2 files changed, 66 insertions, 13 deletions

diff --git a/lib/Makefile.am b/lib/Makefile.am
index 9c73e96..989de28 100644
--- a/lib/Makefile.am
+++ b/lib/Makefile.am
@@ -1,31 +1,31 @@
 lib_LTLIBRARIES = libopkele.la
 
 AM_CPPFLAGS = ${CPPFLAGS_DEBUG}
 DEFAULT_INCLUDES = -I${top_builddir}
 INCLUDES = \
            -I${top_srcdir}/include/ \
            ${KONFORKA_CFLAGS} \
            ${OPENSSL_CFLAGS} \
            ${LIBCURL_CPPFLAGS} \
-           ${PCRE_CFLAGS} ${EXPAT_CFLAGS}
+           ${PCRE_CFLAGS} ${EXPAT_CFLAGS} ${TIDY_CFLAGS}
 libopkele_la_LIBADD = \
         ${LIBCURL} \
         ${PCRE_LIBS} ${EXPAT_LIBS} \
         ${OPENSSL_LIBS} \
-        ${KONFORKA_LIBS}
+        ${KONFORKA_LIBS} ${TIDY_LIBS}
 
 libopkele_la_SOURCES = \
                        params.cc \
                        util.cc \
                        server.cc \
                        secret.cc \
                        data.cc \
                        consumer.cc \
                        exception.cc \
                        extension.cc \
                        sreg.cc \
                        extension_chain.cc \
                        curl.cc expat.cc \
                        discovery.cc
 libopkele_la_LDFLAGS = \
         -version-info 2:0:0
diff --git a/lib/discovery.cc b/lib/discovery.cc
index 8729cfb..a308b56 100644
--- a/lib/discovery.cc
+++ b/lib/discovery.cc
@@ -1,230 +1,283 @@
 #include <list>
 #include <opkele/curl.h>
 #include <opkele/expat.h>
 #include <opkele/uris.h>
 #include <opkele/discovery.h>
 #include <opkele/exception.h>
 #include <opkele/util.h>
+#include <opkele/tidy.h>
 #include <opkele/debug.h>
 
 #include "config.h"
 
 #define XRDS_HEADER "X-XRDS-Location"
 #define CT_HEADER "Content-Type"
 
 namespace opkele {
     using std::list;
     using xrd::XRD_t;
     using xrd::service_t;
 
     static const char *whitespace = " \t\r\n";
     static const char *i_leaders = "=@+$!(";
+    static const size_t max_html = 16384;
 
     static inline bool is_qelement(const XML_Char *n,const char *qen) {
         return !strcasecmp(n,qen);
     }
     static inline bool is_element(const XML_Char *n,const char *en) {
         if(!strcasecmp(n,en)) return true;
         int nl = strlen(n), enl = strlen(en);
         if( (nl>=(enl+1)) && n[nl-enl-1]=='\t'
                 && !strcasecmp(&n[nl-enl],en) )
             return true;
         return false;
     }
 
     static long element_priority(const XML_Char **a) {
         for(;*a;++a)
             if(!strcasecmp(*(a++),"priority")) {
                 long rv;
                 return (sscanf(*a,"%ld",&rv)==1)?rv:-1;
             }
         return -1;
     }
 
     class idigger_t : public util::curl_t, public util::expat_t {
         public:
             string xri_proxy;
 
             enum {
                 xmode_html = 1, xmode_xrd = 2
             };
             int xmode;
 
             string xrds_location;
             string http_content_type;
             service_t html_openid1;
             service_t html_openid2;
             string cdata_buf;
             long status_code;
             string status_string;
 
             typedef list<string> pt_stack_t;
             pt_stack_t pt_stack;
             int skipping;
+            bool parser_choked;
+            string save_html;
 
             XRD_t *xrd;
             service_t *xrd_service;
             string* cdata;
 
             idigger_t()
                 : util::curl_t(easy_init()),
                 util::expat_t(0),
                 xri_proxy(XRI_PROXY_URL) {
                     CURLcode r;
                     (r=misc_sets())
                         || (r=set_write())
                         || (r=set_header())
                         ;
                     if(r)
                         throw exception_curl(OPKELE_CP_ "failed to set curly options",r);
                 }
             ~idigger_t() throw() { }
 
             void discover(idiscovery_t& result,const string& identity) {
                 result.clear();
                 string::size_type fsc = identity.find_first_not_of(whitespace);
                 if(fsc==string::npos)
                     throw bad_input(OPKELE_CP_ "whtiespace-only identity");
                 string::size_type lsc = identity.find_last_not_of(whitespace);
                 assert(lsc!=string::npos);
                 if(!strncasecmp(identity.c_str()+fsc,"xri://",sizeof("xri://")-1))
                     fsc += sizeof("xri://")-1;
                 if((fsc+1)>=lsc)
                     throw bad_input(OPKELE_CP_ "not a character of importance in identity");
                 string id(identity,fsc,lsc-fsc+1);
                 if(strchr(i_leaders,id[0])) {
                     result.normalized_id = id;
                     result.xri_identity = true;
                     /* TODO: further canonicalize xri identity? Like folding case  or whatever... */
                     discover_at(
                             result,
                             xri_proxy + util::url_encode(id)+
                             "?_xrd_r=application/xrd+xml;sep=false", xmode_xrd);
                     if(status_code!=100)
                         throw failed_xri_resolution(OPKELE_CP_
                                 "XRI resolution failed with '"+status_string+"' message",status_code);
                     if(result.xrd.canonical_ids.empty())
                         throw opkele::failed_discovery(OPKELE_CP_ "No CanonicalID for XRI identity found");
                     result.canonicalized_id = result.xrd.canonical_ids.begin()->second;
                 }else{
                     result.xri_identity = false;
                     if(id.find("://")==string::npos)
                         id.insert(0,"http://");
                     string::size_type fp = id.find('#');
                     if(fp!=string::npos) {
                         string::size_type qp = id.find('?');
                         if(qp==string::npos || qp<fp)
                             id.erase(fp);
                         else if(qp>fp)
                             id.erase(fp,qp-fp);
                     }
                     result.normalized_id = util::rfc_3986_normalize_uri(id);
                     discover_at(result,id,xmode_html|xmode_xrd);
                     const char * eu = 0;
                     CURLcode r = easy_getinfo(CURLINFO_EFFECTIVE_URL,&eu);
                     if(r)
                         throw exception_curl(OPKELE_CP_ "failed to get CURLINFO_EFFECTIVE_URL",r);
                     result.canonicalized_id = util::rfc_3986_normalize_uri(eu); /* XXX: strip fragment part? */
                     if(xrds_location.empty()) {
                         html2xrd(result.xrd);
                     }else{
                         discover_at(result,xrds_location,xmode_xrd);
                         if(result.xrd.empty())
                             html2xrd(result.xrd);
                     }
                 }
             }
 
             void discover_at(idiscovery_t& result,const string& url,int xm) {
                 CURLcode r = easy_setopt(CURLOPT_URL,url.c_str());
                 if(r)
                     throw exception_curl(OPKELE_CP_ "failed to set culry urlie",r);
 
-                (*(expat_t*)this) = parser_create_ns();
-                set_user_data(); set_element_handler();
-                set_character_data_handler();
-
                 http_content_type.clear();
                 xmode = xm;
+                prepare_to_parse();
                 if(xmode&xmode_html) {
                     xrds_location.clear();
-                    html_openid1.clear(); html_openid2.clear();
+                    save_html.clear();
+                    save_html.reserve(max_html);
                 }
                 xrd = &result.xrd;
-                cdata = 0; xrd_service = 0; skipping = 0;
-                status_code = 100; status_string.clear();
 
                 r = easy_perform();
                 if(r && r!=CURLE_WRITE_ERROR)
                     throw exception_curl(OPKELE_CP_ "failed to perform curly request",r);
 
-                parse(0,0,true);
+                if(!parser_choked) {
+                    parse(0,0,true);
+                }else{
+                    /* TODO: do not bother if we've seen xml */
+                    try {
+                        util::tidy_doc_t td = util::tidy_doc_t::create();
+                        if(!td)
+                            throw exception_tidy(OPKELE_CP_ "failed to create htmltidy document");
+#ifndef NDEBUG
+                        td.opt_set(TidyQuiet,false);
+                        td.opt_set(TidyShowWarnings,false);
+                        td.opt_set(TidyForceOutput,true);
+                        td.opt_set(TidyXhtmlOut,true);
+                        td.opt_set(TidyDoctypeMode,TidyDoctypeOmit);
+                        td.opt_set(TidyMark,false);
+#endif /* NDEBUG */
+                        if(td.parse_string(save_html)<=0)
+                            throw exception_tidy(OPKELE_CP_ "tidy failed to parse document");
+                        if(td.clean_and_repair()<=0)
+                            throw exception_tidy(OPKELE_CP_ "tidy failed to clean and repair");
+                        util::tidy_buf_t tide;
+                        if(td.save_buffer(tide)<=0)
+                            throw exception_tidy(OPKELE_CP_ "tidy failed to save buffer");
+                        prepare_to_parse();
+                        parse(tide.c_str(),tide.size(),true);
+                    }catch(exception_tidy& et) { }
+                }
+                save_html.clear();
+            }
+
+            void prepare_to_parse() {
+                (*(expat_t*)this) = parser_create_ns();
+                set_user_data(); set_element_handler();
+                set_character_data_handler();
+
+                if(xmode&xmode_html) {
+                    html_openid1.clear(); html_openid2.clear();
+                    parser_choked = false;
+                }
+
+                cdata = 0; xrd_service = 0; skipping = 0;
+                status_code = 100; status_string.clear();
             }
 
             void html2xrd(XRD_t& x) {
                 if(!html_openid1.uris.empty()) {
                     html_openid1.types.insert(STURI_OPENID11);
                     x.services.add(-1,html_openid1);
                 }
                 if(!html_openid2.uris.empty()) {
                     html_openid2.types.insert(STURI_OPENID20);
                     x.services.add(-1,html_openid2);
                 }
             }
 
             size_t write(void *p,size_t s,size_t nm) {
-                if(skipping<0) return 0;
                 /* TODO: limit total size */
                 size_t bytes = s*nm;
-                bool rp = parse((const char *)p,bytes,false);
+                const char *inbuf = (const char*)p;
+                if(xmode&xmode_html) {
+                    size_t mbts = save_html.capacity()-save_html.size();
+                    size_t bts = 0;
+                    if(mbts>0) {
+                        bts = (bytes>mbts)?mbts:bytes;
+                        save_html.append(inbuf,bts);
+                    }
+                    if(skipping<0) return bts;
+                }
+                if(skipping<0) return 0;
+                bool rp = parse(inbuf,bytes,false);
                 if(!rp) {
+                    parser_choked = true;
                     skipping = -1;
-                    bytes = 0;
+                    if(!(xmode&xmode_html))
+                        bytes = 0;
                 }
                 return bytes;
             }
             size_t header(void *p,size_t s,size_t nm) {
                 size_t bytes = s*nm;
                 const char *h = (const char*)p;
                 const char *colon = (const char*)memchr(p,':',bytes);
                 const char *space = (const char*)memchr(p,' ',bytes);
                 if(space && ( (!colon) || space<colon ) ) {
                     xrds_location.clear(); http_content_type.clear();
                 }else if(colon) {
                     const char *hv = ++colon;
                     int hnl = colon-h;
                     int rb;
                     for(rb = bytes-hnl-1;rb>0 && isspace(*hv);++hv,--rb);
                     while(rb>0 && isspace(hv[rb-1])) --rb;
                     if(rb) {
                         if( (hnl>=sizeof(XRDS_HEADER))
                                 && !strncasecmp(h,XRDS_HEADER":",
                                     sizeof(XRDS_HEADER)) ) {
                             xrds_location.assign(hv,rb);
                         }else if( (hnl>=sizeof(CT_HEADER))
                                 && !strncasecmp(h,CT_HEADER":",
                                     sizeof(CT_HEADER)) ) {
                             const char *sc = (const char*)memchr(
                                     hv,';',rb);
                             http_content_type.assign(hv,sc?(sc-hv):rb);
                         }
                     }
                 }
                 return curl_t::header(p,s,nm);
             }
 
             void start_element(const XML_Char *n,const XML_Char **a) {
                 if(skipping<0) return;
                 if(skipping) {
                     if(xmode&xmode_html)
                         html_start_element(n,a);
                     ++skipping; return;
                 }
                 if(pt_stack.empty()) {
                     if(is_qelement(n,NSURI_XRDS "\tXRDS"))
                         return;
                     if(is_qelement(n,NSURI_XRD "\tXRD")) {
                         assert(xrd);
                         xrd->clear();
                         pt_stack.push_back(n);
                     }else if(xmode&xmode_html) {