proxyurl test: percent-encoded specials and UTF-8

Pin that the host/port split is byte-transparent: a percent-encoded '@' (%40) or
':' (%3A) in the userinfo stays encoded and never mis-splits host or port, a
second raw '@' is handled by the last-'@' host rule, and UTF-8 bytes in the
username and host label pass through intact. Decoding happens later at auth time,
not in the parser. Raw UTF-8 CLI args match the existing 01_engine-idna.test.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Xavier Roche <roche@httrack.com>

src/htscoremain.c

@@ -2253,21 +2253,12 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
               htsmain_free();
               return -1;
             } else {
-              char *a;
+              char BIGSTK pname[HTS_URLMAXSIZE * 2];
 
               na++;
               opt->proxy.active = 1;
-              // Rechercher MAIS en partant de la fin à cause de user:pass@proxy:port
-              a = argv[na] + strlen(argv[na]) - 1;
-              while((a > argv[na]) && (*a != ':') && (*a != '@'))
-                a--;
-              if (*a == ':') {  // un port est présent, <proxy>:port
-                sscanf(a + 1, "%d", &opt->proxy.port);
-                StringCopyN(opt->proxy.name, argv[na], (int) (a - argv[na]));
-              } else {          // <proxy>
-                opt->proxy.port = 8080;
-                StringCopy(opt->proxy.name, argv[na]);
-              }
+              hts_parse_proxy(argv[na], pname, sizeof(pname), &opt->proxy.port);
+              StringCopy(opt->proxy.name, pname);
             }
             break;
           case 'F':            // user-agent field

src/htslib.c

@@ -3866,6 +3866,10 @@ const char *jump_protocol_const(const char *source) {
     source += p;
   else if ((p = strfield(source, "file:")))
     source += p;
+  else if ((p = strfield(source, "socks5h:")))
+    source += p;
+  else if ((p = strfield(source, "socks5:")))
+    source += p;
   // net_path
   if (strncmp(source, "//", 2) == 0)
     source += 2;
@@ -3874,6 +3878,43 @@ const char *jump_protocol_const(const char *source) {
 
 DECLARE_NON_CONST_VERSION(jump_protocol)
 
+// default proxy port for a -P argument, keyed on the scheme
+static int proxy_default_port(const char *arg) {
+  if (strfield(arg, "socks5h:") || strfield(arg, "socks5:"))
+    return 1080;
+  return 8080;
+}
+
+void hts_parse_proxy(const char *arg, char *name, size_t name_size, int *port) {
+  const char *authority = strstr(arg, "://");
+  const char *a;
+  size_t namelen;
+
+  if (name_size == 0)
+    return;
+  // scan back to the port ':' (or userinfo '@'), but never past the authority,
+  // so a scheme's own colon is not read as a port separator; inspect a[-1] from
+  // one-past-end so no pointer below the string is ever formed
+  authority = (authority != NULL) ? authority + 3 : arg;
+  a = arg + strlen(arg);
+  while (a > authority && a[-1] != ':' && a[-1] != '@')
+    a--;
+  if (a > authority && a[-1] == ':') {
+    int p = -1;
+
+    sscanf(a, "%d", &p);
+    *port = (p > 0) ? p : proxy_default_port(arg);
+    namelen = (size_t) (a - 1 - arg);
+  } else {
+    *port = proxy_default_port(arg);
+    namelen = strlen(arg);
+  }
+  if (namelen >= name_size) // arg is user-controlled: truncate, don't overflow
+    namelen = name_size - 1;
+  memcpy(name, arg, namelen);
+  name[namelen] = '\0';
+}
+
 // codage base 64 a vers b
 void code64(unsigned char *a, int size_a, unsigned char *b, int crlf) {
   int i1 = 0, i2 = 0, i3 = 0, i4 = 0;

src/htslib.h

@@ -289,6 +289,12 @@ int may_unknown2(httrackp * opt, const char *mime, const char *filename);
 const char *strrchr_limit(const char *s, char c, const char *limit);
 char *jump_protocol(char *source);
 const char *jump_protocol_const(const char *source);
+
+/* Split a -P proxy argument "[scheme://][user:pass@]host[:port]" into the proxy
+   host string (scheme and any user:pass kept, for later stripping and auth),
+   written NUL-terminated into name[name_size] (truncated to fit), and the port
+   in *port. The port defaults by scheme: 1080 for socks5/socks5h, else 8080. */
+void hts_parse_proxy(const char *arg, char *name, size_t name_size, int *port);
 void code64(unsigned char *a, int size_a, unsigned char *b, int crlf);
 
 #define copychar(catbuff,a) concat(catbuff,(a),NULL)

src/htsselftest.c

@@ -1308,6 +1308,22 @@ static int st_identurl(httrackp *opt, int argc, char **argv) {
   return 0;
 }
 
+static int st_proxyurl(httrackp *opt, int argc, char **argv) {
+  char BIGSTK name[HTS_URLMAXSIZE * 2];
+  int port = -1;
+
+  (void) opt;
+  if (argc < 1) {
+    fprintf(stderr, "proxyurl: needs a proxy argument\n");
+    return 1;
+  }
+  hts_parse_proxy(argv[0], name, sizeof(name), &port);
+  // host= is what the connect actually resolves (scheme + user:pass stripped)
+  printf("name=%s port=%d host=%s\n", name, port,
+         jump_identification_const(name));
+  return 0;
+}
+
 /* Regression for the one-byte fil[] overflow: a 2047-byte hostless "?"-URL used
    to abort in strncat_safe_ when the missing leading '/' pushed fil to 2048. */
 static int st_identabs(httrackp *opt, int argc, char **argv) {
@@ -2756,6 +2772,8 @@ static const struct selftest_entry {
     {"resolve", "<link> <adr> <fil>", "resolve a link against an origin",
      st_resolve},
     {"identurl", "<url>", "split an absolute URL into (adr, fil)", st_identurl},
+    {"proxyurl", "<proxy-arg>", "parse a -P proxy URL into host/port",
+     st_proxyurl},
     {"identabs", "", "ident_url_absolute one-byte fil[] overflow self-test",
      st_identabs},
     {"header", "<raw-header-line> ...", "response header-line parsing",

tests/01_engine-proxyurl.test

@@ -0,0 +1,44 @@
+#!/bin/bash
+#
+
+set -euo pipefail
+
+# hts_parse_proxy splits a -P argument "[scheme://][user:pass@]host[:port]" into
+# the proxy host string and port. -#test=proxyurl <arg> prints
+# "name=.. port=.. host=..", where host= is what the connect resolves.
+
+p() {
+    test "$(httrack -O /dev/null -#test=proxyurl "$1")" == "$2" || exit 1
+}
+
+# bare host, with and without an explicit port
+p 'proxy.example.com:8080' 'name=proxy.example.com port=8080 host=proxy.example.com'
+p 'proxy.example.com' 'name=proxy.example.com port=8080 host=proxy.example.com'
+
+# user:pass@ is kept in name (for proxy auth) and stripped from host
+p 'user:pass@proxy:3128' 'name=user:pass@proxy port=3128 host=proxy'
+p 'user:pass@proxy' 'name=user:pass@proxy port=8080 host=proxy'
+
+# a scheme colon must not be read as a port: http://host with no port used to
+# parse to name=http with a garbage port
+p 'http://proxy:8080' 'name=http://proxy port=8080 host=proxy'
+p 'http://proxy' 'name=http://proxy port=8080 host=proxy'
+
+# socks5/socks5h default to 1080 and resolve to the bare host
+p 'socks5://host:1080' 'name=socks5://host port=1080 host=host'
+p 'socks5://host' 'name=socks5://host port=1080 host=host'
+p 'socks5h://host' 'name=socks5h://host port=1080 host=host'
+# explicit port wins over the 1080 default, with and without userinfo
+p 'socks5://host:9050' 'name=socks5://host port=9050 host=host'
+p 'socks5://user:pass@host:9050' 'name=socks5://user:pass@host port=9050 host=host'
+
+# the split is byte-transparent: percent-escapes stay encoded (decoded later at
+# auth time), only structural ASCII ':'/'@' delimit, and the last '@' ends the
+# userinfo. So %40/%3A and UTF-8 bytes never mis-split host or port.
+p 'socks5://user:p%40ss@host:1080' 'name=socks5://user:p%40ss@host port=1080 host=host'
+p 'socks5://us%3Aer:pass@host:1080' 'name=socks5://us%3Aer:pass@host port=1080 host=host'
+p 'socks5://a@b:c@host:1080' 'name=socks5://a@b:c@host port=1080 host=host'
+p 'socks5://naïve:pass@héllo:1080' 'name=socks5://naïve:pass@héllo port=1080 host=héllo'
+
+# a lone ':' must not underflow the backward scan
+p ':' 'name= port=8080 host='

tests/Makefile.am

@@ -46,6 +46,7 @@ TESTS = \
 	01_engine-header.test \
 	01_engine-idna.test \
 	01_engine-identurl.test \
+	01_engine-proxyurl.test \
 	01_engine-identabs.test \
 	01_engine-escape-room.test \
 	01_engine-inplace-escape.test \