1 files changed, 12 insertions, 4 deletions

diff --git a/lib/util.cc b/lib/util.cc
index ac70938..69d37b5 100644
--- a/lib/util.cc
+++ b/lib/util.cc
@@ -128,120 +128,128 @@ namespace opkele {
             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;
         }
 
         /*
          *
          */
 
         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 hext 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 colon = uri.find(':');
+             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(), uri.begin()+colon+1,
+                     uri.begin()+ns, uri.begin()+colon+1,
                      back_inserter(rv), ::tolower );
              bool s;
              if(rv=="http:")
                  s = false;
              else if(rv=="https:")
                  s = true;
+#ifndef NDEBUG
+             else if(rv=="file:")
+                 s = false;
+#endif /* XXX: or try to make tests work some other way */
              else
                  throw not_implemented(OPKELE_CP_ "Only http(s) URIs can be normalized here");
-             string::size_type ul = uri.length();
+             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");
              rv += "//";
              string::size_type interesting = uri.find_first_of(":/#?",colon+3);
              if(interesting==string::npos) {
                  transform(
-                         uri.begin()+colon+3,uri.end(),
+                         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[6];
                      snprintf(tmp,sizeof(tmp),"%ld",port);
                      rv += ':'; 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);