fix to rfc normalization

It kept prepending a '/' to the trailing segment even if the segment was past
[?#]

Signed-off-by: Michael Krelin <hacker@klever.net>

1 files changed, 2 insertions, 1 deletions

diff --git a/lib/util.cc b/lib/util.cc
index ee75d29..a9b9bed 100644
--- a/lib/util.cc
+++ b/lib/util.cc
@@ -116,200 +116,201 @@ namespace opkele {
         time_t w3c_to_time(const string& w) {
             int fraction;
             struct tm tm_t;
             memset(&tm_t,0,sizeof(tm_t));
             if( (
                         sscanf(
                             w.c_str(),
                             "%04d-%02d-%02dT%02d:%02d:%02dZ",
                             &tm_t.tm_year,&tm_t.tm_mon,&tm_t.tm_mday,
                             &tm_t.tm_hour,&tm_t.tm_min,&tm_t.tm_sec
                             ) != 6
                 ) && (
                     sscanf(
                         w.c_str(),
                         "%04d-%02d-%02dT%02d:%02d:%02d.%03dZ",
                         &tm_t.tm_year,&tm_t.tm_mon,&tm_t.tm_mday,
                         &tm_t.tm_hour,&tm_t.tm_min,&tm_t.tm_sec,
                         &fraction
                         ) != 7
                     ) )
                 throw failed_conversion(OPKELE_CP_ "failed to sscanf()");
             tm_t.tm_mon--;
             tm_t.tm_year-=1900;
             time_t rv = mktime(&tm_t);
             if(rv==(time_t)-1)
                 throw failed_conversion(OPKELE_CP_ "failed to mktime()");
             return rv-timezone;
         }
 
         /*
          *
          */
 
         string url_encode(const string& str) {
             char * t = curl_escape(str.c_str(),str.length());
             if(!t)
                 throw failed_conversion(OPKELE_CP_ "failed to curl_escape()");
             string rv(t);
             curl_free(t);
             return rv;
         }
 
         string long_to_string(long l) {
             char rv[32];
             int r=snprintf(rv,sizeof(rv),"%ld",l);
             if(r<0 || r>=(int)sizeof(rv))
                 throw failed_conversion(OPKELE_CP_ "failed to snprintf()");
             return rv;
         }
 
         long string_to_long(const string& s) {
             char *endptr = 0;
             long rv = strtol(s.c_str(),&endptr,10);
             if((!endptr) || endptr==s.c_str())
                 throw failed_conversion(OPKELE_CP_ "failed to strtol()");
             return rv;
         }
 
         /*
          * Normalize URL according to the rules, described in rfc 3986, section 6
          *
          * - uppercase hex triplets (e.g. %ab -> %AB)
          * - lowercase scheme and host
          * - decode %-encoded characters, specified as unreserved in rfc 3986, section 2.3,
          *   that is - [:alpha:][:digit:]._~-
          * - remove dot segments
          * - remove empty and default ports
          * - if there's no path component, add '/'
          */
          string rfc_3986_normalize_uri(const string& uri) {
              static const char *whitespace = " \t\r\n";
              string rv;
              string::size_type ns = uri.find_first_not_of(whitespace);
              if(ns==string::npos)
                  throw bad_input(OPKELE_CP_ "Can't normalize empty URI");
              string::size_type colon = uri.find(':',ns);
              if(colon==string::npos)
                  throw bad_input(OPKELE_CP_ "No scheme specified in URI");
              transform(
                      uri.begin()+ns, uri.begin()+colon+1,
                      back_inserter(rv), ::tolower );
              bool s;
              string::size_type ul = uri.find_last_not_of(whitespace)+1;
              if(ul <= (colon+3))
                  throw bad_input(OPKELE_CP_ "Unexpected end of URI being normalized encountered");
              if(uri[colon+1]!='/' || uri[colon+2]!='/')
                  throw bad_input(OPKELE_CP_ "Unexpected input in URI being normalized after scheme component");
              if(rv=="http:")
                  s = false;
              else if(rv=="https:")
                  s = true;
              else{
                  /* TODO: support more schemes.
                   * e.g. xri. How do we normalize
                   * xri?
                   */
                  rv.append(uri,colon+1,ul-colon-1);
                  return rv;
              }
              rv += "//";
              string::size_type interesting = uri.find_first_of(":/#?",colon+3);
              if(interesting==string::npos) {
                  transform(
                          uri.begin()+colon+3,uri.begin()+ul,
                          back_inserter(rv), ::tolower );
                  rv += '/'; return rv;
              }
              transform(
                      uri.begin()+colon+3,uri.begin()+interesting,
                      back_inserter(rv), ::tolower );
              bool qf = false;
              char ic = uri[interesting];
              if(ic==':') {
                  string::size_type ni = uri.find_first_of("/#?%",interesting+1);
                  const char *nptr = uri.data()+interesting+1;
                  char *eptr = 0;
                  long port = strtol(nptr,&eptr,10);
                  if( (port>0) && (port<65535) && port!=(s?443:80) ) {
                      char tmp[8];
                      snprintf(tmp,sizeof(tmp),":%ld",port);
                      rv += tmp;
                  }
                  if(ni==string::npos) {
                      rv += '/'; return rv;
                  }
                  interesting = ni;
              }else if(ic!='/') {
                  rv += '/'; rv += ic;
                  qf = true;
                  ++interesting;
              }
              string::size_type n = interesting;
              char tmp[3] = { 0,0,0 };
              stack<string::size_type> psegs; psegs.push(rv.length());
              string pseg;
              for(;n<ul;) {
                  string::size_type unsafe = uri.find_first_of(qf?"%":"%/?#",n);
                  if(unsafe==string::npos) {
                      pseg.append(uri,n,ul-n-1); n = ul-1;
                  }else{
                      pseg.append(uri,n,unsafe-n);
                      n = unsafe;
                  }
                  char c = uri[n++];
                  if(c=='%') {
                      if((n+1)>=ul)
                          throw bad_input(OPKELE_CP_ "Unexpected end of URI encountered while parsing percent-encoded character");
                      tmp[0] = uri[n++];
                      tmp[1] = uri[n++];
                      if(!( isxdigit(tmp[0]) && isxdigit(tmp[1]) ))
                          throw bad_input(OPKELE_CP_ "Invalid percent-encoded character in URI being normalized");
                      int cc = strtol(tmp,0,16);
                      if( isalpha(cc) || isdigit(cc) || strchr("._~-",cc) )
                          pseg += cc;
                      else{
                          pseg += '%';
                          pseg += toupper(tmp[0]); pseg += toupper(tmp[1]);
                      }
                  }else if(qf) {
                      rv += pseg; rv += c;
                      pseg.clear();
                  }else if(n>=ul || strchr("?/#",c)) {
                      if(pseg.empty() || pseg==".") {
                      }else if(pseg=="..") {
                          if(psegs.size()>1) {
                              rv.resize(psegs.top()); psegs.pop();
                          }
                      }else{
                          psegs.push(rv.length());
                          if(c!='/') {
                              pseg += c;
                              qf = true;
                          }
                          rv += '/'; rv += pseg;
                      }
                      if(c=='/' && (n>=ul || strchr("?#",uri[n])) ) {
                          rv += '/';
                          if(n<ul)
                              qf = true;
                      }else if(strchr("?#",c)) {
                          if(psegs.size()==1 && psegs.top()==rv.length())
                              rv += '/';
                          if(pseg.empty())
                              rv += c;
                          qf = true;
                      }
                      pseg.clear();
                  }else{
                      pseg += c;
                  }
              }
              if(!pseg.empty()) {
-                 rv += '/'; rv += pseg;
+                 if(!qf) rv += '/';
+                 rv += pseg;
              }
              return rv;
          }
 
     }
 
 }