--- parser3/src/main/untaint.C 2003/03/27 10:53:01 1.115.2.12.2.18 +++ parser3/src/main/untaint.C 2003/09/23 09:54:28 1.118.2.1 @@ -5,7 +5,7 @@ Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char* IDENT_UNTAINT_C="$Date: 2003/03/27 10:53:01 $"; +static const char* IDENT_UNTAINT_C="$Date: 2003/09/23 09:54:28 $"; #include "pa_string.h" @@ -17,6 +17,7 @@ static const char* IDENT_UNTAINT_C="$Dat #include "pa_common.h" #include "pa_charset.h" #include "pa_request_charsets.h" +#include "pa_sapi.h" extern "C" { // author forgot to do that #include "ec.h" @@ -28,10 +29,6 @@ extern "C" { // author forgot to do that #include "pa_sql_connection.h" #endif -// forwards - -void CORD_ec_flush_buf_optimize_whitespace(CORD_ec x); - // defines @@ -84,7 +81,7 @@ inline bool need_file_encode(unsigned ch return strchr( "*?'\"<>|" #ifndef WIN32 - ":\\~" + ":\\" #endif , c)!=0; } @@ -170,6 +167,40 @@ inline bool need_quote_http_header(const return false; } +#ifndef DOXYGEN +struct Append_fragment_info { + String::Language lang; + const StringBody dest_body; + String::Languages* dest_languages; +}; +#endif +int append_fragment_optimizing(char alang, size_t asize, Append_fragment_info& info) { + String::Language lang=(String::Language)alang; + // main idea here: + // tainted piece would get OPTIMIZED bit from 'lang' + // clean piece would be marked OPTIMIZED manually + // pieces with determined languages [not tainted|clean] would retain theirs langs + info.dest_languages->append(info.dest_body, + lang==String::Language::L_TAINTED? + info.lang + :lang==String::Language::L_CLEAN? + (String::Language)(String::Language::L_CLEAN|String::Language::L_OPTIMIZE_BIT) // ORing with OPTIMIZED flag + :lang, + asize); + + return 0; // 0=continue +} +int append_fragment_nonoptimizing(char alang, size_t asize, Append_fragment_info& info) { + String::Language lang=(String::Language)alang; + // The core idea: tainted pieces got marked with context's lang + info.dest_languages->append(info.dest_body, + lang==String::Language::L_TAINTED? + info.lang + :lang, + asize); + + return 0; // 0=continue +} /** appends to other String, @@ -181,39 +212,22 @@ String& String::append_to(String& dest, if(is_empty()) return dest; - // first: letters - dest.body< i(fragments); - while(i.has_next()) { - const Fragment fragment=i.next(); - // main idea here: - // tainted piece would get OPTIMIZED bit from 'lang' - // clean piece would be marked OPTIMIZED manually - // pieces with determined languages [not tainted|clean] would retain theirs langs - dest.fragments+=Fragment(static_cast( - fragment.lang==L_TAINTED?lang:( - fragment.lang==L_CLEAN?L_CLEAN|L_OPTIMIZE_BIT: // ORing with OPTIMIZED flag - fragment.lang)), - fragment.length); - } - } else { // The core idea: tainted pieces got marked with context's lang - Array_iterator i(fragments); - while(i.has_next()) { - const Fragment fragment=i.next(); - dest.fragments+=Fragment( - fragment.lang==L_TAINTED?lang:fragment.lang, - fragment.length); - } + dest.langs.append(dest.body, lang, length()); + else { + Append_fragment_info info={lang, dest.body, &dest.langs}; + langs.for_each(body, lang&L_OPTIMIZE_BIT? + append_fragment_optimizing + :append_fragment_nonoptimizing, info); } + // next: letters + dest.body<='0' && c<='9' || strchr("!*+-/", c); } +inline bool addr_spec_soon(const char *src) { + for(char c; c=*src; src++) + if(c=='<') + return true; + else if(!(c==' ' || c=='\t')) + return false; + return false; +} /** RFC Upper case should be used for hexadecimal digits "A" through "F" @@ -259,12 +281,28 @@ inline bool mail_header_char_valid_withi inline bool mail_header_nonspace_char(char c) { return c != 0x20; } -inline void ec_append(CORD_ec& result, bool& optimize, bool& whitespace, CORD_pos& pos, size_t size) { +inline void ec_append(CORD_ec& result, bool& optimize, bool& whitespace, CORD_pos pos, size_t size) { while(size--) { CORD_ec_append(result, CORD_pos_fetch(pos)); CORD_next(pos); } } +inline void pa_CORD_pos_advance(CORD_pos pos, size_t n) { + while(true) { + size_t avail=CORD_pos_chars_left(pos); + if(avail==0) { + CORD_next(pos); + if(!--n) + break; + } else if(avail=n + CORD_pos_advance(pos, n); + break; + } + } +} StringBody String::cstr_to_string_body(Language lang, SQL_Connection* connection, const Request_charsets *charsets) const { @@ -285,7 +323,7 @@ StringBody String::cstr_to_string_body(L bool optimize=(to_lang & L_OPTIMIZE_BIT)!=0; if(!optimize) whitespace=false; - + switch(to_lang & ~L_OPTIMIZE_BIT) { case L_CLEAN: case L_TAINTED: @@ -303,7 +341,7 @@ StringBody String::cstr_to_string_body(L // tainted, untaint language: file [name] escape( // Macintosh has problems with small Russian letter 'r' - if( c=='\xF0' && charsets && charsets->source().name()=="windows-1251" ) { + if( c=='\xF0' && charsets && charsets->source().NAME()=="WINDOWS-1251" ) { // fixing that letter for most common charset to_char('p'); } else // fallback to default @@ -315,20 +353,15 @@ StringBody String::cstr_to_string_body(L { const char *fragment_str=body.mid(fragment_begin, fragment_length).cstr(); // skip source [we use recoded version] - CORD_pos_advance(pos, fragment_length); - const void *client_ptr; - size_t client_size; + pa_CORD_pos_advance(pos, fragment_length); + String::C output(fragment_str, fragment_length); if(charsets) - Charset::transcode( - charsets->source(), fragment_str, fragment_length, - charsets->client(), client_ptr, client_size); - else { - client_ptr=fragment_str; - client_size=fragment_length; - } + output=Charset::transcode(output, + charsets->source(), + charsets->client()); char c; - for(const char* src=(const char* )client_ptr; c=*src++; ) + for(const char* src=output.str; c=*src++; ) encode(need_uri_encode, '%', c); } break; @@ -346,9 +379,9 @@ StringBody String::cstr_to_string_body(L const char *mail_ptr= body.mid(fragment_begin, mail_size=fragment_length).cstr(); // skip source [we use recoded version] - CORD_pos_advance(pos, mail_size); + pa_CORD_pos_advance(pos, mail_size); - const char* charset_name=charsets->mail().name().cstr(); + const char* charset_name=charsets->mail().NAME().cstr(); // Subject: Re: parser3: =?koi8-r?Q?=D3=C5=CD=C9=CE=C1=D2?= bool to_quoted_printable=false; @@ -357,7 +390,7 @@ StringBody String::cstr_to_string_body(L uchar c; for(const char* src=mail_ptr; c=(uchar)*src++; ) { //RFC + An 'encoded-word' MUST NOT appear in any portion of an 'addr-spec'. - if(to_quoted_printable && (c==',' || c=='<')) { + if(to_quoted_printable && (c==',' || addr_spec_soon(src) || c == '"')) { email=c=='<'; to_string("?="); to_quoted_printable=false; @@ -398,7 +431,7 @@ StringBody String::cstr_to_string_body(L if(connection) { const char *fragment_str=body.mid(fragment_begin, fragment_length).cstr(); // skip source [we use recoded version] - CORD_pos_advance(pos, fragment_length); + pa_CORD_pos_advance(pos, fragment_length); to_string(connection->quote(fragment_str, fragment_length)); } else @@ -437,10 +470,8 @@ StringBody String::cstr_to_string_body(L }); break; default: - throw Exception(0, - 0, - "unknown untaint language #%d", - static_cast(to_lang)); // should never + SAPI::die("unknown untaint language #%d", + static_cast(to_lang)); // should never break; // never } }