robustness improvements in link descovery mechanisms

accept stray spaces in link tags and inside relevant attribute values.

Thanks, Gen, for pointing it out.

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

1 files changed, 4 insertions, 4 deletions

diff --git a/lib/consumer.cc b/lib/consumer.cc
index 299b3bc..ff5da91 100644
--- a/lib/consumer.cc
+++ b/lib/consumer.cc
@@ -194,196 +194,196 @@ namespace opkele {
 	    unsigned char *md = HMAC(
 		    EVP_sha1(),
 		    &(secret.front()),secret.size(),
 		    (const unsigned char *)kv.data(),kv.length(),
 		    0,&md_len);
 	    if(sig.size()!=md_len || memcmp(&(sig.front()),md,md_len))
 		throw id_res_mismatch(OPKELE_CP_ "signature mismatch");
 	}catch(failed_lookup& e) { /* XXX: more specific? */
 	    const string& slist = pin.get_param("openid.signed");
 	    string::size_type pp = 0;
 	    params_t p;
 	    while(true) {
 		string::size_type co = slist.find(',',pp);
 		string f = "openid.";
 		f += (co==string::npos)?slist.substr(pp):slist.substr(pp,co-pp);
 		p[f] = pin.get_param(f);
 		if(co==string::npos)
 		    break;
 		pp = co+1;
 	    }
 	    p["openid.assoc_handle"] = pin.get_param("openid.assoc_handle");
 	    p["openid.sig"] = pin.get_param("openid.sig");
 	    p["openid.signed"] = pin.get_param("openid.signed");
 	    try {
 		string ih = pin.get_param("openid.invalidate_handle");
 		p["openid.invalidate_handle"] = ih;
 	    }catch(failed_lookup& fl) { }
 	    try {
 		check_authentication(server,p);
 	    }catch(failed_check_authentication& fca) {
 		throw id_res_failed(OPKELE_CP_ "failed to check_authentication()");
 	    }
 	}
 	if(ext) ext->id_res_hook(pin,ps,identity);
     }
 
     void consumer_t::check_authentication(const string& server,const params_t& p) {
 	string request = "openid.mode=check_authentication";
 	for(params_t::const_iterator i=p.begin();i!=p.end();++i) {
 	    if(i->first!="openid.mode") {
 		request += '&';
 		request += i->first;
 		request += '=';
 		request += util::url_encode(i->second);
 	    }
 	}
 	curl_t curl = curl_easy_init();
 	if(!curl)
 	    throw exception_curl(OPKELE_CP_ "failed to curl_easy_init()");
 	string response;
 	CURLcode r;
 	(r=curl_misc_sets(curl))
 	|| (r=curl_easy_setopt(curl,CURLOPT_URL,server.c_str()))
 	|| (r=curl_easy_setopt(curl,CURLOPT_POST,1))
 	|| (r=curl_easy_setopt(curl,CURLOPT_POSTFIELDS,request.data()))
 	|| (r=curl_easy_setopt(curl,CURLOPT_POSTFIELDSIZE,request.length()))
 	|| (r=curl_easy_setopt(curl,CURLOPT_WRITEFUNCTION,_curl_tostring))
 	|| (r=curl_easy_setopt(curl,CURLOPT_WRITEDATA,&response))
 	;
 	if(r)
 	    throw exception_curl(OPKELE_CP_ "failed to curl_easy_setopt()",r);
 	if(r=curl_easy_perform(curl))
 	    throw exception_curl(OPKELE_CP_ "failed to curl_easy_perform()",r);
 	params_t pp; pp.parse_keyvalues(response);
 	if(pp.has_param("invalidate_handle"))
 	    invalidate_assoc(server,pp.get_param("invalidate_handle"));
 	if(pp.has_param("is_valid")) {
 	    if(pp.get_param("is_valid")=="true")
 		return;
 	}else if(pp.has_param("lifetime")) {
 	    if(util::string_to_long(pp.get_param("lifetime")))
 		return;
 	}
 	throw failed_check_authentication(OPKELE_CP_ "failed to verify response");
     }
 
     void consumer_t::retrieve_links(const string& url,string& server,string& delegate) {
 #if defined(USE_LIBPCRECPP) || defined(USE_PCREPP)
 	server.erase();
 	delegate.erase();
 	curl_t curl = curl_easy_init();
 	if(!curl)
 	    throw exception_curl(OPKELE_CP_ "failed to curl_easy_init()");
 	string html;
 	CURLcode r;
 	(r=curl_misc_sets(curl))
 	|| (r=curl_easy_setopt(curl,CURLOPT_URL,url.c_str()))
 	|| (r=curl_easy_setopt(curl,CURLOPT_WRITEFUNCTION,_curl_tostring))
 	|| (r=curl_easy_setopt(curl,CURLOPT_WRITEDATA,&html))
 	;
 	if(r)
 	    throw exception_curl(OPKELE_CP_ "failed to curl_easy_setopt()",r);
 	r = curl_easy_perform(curl);
 	if(r && r!=CURLE_WRITE_ERROR)
 	    throw exception_curl(OPKELE_CP_ "failed to curl_easy_perform()",r);
 	// strip out everything past body
-	static const char *re_hdre = "<head[^>]*>",
-		     *re_lre = "<link\\b([^>]+)>",
-		     *re_rre = "\\brel=['\"]([^'\"]+)['\"]",
-		     *re_hre = "\\bhref=['\"]([^'\"]+)['\"]";
+	static const char *re_hdre = "<\\s*head[^>]*>",
+		     *re_lre = "<\\s*link\\b([^>]+)>",
+		     *re_rre = "\\brel\\s*=\\s*['\"]\\s*([^'\"\\s]+)\\s*['\"]",
+		     *re_hre = "\\bhref\\s*=\\s*['\"]\\s*([^'\"\\s]+)\\s*['\"]";
 #if defined(USE_LIBPCRECPP)
 	static pcrecpp::RE_Options ro(PCRE_CASELESS|PCRE_DOTALL);
 	static pcrecpp::RE
 	    bre("<body\\b.*",ro), hdre(re_hdre,ro),
 	    lre(re_lre,ro), rre(re_rre), hre(re_hre,ro);
 	bre.Replace("",&html);
 	pcrecpp::StringPiece hpiece(html);
 	if(!hdre.FindAndConsume(&hpiece))
 	    throw bad_input(OPKELE_CP_ "failed to find head");
 	string attrs;
 	while(lre.FindAndConsume(&hpiece,&attrs)) {
 	    pcrecpp::StringPiece rel, href;
 	    if(!(rre.PartialMatch(attrs,&rel) && hre.PartialMatch(attrs,&href)))
 		continue;
 	    if(rel=="openid.server") {
 		href.CopyToString(&server);
 		if(!delegate.empty())
 		    break;
 	    }else if(rel=="openid.delegate") {
 		href.CopyToString(&delegate);
 		if(!server.empty())
 		    break;
 	    }
 	}
 #elif defined(USE_PCREPP)
 	pcrepp::Pcre bre("<body\\b",PCRE_CASELESS);
 	if(bre.search(html))
 	    html.erase(bre.get_match_start());
 	pcrepp::Pcre hdre(re_hdre,PCRE_CASELESS);
 	if(!hdre.search(html))
 	    throw bad_input(OPKELE_CP_ "failed to find head");
 	html.erase(0,hdre.get_match_end()+1);
 	pcrepp::Pcre lre(re_lre,PCRE_CASELESS), rre(re_rre,PCRE_CASELESS), hre(re_hre,PCRE_CASELESS);
 	while(lre.search(html)) {
 	    string attrs = lre[0];
 	    html.erase(0,lre.get_match_end()+1);
 	    if(!(rre.search(attrs)&&hre.search(attrs)))
 		continue;
 	    if(rre[0]=="openid.server") {
 		server = hre[0];
 		if(!delegate.empty())
 		    break;
 	    }else if(rre[0]=="openid.delegate") {
 		delegate = hre[0];
 		if(!server.empty())
 		    break;
 	    }
 	}
 #else
 	#error "I must have gone crazy"
 #endif
 	if(server.empty())
 	    throw failed_assertion(OPKELE_CP_ "The location has no openid.server declaration");
 #else /* none of the RE bindings enabled */
 	throw not_implemented(OPKELE_CP_ "No internal implementation of retrieve_links were provided at compile-time");
 #endif
     }
 
     assoc_t consumer_t::find_assoc(const string& server) {
 	throw failed_lookup(OPKELE_CP_ "no find_assoc() provided");
     }
 
     string consumer_t::normalize(const string& url) {
 	string rv = url;
 	// strip leading and trailing spaces
 	string::size_type i = rv.find_first_not_of(" \t\r\n");
 	if(i==string::npos)
 	    throw bad_input(OPKELE_CP_ "empty URL");
 	if(i)
 	    rv.erase(0,i);
 	i = rv.find_last_not_of(" \t\r\n");
 	assert(i!=string::npos);
 	if(i<(rv.length()-1))
 	    rv.erase(i+1);
 	// add missing http://
 	i = rv.find("://");
 	if(i==string::npos) { // primitive. but do we need more?
 	    rv.insert(0,"http://");
 	    i = sizeof("http://")-1;
 	}else{
 	    i += sizeof("://")-1;
 	}
 	string::size_type qm = rv.find('?',i);
 	string::size_type sl = rv.find('/',i);
 	if(qm!=string::npos) {
 	    if(sl==string::npos || sl>qm)
 		rv.insert(qm,1,'/');
 	}else{
 	    if(sl==string::npos)
 		rv += '/';
 	}
 	return rv;
     }
 
     string consumer_t::canonicalize(const string& url) {
 	string rv = normalize(url);