURLU_GUESS_SCHEME)))
      return CURLUE_BAD_SCHEME;

    if(flags & CURLU_DEFAULT_SCHEME)
      schemep = DEFAULT_SCHEME;

    /*
     * The URL was badly formatted, let's try without scheme specified.
     */
    *hostpp = url;
  }

  if(schemep) {
    u->scheme = curlx_strdup(schemep);
    if(!u->scheme)
      return CURLUE_OUT_OF_MEMORY;
  }
  return CURLUE_OK;
}

static CURLUcode guess_scheme(CURLU *u, struct dynbuf *host)
{
  const char *hostname = curlx_dyn_ptr(host);
  const char *schemep = NULL;
  /* legacy curl-style guess based on hostname */
  if(checkprefix("ftp.", hostname))
    schemep = "ftp";
  else if(checkprefix("dict.", hostname))
    schemep = "dict";
  else if(checkprefix("ldap.", hostname))
    schemep = "ldap";
  else if(checkprefix("imap.", hostname))
    schemep = "imap";
  else if(checkprefix("smtp.", hostname))
    schemep = "smtp";
  else if(checkprefix("pop3.", hostname))
    schemep = "pop3";
  else
    schemep = "http";

  u->scheme = curlx_strdup(schemep);
  if(!u->scheme)
    return CURLUE_OUT_OF_MEMORY;

  u->guessed_scheme = TRUE;
  return CURLUE_OK;
}

static CURLUcode handle_fragment(CURLU *u, const char *fragment,
                                 size_t fraglen, unsigned int flags)
{
  CURLUcode ures;
  u->fragment_present = TRUE;
  if(fraglen > 1) {
    /* skip the leading '#' in the copy but include the null-terminator */
    if(flags & CURLU_URLENCODE) {
      struct dynbuf enc;
      curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH);
      ures = urlencode_str(&enc, fragment + 1, fraglen - 1, TRUE, QUERY_NO);
      if(ures)
        return ures;
      u->fragment = curlx_dyn_ptr(&enc);
    }
    else {
      if(badoctets(fragment, fraglen, flags))
        return CURLUE_BAD_FRAGMENT;
      u->fragment = curlx_memdup0(fragment + 1, fraglen - 1);
      if(!u->fragment)
        return CURLUE_OUT_OF_MEMORY;
    }
  }
  return CURLUE_OK;
}

static CURLUcode handle_query(CURLU *u, const char *query,
                              size_t qlen, unsigned int flags)
{
  u->query_present = TRUE;
  if(qlen > 1) {
    if(flags & CURLU_URLENCODE) {
      struct dynbuf enc;
      CURLUcode ures;
      curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH);
      /* skip the leading question mark */
      ures = urlencode_str(&enc, query + 1, qlen - 1, TRUE, QUERY_YES);
      if(ures)
        return ures;
      u->query = curlx_dyn_ptr(&enc);
    }
    else {
      if(badoctets(query, qlen, flags))
        return CURLUE_BAD_QUERY;

      u->query = curlx_memdup0(query + 1, qlen - 1);
      if(!u->query)
        return CURLUE_OUT_OF_MEMORY;
    }
  }
  else {
    /* single byte query */
    u->query = curlx_strdup("");
    if(!u->query)
      return CURLUE_OUT_OF_MEMORY;
  }
  return CURLUE_OK;
}

static CURLUcode handle_path(CURLU *u, const char *path,
                             size_t pathlen, unsigned int flags,
                             bool is_file)
{
  CURLUcode ures;
  if(pathlen && (flags & CURLU_URLENCODE)) {
    struct dynbuf enc;
    curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH);
    ures = urlencode_str(&enc, path, pathlen, TRUE, QUERY_NO);
    if(ures)
      return ures;
    pathlen = curlx_dyn_len(&enc);
    path = u->path = curlx_dyn_ptr(&enc);
  }

  if(pathlen >= (size_t)(1 + !is_file)) {
    if(badoctets(path, pathlen, flags))
      return CURLUE_BAD_PATH;

    /* paths for file:// scheme can be one byte, others need to be two */
    if(!u->path) {
      u->path = curlx_memdup0(path, pathlen);
      if(!u->path)
        return CURLUE_OUT_OF_MEMORY;
      path = u->path;
    }
    else if(flags & CURLU_URLENCODE)
      /* it might have encoded more than the path so cut it */
      u->path[pathlen] = 0;

    if(!(flags & CURLU_PATH_AS_IS)) {
      /* remove ../ and ./ sequences according to RFC3986 */
      char *dedot;
      int err = dedotdotify(path, pathlen, &dedot);
      if(err)
        return CURLUE_OUT_OF_MEMORY;
      if(dedot) {
        curlx_free(u->path);
        u->path = dedot;
      }
    }
  }
  return CURLUE_OK;
}

static CURLUcode parseurl(const char *url, CURLU *u, unsigned int flags)
{
  const char *path;
  size_t pathlen;
  char schemebuf[MAX_SCHEME_LEN + 1];
  size_t schemelen = 0;
  size_t urllen;
  CURLUcode ures = CURLUE_OK;
  struct dynbuf host;
  bool is_file = FALSE;

  DEBUGASSERT(url);

  urllen = strlen(url);
  if(urllen > CURL_MAX_INPUT_LENGTH)
    return CURLUE_MALFORMED_INPUT;

  curlx_dyn_init(&host, CURL_MAX_INPUT_LENGTH);

  schemelen = Curl_is_absolute_url(url, schemebuf, sizeof(schemebuf),
                                   flags & (CURLU_GUESS_SCHEME |
                                            CURLU_DEFAULT_SCHEME));

  /* handle the file: scheme */
  if(schemelen == 4 && !memcmp(schemebuf, "file", 4)) {
    is_file = TRUE;
    ures = parse_file(url, urllen, u, &path, &pathlen);
  }
  else {
    const char *hostp = NULL;
    const char *p;
    size_t hostlen;
    ures = parse_scheme(url, u, schemebuf, schemelen, flags, &hostp);
    if(ures)
      goto fail;

    /* find the end of the hostname + port number */
    p = hostp;
    while(*p && *p != '/' && *p != '?' && *p != '#')
      p++;
    hostlen = p - hostp;
    path = p;

    /* this pathlen also contains the query and the fragment */
    pathlen = urllen - (path - url);
    if(hostlen) {
      ures = parse_authority(u, hostp, hostlen, flags, &host, !!u->scheme);
      if(!ures && (flags & CURLU_GUESS_SCHEME) && !u->scheme)
        ures = guess_scheme(u, &host);
    }
    else if(flags & CURLU_NO_AUTHORITY) {
      /* allowed to be empty. */
      if(curlx_dyn_add(&host, ""))
        ures = CURLUE_OUT_OF_MEMORY;
    }
    else
      ures = CURLUE_NO_HOST;
  }
  if(!ures) {
    /* The path might at this point contain a fragment and/or a query to
       handle */
    const char *fragment = memchr(path, '#', pathlen);
    if(fragment) {
      size_t fraglen = pathlen - (fragment - path);
      ures = handle_fragment(u, fragment, fraglen, flags);
      /* after this, pathlen still contains the query */
      pathlen -= fraglen;
    }
  }
  if(!ures) {
    const char *query = memchr(path, '?', pathlen);
    if(query) {
      size_t qlen = pathlen - (query - path);
      ures = handle_query(u, query, qlen, flags);
      pathlen -= qlen;
    }
  }
  if(!ures)
    /* the fragment and query parts are trimmed off from the path */
    ures = handle_path(u, path, pathlen, flags, is_file);
  if(!ures) {
    u->host = curlx_dyn_ptr(&host);
    return CURLUE_OK;
  }
fail:
  curlx_dyn_free(&host);
  free_urlhandle(u);
  return ures;
}

/*
 * Parse the URL and, if successful, replace everything in the Curl_URL struct.
 */
static CURLUcode parseurl_and_replace(const char *url, CURLU *u,
                                      unsigned int flags)
{
  CURLUcode ures;
  CURLU tmpurl;
  memset(&tmpurl, 0, sizeof(tmpurl));
  ures = parseurl(url, &tmpurl, flags);
  if(!ures) {
    free_urlhandle(u);
    *u = tmpurl;
  }
  return ures;
}

/*
 * Concatenate a relative URL onto a base URL making it absolute.
 */
static CURLUcode redirect_url(const char *base, const char *relurl,
                              CURLU *u, unsigned int flags)
{
  struct dynbuf urlbuf;
  bool host_changed = FALSE;
  const char *useurl = relurl;
  const char *cutoff = NULL;
  size_t prelen;
  CURLUcode uc;
  /* this can get here with a NULL u->scheme only if asked to use the default
     scheme, so allow fallback to that */
  const char *scheme = u->scheme ? u->scheme : DEFAULT_SCHEME;

  /* protsep points to the start of the hostname, after [scheme]:// */
  const char *protsep = base + strlen(scheme) + 3;
  DEBUGASSERT(base && relurl && u); /* all set here */
  if(!base)
    return CURLUE_MALFORMED_INPUT; /* should never happen */

  /* handle different relative URL types */
  switch(relurl[0]) {
  case '/':
    if(relurl[1] == '/') {
      /* protocol-relative URL: //example.com/path */
      cutoff = protsep;
      useurl = &relurl[2];
      host_changed = TRUE;
    }
    else
      /* absolute /path */
      cutoff = strchr(protsep, '/');
    break;

  case '#':
    /* fragment-only change */
    if(u->fragment_present)
      cutoff = strchr(protsep, '#');
    break;

  default:
    /* path or query-only change */
    if(u->query_present)
      /* remove existing query */
      cutoff = strchr(protsep, '?');
    else if(u->fragment_present)
      /* Remove existing fragment */
      cutoff = strchr(protsep, '#');

    if(relurl[0] != '?') {
      /* append a relative path after the last slash */
      cutoff = memrchr(protsep, '/',
                       cutoff ? (size_t)(cutoff - protsep) : strlen(protsep));
      if(cutoff)
        cutoff++; /* truncate after last slash */
    }
    break;
  }

  prelen = cutoff ? (size_t)(cutoff - base) : strlen(base);

  /* build new URL */
  curlx_dyn_init(&urlbuf, CURL_MAX_INPUT_LENGTH);

  if(!curlx_dyn_addn(&urlbuf, base, prelen) &&
     !urlencode_str(&urlbuf, useurl, strlen(useurl), !host_changed,
                    QUERY_NOT_YET)) {
    uc = parseurl_and_replace(curlx_dyn_ptr(&urlbuf), u,
                              flags & ~U_CURLU_PATH_AS_IS);
  }
  else
    uc = CURLUE_OUT_OF_MEMORY;

  curlx_dyn_free(&urlbuf);
  return uc;
}

/*
 */
CURLU *curl_url(void)
{
  return curlx_calloc(1, sizeof(struct Curl_URL));
}

void curl_url_cleanup(CURLU *u)
{
  if(u) {
    free_urlhandle(u);
    curlx_free(u);
  }
}

#define DUP(dest, src, name)                    \
  do {                                          \
    if((src)->name) {                           \
      (dest)->name = curlx_strdup((src)->name); \
      if(!(dest)->name)                         \
        goto fail;                              \
    }                                           \
  } while(0)

CURLU *curl_url_dup(const CURLU *in)
{
  struct Curl_URL *u = curlx_calloc(1, sizeof(struct Curl_URL));
  if(u) {
    DUP(u, in, scheme);
    DUP(u, in, user);
    DUP(u, in, password);
    DUP(u, in, options);
    DUP(u, in, host);
    DUP(u, in, path);
    DUP(u, in, query);
    DUP(u, in, fragment);
    DUP(u, in, zoneid);
    u->portnum = in->portnum;
    u->port_present = in->port_present;
    u->fragment_present = in->fragment_present;
    u->query_present = in->query_present;
  }
  return u;
fail:
  curl_url_cleanup(u);
  return NULL;
}

#ifndef USE_IDN
#define host_decode(x, y) CURLUE_LACKS_IDN
#define host_encode(x, y) CURLUE_LACKS_IDN
#else
static CURLUcode host_decode(const char *host, char **allochost)
{
  CURLcode result = Curl_idn_decode(host, allochost);
  if(result)
    return (result == CURLE_OUT_OF_MEMORY) ?
      CURLUE_OUT_OF_MEMORY : CURLUE_BAD_HOSTNAME;
  return CURLUE_OK;
}

static CURLUcode host_encode(const char *host, char **allochost)
{
  CURLcode result = Curl_idn_encode(host, allochost);
  if(result)
    return (result == CURLE_OUT_OF_MEMORY) ?
      CURLUE_OUT_OF_MEMORY : CURLUE_BAD_HOSTNAME;
  return CURLUE_OK;
}
#endif

static CURLUcode urlget_format(const CURLU *u, CURLUPart what,
                               const char *ptr, char **partp,
                               bool plusdecode, unsigned int flags)
{
  CURLUcode uc = CURLUE_OK;
  size_t partlen = strlen(ptr);
  bool urldecode = (flags & CURLU_URLDECODE) ? 1 : 0;
  bool urlencode = (flags & CURLU_URLENCODE) ? 1 : 0;
  bool punycode = (flags & CURLU_PUNYCODE) && (what == CURLUPART_HOST);
  bool depunyfy = (flags & CURLU_PUNY2IDN) && (what == CURLUPART_HOST);
  char *part = curlx_memdup0(ptr, partlen);
  *partp = NULL;
  if(!part)
    return CURLUE_OUT_OF_MEMORY;
  if(plusdecode) {
    /* convert + to space */
    char *plus = part;
    size_t i = 0;
    for(i = 0; i < partlen; ++plus, i++) {
      if(*plus == '+')
        *plus = ' ';
    }
  }
  if(urldecode) {
    char *decoded;
    size_t dlen;
    /* this unconditional rejection of control bytes is documented API
       behavior */
    CURLcode result = Curl_urldecode(part, partlen, &decoded, &dlen,
                                     REJECT_CTRL);
    curlx_free(part);
    if(result)
      return CURLUE_URLDECODE;
    part = decoded;
    partlen = dlen;
  }
  if(urlencode) {
    struct dynbuf enc;
    curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH);
    uc = urlencode_str(&enc, part, partlen, TRUE, what == CURLUPART_QUERY ?
                       QUERY_YES : QUERY_NO);
    curlx_free(part);
    if(uc)
      return uc;
    part = curlx_dyn_ptr(&enc);
  }
  else if(punycode) {
    if(!Curl_is_ASCII_name(u->host)) {
      char *punyversion = NULL;
      uc = host_decode(part, &punyversion);
      curlx_free(part);
      if(uc)
        return uc;
      part = punyversion;
    }
  }
  else if(depunyfy && Curl_is_ASCII_name(u->host)) {
    char *unpunified = NULL;
    uc = host_encode(part, &unpunified);
    curlx_free(part);
    if(uc)
      return uc;
    part = unpunified;
  }
  *partp = part;
  return CURLUE_OK;
}

static CURLUcode file_url(const CURLU *u, char **part,
                          const char *fragmentsep,
                          const char *querysep)
{
  char *url = curl_maprintf("file://%s%s%s%s%s",
                            u->path, querysep, u->query ? u->query : "",
                            fragmentsep, u->fragment ? u->fragment : "");
  if(!url)
    return CURLUE_OUT_OF_MEMORY;

  *part = url;
  return CURLUE_OK;
}

static CURLUcode urlget_url(const CURLU *u, char **part, unsigned int flags)
{
  char *url;
  char *allochost = NULL;
  const char *fragmentsep =
    (u->fragment || (u->fragment_present && flags & CURLU_GET_EMPTY)) ?
    "#" : "";
  const char *querysep = ((u->query && u->query[0]) ||
                          (u->query_present && flags & CURLU_GET_EMPTY)) ?
    "?" : "";
  char portbuf[7];
  if(curl_strequal("file", u->scheme))
    return file_url(u, part, fragmentsep, querysep);
  else if(!u->host)
    return CURLUE_NO_HOST;
  else {
    const char *scheme;
    char *options = u->options;
    char *port = NULL;
    const struct Curl_scheme *h = NULL;
    char schemebuf[MAX_SCHEME_LEN + 5];
    if(u->scheme)
      scheme = u->scheme;
    else if(flags & CURLU_DEFAULT_SCHEME)
      scheme = DEFAULT_SCHEME;
    else
      return CURLUE_NO_SCHEME;

    if(u->port_present) {
      curl_msnprintf(portbuf, sizeof(portbuf), "%u", u->portnum);
      port = portbuf;
    }

    h = Curl_get_scheme(scheme);
    if(h) {
      if(!u->port_present && (flags & CURLU_DEFAULT_PORT)) {
        /* there is no stored port number, but asked to deliver a default one
           for the scheme */
        curl_msnprintf(portbuf, sizeof(portbuf), "%u", h->defport);
        port = portbuf;
      }
      else if(u->port_present && (h->defport == u->portnum) &&
              (flags & CURLU_NO_DEFAULT_PORT)) {
        /* there is a stored port number, but asked to inhibit if it matches
           the default port for the scheme */
        port = NULL;
      }

      if(!(h->flags & PROTOPT_URLOPTIONS))
        options = NULL;
    }

    if(u->host[0] == '[') {
      if(u->zoneid) {
        /* make it '[ host %25 zoneid ]' */
        struct dynbuf enc;
        size_t hostlen = strlen(u->host);
        curlx_dyn_init(&enc, CURL_MAX_INPUT_LENGTH);
        if(curlx_dyn_addf(&enc, "%.*s%%25%s]", (int)hostlen - 1, u->host,
                          u->zoneid))
          return CURLUE_OUT_OF_MEMORY;
        allochost = curlx_dyn_ptr(&enc);
      }
    }
    else if(flags & CURLU_URLENCODE) {
      allochost = curl_easy_escape(NULL, u->host, 0);
      if(!allochost)
        return CURLUE_OUT_OF_MEMORY;
    }
    else if(flags & CURLU_PUNYCODE) {
      if(!Curl_is_ASCII_name(u->host)) {
        CURLUcode ret = host_decode(u->host, &allochost);
        if(ret)
          return ret;
      }
    }
    else if(flags & CURLU_PUNY2IDN) {
      if(Curl_is_ASCII_name(u->host)) {
        CURLUcode ret = host_encode(u->host, &allochost);
        if(ret)
          return ret;
      }
    }

    if(!(flags & CURLU_NO_GUESS_SCHEME) || !u->guessed_scheme)
      curl_msnprintf(schemebuf, sizeof(schemebuf), "%s://", scheme);
    else
      schemebuf[0] = 0;

    url = curl_maprintf("%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s",
                        schemebuf,
                        u->user ? u->user : "",
                        u->password ? ":" : "",
                        u->password ? u->password : "",
                        options ? ";" : "",
                        options ? options : "",
                        (u->user || u->password || options) ? "@" : "",
                        allochost ? allochost : u->host,
                        port ? ":" : "",
                        port ? port : "",
                        u->path ? u->path : "/",
                        querysep,
                        u->query ? u->query : "",
                        fragmentsep,
                        u->fragment ? u->fragment : "");
    curlx_free(allochost);
  }
  if(!url)
    return CURLUE_OUT_OF_MEMORY;
  *part = url;
  return CURLUE_OK;
}

CURLUcode curl_url_get(const CURLU *u, CURLUPart what,
                       char **part, unsigned int flags)
{
  const char *ptr;
  CURLUcode ifmissing = CURLUE_UNKNOWN_PART;
  char portbuf[7];
  bool plusdecode = FALSE;
  if(!u)
    return CURLUE_BAD_HANDLE;
  if(!part)
    return CURLUE_BAD_PARTPOINTER;
  *part = NULL;

  switch(what) {
  case CURLUPART_SCHEME:
    ptr = u->scheme;
    ifmissing = CURLUE_NO_SCHEME;
    flags &= ~U_CURLU_URLDECODE; /* never for schemes */
    if((flags & CURLU_NO_GUESS_SCHEME) && u->guessed_scheme)
      return CURLUE_NO_SCHEME;
    break;
  case CURLUPART_USER:
    ptr = u->user;
    ifmissing = CURLUE_NO_USER;
    break;
  case CURLUPART_PASSWORD:
    ptr = u->password;
    ifmissing = CURLUE_NO_PASSWORD;
    break;
  case CURLUPART_OPTIONS:
    ptr = u->options;
    ifmissing = CURLUE_NO_OPTIONS;
    break;
  case CURLUPART_HOST:
    ptr = u->host;
    ifmissing = CURLUE_NO_HOST;
    break;
  case CURLUPART_ZONEID:
    ptr = u->zoneid;
    ifmissing = CURLUE_NO_ZONEID;
    break;
  case CURLUPART_PORT:
    ptr = NULL;
    ifmissing = CURLUE_NO_PORT;
    flags &= ~U_CURLU_URLDECODE; /* never for port */
    if(u->port_present) {
      const struct Curl_scheme *h = u->scheme ?
                                    Curl_get_scheme(u->scheme) : NULL;
      /* there is a stored port number, but ask to inhibit if
         it matches the default one for the scheme */
      if(h && (h->defport == u->portnum) &&
         (flags & CURLU_NO_DEFAULT_PORT)) {
        ptr = NULL;
      }
      else {
        curl_msnprintf(portbuf, sizeof(portbuf), "%u", u->portnum);
        ptr = portbuf;
      }
    }
    else if((flags & CURLU_DEFAULT_PORT) && u->scheme) {
      /* there is no stored port number, but asked to deliver
         a default one for the scheme */
      const struct Curl_scheme *h = Curl_get_scheme(u->scheme);
      if(h) {
        curl_msnprintf(portbuf, sizeof(portbuf), "%u", h->defport);
        ptr = portbuf;
      }
    }
    break;
  case CURLUPART_PATH:
    ptr = u->path;
    if(!ptr)
      ptr = "/";
    break;
  case CURLUPART_QUERY:
    ptr = u->query;
    ifmissing = CURLUE_NO_QUERY;
    plusdecode = flags & CURLU_URLDECODE;
    if(ptr && !ptr[0] && !(flags & CURLU_GET_EMPTY))
      /* there was a blank query and the user does not ask for it */
      ptr = NULL;
    break;
  case CURLUPART_FRAGMENT:
    ptr = u->fragment;
    ifmissing = CURLUE_NO_FRAGMENT;
    if(!ptr && u->fragment_present && flags & CURLU_GET_EMPTY)
      /* there was a blank fragment and the user asks for it */
      ptr = "";
    break;
  case CURLUPART_URL:
    return urlget_url(u, part, flags);
  default:
    ptr = NULL;
    break;
  }
  if(ptr)
    return urlget_format(u, what, ptr, part, plusdecode, flags);

  return ifmissing;
}

static CURLUcode set_url_scheme(CURLU *u, const char *scheme,
                                unsigned int flags)
{
  size_t plen = strlen(scheme);
  const struct Curl_scheme *h = NULL;
  if((plen > MAX_SCHEME_LEN) || (plen < 1))
    /* too long or too short */
    return CURLUE_BAD_SCHEME;
  /* verify that it is a fine scheme */
  h = Curl_get_scheme(scheme);
  if(!(flags & CURLU_NON_SUPPORT_SCHEME) && (!h || !h->run))
    return CURLUE_UNSUPPORTED_SCHEME;
  if(!h) {
    const char *s = scheme;
    if(ISALPHA(*s)) {
      /* ALPHA *( ALPHA / DIGIT / "+" / "-" / "." ) */
      s++;
      while(--plen) {
        if(ISALNUM(*s) || (*s == '+') || (*s == '-') || (*s == '.'))
          s++; /* fine */
        else
          return CURLUE_BAD_SCHEME;
      }
    }
    else
      return CURLUE_BAD_SCHEME;
  }
  u->guessed_scheme = FALSE;
  return CURLUE_OK;
}

static CURLUcode set_url_port(CURLU *u, const char *provided_port)
{
  curl_off_t port;
  if(!ISDIGIT(provided_port[0]))
    /* not a number */
    return CURLUE_BAD_PORT_NUMBER;
  if(curlx_str_number(&provided_port, &port, 0xffff) || *provided_port)
    /* weirdly provided number, not good! */
    return CURLUE_BAD_PORT_NUMBER;
  u->portnum = (uint16_t)port;
  u->port_present = TRUE;
  return CURLUE_OK;
}

static CURLUcode set_url(CURLU *u, const char *url, size_t part_size,
                         unsigned int flags)
{
  /*
   * Allow a new URL to replace the existing (if any) contents.
   *
   * If the existing contents is enough for a URL, allow a relative URL to
   * replace it.
   */
  CURLUcode uc;
  char *oldurl = NULL;

  if(!part_size) {
    /* a blank URL is not a valid URL unless we already have a complete one
       and this is a redirect */
    uc = curl_url_get(u, CURLUPART_URL, &oldurl, flags);
    if(!uc) {
      /* success, meaning the "" is a fine relative URL, and the new URL
         inherits scheme/authority/path/query, but not fragment, from the
         existing URL (RFC 3986 section 5.2.2) */
      curlx_safefree(u->fragment);
      u->fragment_present = FALSE;
      curlx_free(oldurl);
      return CURLUE_OK;
    }
    if(uc == CURLUE_OUT_OF_MEMORY)
      return uc;
    return CURLUE_MALFORMED_INPUT;
  }

  /* if the new URL is absolute replace the existing with the new. */
  if(Curl_is_absolute_url(url, NULL, 0,
                          flags & (CURLU_GUESS_SCHEME | CURLU_DEFAULT_SCHEME)))
    return parseurl_and_replace(url, u, flags);

  /* if the old URL is incomplete (we cannot get an absolute URL in
     'oldurl'), replace the existing with the new.
     Always include "scheme://" to make the URL "complete" */
  /* Preserve empty query/fragment separators: they affect where relative
     references splice into the base URL. */
  uc = curl_url_get(u, CURLUPART_URL, &oldurl,
                    (flags & ~CURLU_NO_GUESS_SCHEME) | CURLU_GET_EMPTY);
  if(uc == CURLUE_OUT_OF_MEMORY)
    return uc;
  else if(uc)
    return parseurl_and_replace(url, u, flags);

  DEBUGASSERT(oldurl); /* it is set here */
  /* apply the relative part to create a new URL */
  uc = redirect_url(oldurl, url, u, flags);
  curlx_free(oldurl);
  return uc;
}

static CURLUcode urlset_clear(CURLU *u, CURLUPart what)
{
  switch(what) {
  case CURLUPART_URL:
    free_urlhandle(u);
    memset(u, 0, sizeof(struct Curl_URL));
    break;
  case CURLUPART_SCHEME:
    curlx_safefree(u->scheme);
    u->guessed_scheme = FALSE;
    break;
  case CURLUPART_USER:
    curlx_safefree(u->user);
    break;
  case CURLUPART_PASSWORD:
    curlx_strzero(u->password);
    curlx_safefree(u->password);
    break;
  case CURLUPART_OPTIONS:
    curlx_safefree(u->options);
    break;
  case CURLUPART_HOST:
    curlx_safefree(u->host);
    break;
  case CURLUPART_ZONEID:
    curlx_safefree(u->zoneid);
    break;
  case CURLUPART_PORT:
    u->portnum = 0;
    u->port_present = FALSE;
    break;
  case CURLUPART_PATH:
    curlx_safefree(u->path);
    break;
  case CURLUPART_QUERY:
    curlx_safefree(u->query);
    u->query_present = FALSE;
    break;
  case CURLUPART_FRAGMENT:
    curlx_safefree(u->fragment);
    u->fragment_present = FALSE;
    break;
  default:
    return CURLUE_UNKNOWN_PART;
  }
  return CURLUE_OK;
}

static bool allowed_in_path(unsigned char x)
{
  switch(x) {
  case '!':
  case '$':
  case '&':
  case '\'':
  case '(':
  case ')':
  case '{':
  case '}':
  case '[':
  case ']':
  case '*':
  case '+':
  case ',':
  case ';':
  case '=':
  case ':':
  case '@':
  case '/':
    return TRUE;
  }
  return FALSE;
}

static CURLUcode url_encode_part(struct dynbuf *encp,
                                 const char *part,
                                 bool plusencode,
                                 bool pathmode,
                                 bool equalsencode)
{
  const unsigned char *i;

  for(i = (const unsigned char *)part; *i; i++) {
    CURLcode result;
    if((*i == ' ') && plusencode)
      result = curlx_dyn_addn(encp, "+", 1);
    else if(ISUNRESERVED(*i) ||
            (pathmode && allowed_in_path(*i)) ||
            ((*i == '=') && equalsencode)) {
      if((*i == '=') && equalsencode)
        /* only skip the first equals sign */
        equalsencode = FALSE;
      result = curlx_dyn_addn(encp, i, 1);
    }
    else {
      unsigned char out[3] = { '%' };
      Curl_hexbyte(&out[1], *i);
      result = curlx_dyn_addn(encp, out, 3);
    }
    if(result)
      return cc2cu(result);
  }
  return CURLUE_OK;
}

static CURLUcode url_uppercasehex_part(struct dynbuf *encp,
                                       const char *part)
{
  char *p;
  CURLcode result = curlx_dyn_add(encp, part);
  if(result)
    return cc2cu(result);
  p = curlx_dyn_ptr(encp);
  while(*p) {
    /* make sure percent encoded are upper case */
    if((*p == '%') && ISXDIGIT(p[1]) && ISXDIGIT(p[2]) &&
       (ISLOWER(p[1]) || ISLOWER(p[2]))) {
      p[1] = Curl_raw_toupper(p[1]);
      p[2] = Curl_raw_toupper(p[2]);
      p += 3;
    }
    else
      p++;
  }
  return CURLUE_OK;
}

static CURLUcode url_append_query(CURLU *u, struct dynbuf *encp)
{
  /* Append the 'encp' string onto the old query. Add a '&' separator if none
     is already present at the end of the existing query */

  size_t querylen = u->query ? strlen(u->query) : 0;
  bool addamperand = querylen && (u->query[querylen - 1] != '&');
  if(querylen) {
    struct dynbuf qbuf;
    CURLcode result;
    const char *newp = curlx_dyn_ptr(encp);
    curlx_dyn_init(&qbuf, CURL_MAX_INPUT_LENGTH);

    /* add original query */
    result = curlx_dyn_addn(&qbuf, u->query, querylen);
    if(!result && addamperand)
      /* add ampersand */
      result = curlx_dyn_addn(&qbuf, "&", 1);
    if(!result)
      /* add new query part */
      result = curlx_dyn_add(&qbuf, newp);
    if(result)
      goto nomem;
    curlx_dyn_free(encp);
    curlx_free(u->query);
    u->query = curlx_dyn_ptr(&qbuf);
    return CURLUE_OK;
nomem:
    curlx_dyn_free(encp);
    return cc2cu(result);
  }
  else {
    curlx_free(u->query);
    u->query = curlx_dyn_ptr(encp);
  }
  return CURLUE_OK;
}

static CURLUcode url_sethost(CURLU *u, struct dynbuf *encp,
                             bool urlencode,
                             unsigned int flags)
{
  size_t n = curlx_dyn_len(encp);
  bool bad = FALSE;
  char *newp = curlx_dyn_ptr(encp);
  if(!n)
    /* an empty hostname is okay if told so */
    bad = (flags & CURLU_NO_AUTHORITY) ? FALSE : TRUE;
  else if(!urlencode) {
    /* if the hostname part was not URL encoded here, it was set already URL
       encoded so we need to decode it to check */
    size_t dlen;
    char *decoded = NULL;
    CURLcode result = Curl_urldecode(newp, n, &decoded, &dlen, REJECT_CTRL);
    if(result || hostname_check6(u, decoded, dlen))
      bad = TRUE;
    curlx_free(decoded);
  }
  else if(hostname_check6(u, newp, n))
    bad = TRUE;
  if(bad) {
    curlx_dyn_free(encp);
    return CURLUE_BAD_HOSTNAME;
  }
  return CURLUE_OK;
}

CURLUcode curl_url_set(CURLU *u, CURLUPart what,
                       const char *part, unsigned int flags)
{
  char **storep = NULL;
  bool urlencode = (flags & CURLU_URLENCODE) ? 1 : 0;
  bool plusencode = FALSE;
  bool pathmode = FALSE;
  bool leadingslash = FALSE;
  bool appendquery = FALSE;
  bool equalsencode = FALSE;
  size_t nalloc;

  if(!u)
    return CURLUE_BAD_HANDLE;
  if(!part)
    /* setting a part to NULL clears it */
    return urlset_clear(u, what);

  nalloc = strlen(part);
  if(nalloc > CURL_MAX_INPUT_LENGTH)
    /* excessive input length */
    return CURLUE_MALFORMED_INPUT;

  switch(what) {
  case CURLUPART_SCHEME: {
    CURLUcode status = set_url_scheme(u, part, flags);
    if(status)
      return status;
    storep = &u->scheme;
    urlencode = FALSE; /* never */
    break;
  }
  case CURLUPART_USER:
    storep = &u->user;
    break;
  case CURLUPART_PASSWORD:
    storep = &u->password;
    break;
  case CURLUPART_OPTIONS:
    storep = &u->options;
    break;
  case CURLUPART_HOST:
    storep = &u->host;
    curlx_safefree(u->zoneid);
    break;
  case CURLUPART_ZONEID:
    storep = &u->zoneid;
    break;
  case CURLUPART_PORT:
    return set_url_port(u, part);
  case CURLUPART_PATH:
    pathmode = TRUE;
    leadingslash = TRUE; /* enforce */
    storep = &u->path;
    break;
  case CURLUPART_QUERY:
    plusencode = urlencode;
    appendquery = (flags & CURLU_APPENDQUERY) ? 1 : 0;
    equalsencode = appendquery;
    storep = &u->query;
    u->query_present = TRUE;
    break;
  case CURLUPART_FRAGMENT:
    storep = &u->fragment;
    u->fragment_present = TRUE;
    break;
  case CURLUPART_URL:
    return set_url(u, part, nalloc, flags);
  default:
    return CURLUE_UNKNOWN_PART;
  }
  DEBUGASSERT(storep);
  {
    const char *newp = NULL;
    struct dynbuf enc;
    CURLUcode status;
    curlx_dyn_init(&enc, (nalloc * 3) + 1 + leadingslash);

    if(leadingslash && (part[0] != '/')) {
      CURLcode result = curlx_dyn_addn(&enc, "/", 1);
      if(result)
        return cc2cu(result);
    }
    if(urlencode)
      status = url_encode_part(&enc, part, plusencode, pathmode, equalsencode);
    else
      status = url_uppercasehex_part(&enc, part);
    if(!status) {
      newp = curlx_dyn_ptr(&enc);

      if(appendquery && newp)
        return url_append_query(u, &enc);
      else if(what == CURLUPART_HOST)
        status = url_sethost(u, &enc, urlencode, flags);
    }
    if(status)
      return status;

    if(what == CURLUPART_PASSWORD)
      curlx_strzero(*storep);
    curlx_free(*storep);
    *storep = (char *)CURL_UNCONST(newp);
  }
  return CURLUE_OK;
}

bool Curl_url_same_origin(CURLU *base, CURLU *href)
{
  const struct Curl_scheme *s = NULL;

  /* base must be an absolute URL */
  if(!base->scheme || !base->host)
    return FALSE;
  if(href->scheme && !curl_strequal(base->scheme, href->scheme))
    return FALSE;
  if(href->host) {
    if(!curl_strequal(base->host, href->host))
      return FALSE;

    if(base->port_present != href->port_present) {
      /* one is present, one is not */
      s = Curl_get_scheme(base->scheme);
      if(!s) /* Cannot match default port for unknown scheme */
        return FALSE;
      /* to match, the present one must be the default port */
      if((base->port_present && (base->portnum != s->defport)) ||
         (href->port_present && (href->portnum != s->defport)))
        return FALSE;
    }
    else if(base->portnum != href->portnum) /* both present or missing */
      return FALSE;

    if(!curl_strequal(base->zoneid ? base->zoneid : "",
                      href->zoneid ? href->zoneid : ""))
      return FALSE;
  }
  else if(href->port_present) /* no host in href, then there must be no port */
    return FALSE;
  return TRUE;
}

CURLUcode Curl_url_get_port(CURLU *u, uint16_t *pport)
{
  if(u->port_present) {
    *pport = u->portnum;
    return CURLUE_OK;
  }
  else if(u->scheme) {
    const struct Curl_scheme *s = Curl_get_scheme(u->scheme);
    if(s && s->defport) {
      *pport = s->defport;
      return CURLUE_OK;
    }
  }
  *pport = 0;
  return CURLUE_NO_PORT;
}

Coverage Report

Created: 2026-09-14 07:06