--- parser3/src/main/pa_common.C 2009/07/06 12:07:04 1.247 +++ parser3/src/main/pa_common.C 2012/06/04 05:47:29 1.268 @@ -1,7 +1,7 @@ /** @file Parser: commonly functions. - Copyright(c) 2001-2009 ArtLebedev Group (http://www.artlebedev.com) + Copyright (c) 2000-2012 Art. Lebedev Studio (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) * BASE64 part @@ -26,8 +26,6 @@ * */ -static const char * const IDENT_COMMON_C="$Date: 2009/07/06 12:07:04 $"; - #include "pa_common.h" #include "pa_exception.h" #include "pa_hash.h" @@ -38,6 +36,8 @@ static const char * const IDENT_COMMON_C #include "pcre.h" #include "pa_request.h" +volatile const char * IDENT_PA_COMMON_C="$Id: pa_common.C,v 1.268 2012/06/04 05:47:29 misha Exp $" IDENT_PA_COMMON_H IDENT_PA_HASH_H IDENT_PA_ARRAY_H IDENT_PA_STACK_H; + // some maybe-undefined constants #ifndef _O_TEXT @@ -67,6 +67,31 @@ const String file_status_name(FILE_STATU // functions +bool capitalized(const char* s){ + bool upper=true; + for(const char* c=s; *c; c++){ + if(*c != (upper ? toupper((unsigned char)*c) : tolower((unsigned char)*c))) + return false; + upper=strchr("-_ ", *c) != 0; + } + return true; +} + +const char* capitalize(const char* s){ + if(!s || capitalized(s)) + return s; + + char* result=pa_strdup(s); + if(result){ + bool upper=true; + for(char* c=result; *c; c++){ + *c=upper ? (char)toupper((unsigned char)*c) : (char)tolower((unsigned char)*c); + upper=strchr("-_ ", *c) != 0; + } + } + return (const char*)result; +} + void fix_line_breaks(char *str, size_t& length) { //_asm int 3; const char* const eob=str+length; @@ -77,7 +102,7 @@ void fix_line_breaks(char *str, size_t& while(char* eol=(char*)memchr(bol, '\r', eob -bol)) { size_t len=eol-bol; if(dest!=bol) - memcpy(dest, bol, len); + memmove(dest, bol, len); dest+=len; *dest++='\n'; @@ -89,7 +114,7 @@ void fix_line_breaks(char *str, size_t& } // last piece without \r if(dest!=bol) - memcpy(dest, bol, eob-bol); + memmove(dest, bol, eob-bol); str[length]=0; // terminating } @@ -150,7 +175,7 @@ static void file_read_action( lseek(f, info.offset, SEEK_SET); *info.data=info.buf ? info.buf - : new(PointerFreeGC) char[to_read_size+(as_text?1:0)]; + : (char *)pa_malloc_atomic(to_read_size+1); *info.data_size=(size_t)read(f, *info.data, to_read_size); if(ssize_t(*info.data_size)<0 || *info.data_size>to_read_size) @@ -160,7 +185,7 @@ static void file_read_action( *info.data_size, to_read_size); } else { // empty file // for both, text and binary: for text we need that terminator, for binary we need nonzero pointer to be able to save such files - *info.data=new(PointerFreeGC) char[1]; + *info.data=(char *)pa_malloc_atomic(1); *(char*)(*info.data)=0; *info.data_size=0; return; @@ -175,9 +200,7 @@ File_read_result file_read(Request_chars if(params){ int valid_options=pa_get_valid_file_options_count(*params); if(valid_options!=params->count()) - throw Exception(PARSER_RUNTIME, - 0, - "invalid option passed"); + throw Exception(PARSER_RUNTIME, 0, CALLED_WITH_INVALID_OPTION); } File_read_action_info info={&result.str, &result.length, buf, offset, count}; @@ -188,23 +211,24 @@ File_read_result file_read(Request_chars if(as_text){ if(result.success){ + Charset* asked_charset=0; if(result.length>=3 && strncmp(result.str, "\xEF\xBB\xBF", 3)==0){ // skip UTF-8 signature (BOM code) result.str+=3; result.length-=3; + asked_charset=&UTF8_charset; } - if(result.length && transcode_text_result && params){ // must be checked because transcode returns CONST string in case length==0, which contradicts hacking few lines below - if(Value* vcharset_name=params->get(PA_CHARSET_NAME)){ - Charset asked_charset=::charsets.get(vcharset_name->as_string(). - change_case(charsets.source(), String::CC_UPPER)); - - String::C body=String::C(result.str, result.length); - body=Charset::transcode(body, asked_charset, charsets.source()); + if(params) + if(Value* vcharset_name=params->get(PA_CHARSET_NAME)) + asked_charset=&::charsets.get(vcharset_name->as_string().change_case(charsets.source(), String::CC_UPPER)); + + if(result.length && transcode_text_result && asked_charset){ // length must be checked because transcode returns CONST string in case length==0, which contradicts hacking few lines below + String::C body=String::C(result.str, result.length); + body=Charset::transcode(body, *asked_charset, charsets.source()); - result.str=const_cast(body.str); // hacking a little - result.length=body.length; - } + result.str=const_cast(body.str); // hacking a little + result.length=body.length; } } if(result.length) @@ -239,8 +263,9 @@ File_read_result file_load(Request& r, c return result; } + #ifdef PA_SAFE_MODE -void check_safe_mode(struct stat finfo, const String& file_spec, const char* fname) { +void check_safe_mode(struct stat finfo, const String& file_spec, const char* fname) { if(finfo.st_uid/*foreign?*/!=geteuid() && finfo.st_gid/*foreign?*/!=getegid()) throw Exception(PARSER_RUNTIME, @@ -252,9 +277,14 @@ void check_safe_mode(struct stat finfo, "fuid(%d)!=euid(%d) or fgid(%d)!=egid(%d)", fname, finfo.st_uid, geteuid(), - finfo.st_gid, getegid()); -} -#endif + finfo.st_gid, getegid()); +} +#else +void check_safe_mode(struct stat, const String&, const char*) { +} +#endif + + bool file_read_action_under_lock(const String& file_spec, const char* action_name, File_read_action action, void *context, @@ -281,15 +311,13 @@ bool file_read_action_under_lock(const S strerror(errno), errno, fname); struct stat finfo; - if(stat(fname, &finfo)!=0) + if(fstat(f, &finfo)!=0) throw Exception("file.missing", // hardly possible: we just opened it OK &file_spec, "stat failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname); -#ifdef PA_SAFE_MODE check_safe_mode(finfo, file_spec, fname); -#endif action(finfo, f, file_spec, fname, as_text, context); } catch(...) { @@ -303,7 +331,8 @@ bool file_read_action_under_lock(const S return true; } else { if(fail_on_read_problem) - throw Exception(errno==EACCES?"file.access":errno==ENOENT?"file.missing":0, + throw Exception(errno==EACCES?"file.access" + :(errno==ENOENT || errno==ENOTDIR || errno==ENODEV)?"file.missing":0, &file_spec, "%s failed: %s (%d), actual filename '%s'", action_name, strerror(errno), errno, fname); @@ -350,7 +379,12 @@ bool file_write_action_under_lock( } try { - action(f, context); +#if (defined(HAVE_FCHMOD) && defined(PA_SAFE_MODE)) + struct stat finfo; + if(fstat(f, &finfo)==0 && finfo.st_mode & 0111) + fchmod(f, finfo.st_mode & 0666/*clear executable bits*/); // backward: ignore errors if any +#endif + action(f, context); } catch(...) { #ifdef HAVE_FTRUNCATE if(!do_append) @@ -376,7 +410,8 @@ bool file_write_action_under_lock( #ifndef DOXYGEN struct File_write_action_info { - const char* str; size_t length; + const char* str; + size_t length; }; #endif static void file_write_action(int f, void *context) { @@ -390,10 +425,21 @@ static void file_write_action(int f, voi } } void file_write( - const String& file_spec, - const char* data, size_t size, + Request_charsets& charsets, + const String& file_spec, + const char* data, + size_t size, bool as_text, - bool do_append) { + bool do_append, + Charset* asked_charset) { + + if(as_text && asked_charset){ + String::C body=String::C(data, size); + body=Charset::transcode(body, charsets.source(), *asked_charset); + data=body.str; + size=body.length; + }; + File_write_action_info info={data, size}; file_write_action_under_lock( @@ -405,13 +451,62 @@ void file_write( do_append); } +static size_t get_dir(char* fname, size_t helper_length){ + bool dir=false; + size_t pos=0; + for(pos=helper_length; pos; pos--){ + char c=fname[pos-1]; + if(c=='/' || c=='\\'){ + fname[pos-1]=0; + dir=true; + } else if(dir) break; + } + return pos; +} + +static bool entry_readable(char* fname, bool need_dir) { + if(need_dir){ + size_t size=strlen(fname); + while(size) { + char c=fname[size-1]; + if(c=='/' || c=='\\') + fname[--size]=0; + else + break; + } + } + + struct stat finfo; + if(access(fname, R_OK)==0 && entry_exists(fname, &finfo)) { + bool is_dir=(finfo.st_mode&S_IFDIR) != 0; + return is_dir==need_dir; + } + return false; +} + +static bool entry_readable(const String& file_spec, bool need_dir) { + return entry_readable(file_spec.taint_cstrm(String::L_FILE_SPEC), need_dir); +} + // throws nothing! [this is required in file_move & file_delete] -static void rmdir(const String& file_spec, size_t pos_after) { - size_t pos_before; - if((pos_before=file_spec.pos('/', pos_after))!=STRING_NOT_FOUND) - rmdir(file_spec, pos_before+1); - - rmdir(file_spec.mid(0, pos_after-1/* / */).taint_cstr(String::L_FILE_SPEC)); +static void rmdir(const String& file_spec, size_t pos_after=0) { + char* dir_spec=file_spec.taint_cstrm(String::L_FILE_SPEC); + size_t length=strlen(dir_spec); + while( (length=get_dir(dir_spec, length)) && (length > pos_after) ){ +#ifdef WIN32 + if(!entry_readable(dir_spec, true)) + break; + DWORD attrs=GetFileAttributes(dir_spec); + if( + (attrs==INVALID_FILE_ATTRIBUTES) + || !(attrs & FILE_ATTRIBUTE_DIRECTORY) + || (attrs & FILE_ATTRIBUTE_REPARSE_POINT) + ) + break; +#endif + if( rmdir(dir_spec) ) + break; + }; } bool file_delete(const String& file_spec, bool fail_on_problem) { @@ -458,26 +553,6 @@ bool entry_exists(const String& file_spe return entry_exists(fname, 0); } -static bool entry_readable(const String& file_spec, bool need_dir) { - char* fname=file_spec.taint_cstrm(String::L_FILE_SPEC); - if(need_dir) { - size_t size=strlen(fname); - while(size) { - char c=fname[size-1]; - if(c=='/' || c=='\\') - fname[--size]=0; - else - break; - } - } - struct stat finfo; - if(access(fname, R_OK)==0 && entry_exists(fname, &finfo)) { - bool is_dir=(finfo.st_mode&S_IFDIR) != 0; - return is_dir==need_dir; - } - return false; -} - bool file_exist(const String& file_spec) { return entry_readable(file_spec, false); } @@ -683,7 +758,7 @@ enum EscapeState { }; // @todo prescan for reduce required size (unescaped sting in 1 byte charset requires less memory usually) -char* unescape_chars(const char* cp, int len, Charset* charset, bool ignore_plus){ +char* unescape_chars(const char* cp, int len, Charset* charset, bool js){ char* s=new(PointerFreeGC) char[len+1]; // must be enough (%uXXXX==6 bytes, max utf-8 char length==6 bytes) char* dst=s; EscapeState escapeState=EscapeRest; @@ -692,12 +767,12 @@ char* unescape_chars(const char* cp, int short int jsCnt=0; while(srcPos> 16); + *outptr++ = (unsigned char)(saved >> 8); + *outptr++ = (unsigned char)(saved); + i = 0; + } } } @@ -1159,7 +1264,7 @@ g_mime_utils_base64_decode_step (const u i = 2; while (inptr > in && i) { inptr--; - if (gmime_base64_rank[*inptr] != 0xff) { + if (gmime_base64_rank[*inptr] <= 0xfe) { if (*inptr == '=' && outptr > out) outptr--; i--; @@ -1172,16 +1277,18 @@ g_mime_utils_base64_decode_step (const u char* pa_base64_encode(const char *in, size_t in_size){ - /* wont go to more than 2x size (overly conservative) */ - char* result=new(PointerFreeGC) char[in_size * 2 + 6]; + size_t new_size = ((in_size / 3 + 1) * 4); + new_size += new_size / (BASE64_GROUPS_IN_LINE * 4)/*new lines*/ + 1/*zero terminator*/; + char* result = new(PointerFreeGC) char[new_size]; int state=0; int save=0; #ifndef NDEBUG size_t filled= #endif - g_mime_utils_base64_encode_close ((const unsigned char*)in, in_size, - (unsigned char*)result, &state, &save); - assert(filled <= in_size * 2 + 6); + g_mime_utils_base64_encode_close ((const unsigned char*)in, in_size, (unsigned char*)result, &state, &save); + + //throw Exception(PARSER_RUNTIME, 0, "%d %d %d", in_size, new_size, filled); + assert(filled <= new_size); return result; } @@ -1223,16 +1330,22 @@ static void file_base64_file_action( } } -void pa_base64_decode(const char *in, size_t in_size, char*& result, size_t& result_size) { - /* wont go to more than had (overly conservative) */ - result=new(PointerFreeGC) char[in_size+1/*terminator*/]; +void pa_base64_decode(const char *in, size_t in_size, char*& result, size_t& result_size, bool strict) { + // every 4 base64 bytes are converted into 3 normal bytes + // not full set (tail) of 4-bytes set is ignored + size_t new_size=in_size/4*3; + result=new(PointerFreeGC) char[new_size+1/*terminator*/]; + int state=0; int save=0; result_size= - g_mime_utils_base64_decode_step ((const unsigned char*)in, in_size, - (unsigned char*)result, &state, &save); - assert(result_size <= in_size); + g_mime_utils_base64_decode_step ((const unsigned char*)in, in_size, + (unsigned char*)result, &state, &save, strict); + assert(result_size <= new_size); result[result_size]=0; // for text files + + if(strict && state!=0) + throw Exception(BASE64_FORMAT, 0, "Unexpected end of chars"); }