Annotation of parser3/src/main/pa_common.C, revision 1.310
1.15 paf 1: /** @file
1.16 paf 2: Parser: commonly functions.
3:
1.303 moko 4: Copyright (c) 2000-2017 Art. Lebedev Studio (http://www.artlebedev.com)
1.101 paf 5: Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
1.306 moko 6: */
1.210 paf 7:
1.1 paf 8: #include "pa_common.h"
1.4 paf 9: #include "pa_exception.h"
1.154 paf 10: #include "pa_hash.h"
1.14 paf 11: #include "pa_globals.h"
1.154 paf 12: #include "pa_charsets.h"
1.214 paf 13: #include "pa_http.h"
1.223 misha 14: #include "pa_request_charsets.h"
1.237 misha 15: #include "pcre.h"
1.241 misha 16: #include "pa_request.h"
1.98 paf 17:
1.283 moko 18: #include "pa_idna.h"
19: #include "pa_convert_utf.h"
20:
1.273 moko 21: #ifdef _MSC_VER
1.276 moko 22: #include <windows.h>
1.273 moko 23: #include <direct.h>
24: #endif
25:
1.277 moko 26: #ifdef _MSC_VER
27: #define pa_mkdir(path, mode) _mkdir(path)
28: #else
29: #define pa_mkdir(path, mode) mkdir(path, mode)
30: #endif
31:
1.310 ! moko 32: volatile const char * IDENT_PA_COMMON_C="$Id: pa_common.C,v 1.309 2019/11/22 23:11:25 moko Exp $" IDENT_PA_COMMON_H IDENT_PA_HASH_H IDENT_PA_ARRAY_H IDENT_PA_STACK_H;
1.267 moko 33:
1.93 paf 34: // some maybe-undefined constants
35:
1.82 paf 36: #ifndef _O_TEXT
37: # define _O_TEXT 0
38: #endif
39: #ifndef _O_BINARY
40: # define _O_BINARY 0
1.47 paf 41: #endif
1.80 paf 42:
1.138 paf 43: #ifdef HAVE_FTRUNCATE
44: # define PA_O_TRUNC 0
45: #else
46: # ifdef _O_TRUNC
47: # define PA_O_TRUNC _O_TRUNC
48: # else
49: # error you must have either ftruncate function or _O_TRUNC bit declared
50: # endif
1.154 paf 51: #endif
1.176 paf 52:
1.154 paf 53: // defines for globals
54:
55: #define FILE_STATUS_NAME "status"
56:
57: // globals
58:
59: const String file_status_name(FILE_STATUS_NAME);
60:
1.305 moko 61: String sql_bind_name(SQL_BIND_NAME);
62: String sql_limit_name(PA_SQL_LIMIT_NAME);
63: String sql_offset_name(PA_SQL_OFFSET_NAME);
64: String sql_default_name(SQL_DEFAULT_NAME);
65: String sql_distinct_name(SQL_DISTINCT_NAME);
66: String sql_value_type_name(SQL_VALUE_TYPE_NAME);
67:
1.301 moko 68: // forwards
69:
70: const UTF16* pa_utf16_encode(const char* in, Charset& source_charset);
71:
1.154 paf 72: // functions
1.127 paf 73:
1.301 moko 74: #ifdef _MSC_VER
75:
76: int pa_stat(const char *pathname, struct stat *buffer){
77: const UTF16* utf16name=pa_utf16_encode(pathname, pa_thread_request().charsets.source());
78: return _wstat64((const wchar_t *)utf16name, buffer);
79: }
80:
81: int pa_open(const char *pathname, int flags, int mode){
82: const UTF16* utf16name=pa_utf16_encode(pathname, pa_thread_request().charsets.source());
83: return _wopen((const wchar_t *)utf16name, flags, mode);
84: }
85:
86: FILE *pa_fopen(const char *pathname, const char *mode){
87: const UTF16* utf16name=pa_utf16_encode(pathname, pa_thread_request().charsets.source());
88: const UTF16* utf16mode=pa_utf16_encode(mode, pa_thread_request().charsets.source());
89: return _wfopen((const wchar_t *)utf16name, (const wchar_t *)utf16mode);
90: }
91:
92: #endif
93:
1.206 paf 94: /// these options were handled but not checked elsewhere, now check them
1.239 misha 95: int pa_get_valid_file_options_count(HashStringValue& options) {
1.206 paf 96: int result=0;
97: if(options.get(PA_SQL_LIMIT_NAME))
98: result++;
99: if(options.get(PA_SQL_OFFSET_NAME))
100: result++;
101: if(options.get(PA_COLUMN_SEPARATOR_NAME))
102: result++;
103: if(options.get(PA_COLUMN_ENCLOSER_NAME))
104: result++;
1.223 misha 105: if(options.get(PA_CHARSET_NAME))
106: result++;
1.206 paf 107: return result;
108: }
109:
1.123 paf 110: #ifndef DOXYGEN
111: struct File_read_action_info {
1.154 paf 112: char **data; size_t *data_size;
1.310 ! moko 113: char* buf; size_t offset; size_t limit;
1.126 paf 114: };
1.123 paf 115: #endif
1.271 moko 116:
1.289 moko 117: static void file_read_action(struct stat& finfo, int f, const String& file_spec, void *context) {
1.310 ! moko 118: File_read_action_info& info = *static_cast<File_read_action_info *>(context);
! 119: size_t to_read_size = info.limit;
1.188 paf 120: if(!to_read_size)
1.310 ! moko 121: to_read_size = check_file_size(finfo.st_size, file_spec);
1.271 moko 122: if(to_read_size) {
1.188 paf 123: if(info.offset)
124: lseek(f, info.offset, SEEK_SET);
1.310 ! moko 125: *info.data = info.buf ? info.buf : (char *)pa_malloc_atomic(to_read_size+1);
! 126: ssize_t result = read(f, *info.data, to_read_size);
1.271 moko 127: if(result<0)
128: throw Exception("file.read", &file_spec, "read failed: %s (%d)", strerror(errno), errno);
1.310 ! moko 129: *info.data_size = result;
1.123 paf 130: } else { // empty file
1.209 paf 131: // for both, text and binary: for text we need that terminator, for binary we need nonzero pointer to be able to save such files
1.310 ! moko 132: *info.data = (char *)pa_malloc_atomic(1);
! 133: *(char*)(*info.data) = 0;
! 134: *info.data_size = 0;
1.123 paf 135: return;
136: }
1.126 paf 137: }
1.241 misha 138:
1.310 ! moko 139: File_read_result file_read_binary(const String& file_spec, bool fail_on_read_problem, char* buf, size_t offset, size_t limit) {
! 140: File_read_result result = {false, 0, 0, 0};
! 141: File_read_action_info info = {&result.str, &result.length, buf, offset, limit};
1.309 moko 142:
1.310 ! moko 143: result.success = file_read_action_under_lock(file_spec, "read", file_read_action, &info, 0, fail_on_read_problem);
1.309 moko 144: return result;
145: }
146:
1.154 paf 147: File_read_result file_read(Request_charsets& charsets, const String& file_spec,
1.310 ! moko 148: bool as_text, HashStringValue *options,
1.229 misha 149: bool fail_on_read_problem,
1.310 ! moko 150: size_t offset = 0, size_t limit = 0, bool transcode_text_result = true) {
! 151: File_read_result result = {false, 0, 0, 0};
! 152: if(options){
! 153: int valid_options = pa_get_valid_file_options_count(*options);
! 154: if(valid_options != options->count())
1.262 misha 155: throw Exception(PARSER_RUNTIME, 0, CALLED_WITH_INVALID_OPTION);
1.241 misha 156: }
1.203 paf 157:
1.310 ! moko 158: File_read_action_info info = {&result.str, &result.length, 0, offset, limit};
1.161 paf 159:
1.310 ! moko 160: result.success = file_read_action_under_lock(file_spec, "read", file_read_action, &info, as_text, fail_on_read_problem);
1.223 misha 161:
1.241 misha 162: if(as_text){
163: if(result.success){
1.310 ! moko 164: Charset* asked_charset = 0;
! 165: if(options)
! 166: if(Value* vcharset_name = options->get(PA_CHARSET_NAME))
! 167: asked_charset = &pa_charsets.get(vcharset_name->as_string());
1.263 misha 168:
1.310 ! moko 169: asked_charset = pa_charsets.checkBOM(result.str, result.length, asked_charset);
1.287 moko 170:
1.263 misha 171: 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
1.310 ! moko 172: String::C body = String::C(result.str, result.length);
1.263 misha 173: body=Charset::transcode(body, *asked_charset, charsets.source());
1.236 misha 174:
1.310 ! moko 175: result.str = const_cast<char*>(body.str); // hacking a little
! 176: result.length = body.length;
1.131 paf 177: }
178: }
1.241 misha 179: if(result.length)
180: fix_line_breaks(result.str, result.length);
1.123 paf 181: }
1.241 misha 182:
183: return result;
184: }
185:
186: File_read_result file_load(Request& r, const String& file_spec,
1.310 ! moko 187: bool as_text, HashStringValue *options,
1.241 misha 188: bool fail_on_read_problem,
1.310 ! moko 189: bool transcode_text_result) {
! 190:
! 191: size_t offset = 0;
! 192: size_t limit = 0;
! 193:
! 194: if(options){
! 195: if(Value *voffset = (Value *)options->get(sql_offset_name))
! 196: offset = r.process(*voffset).as_int();
! 197: if(Value *vlimit = (Value *)options->get(sql_limit_name))
! 198: limit = r.process(*vlimit).as_int();
! 199: // no check on options count here
! 200: }
1.241 misha 201:
202: if(file_spec.starts_with("http://")) {
1.310 ! moko 203: if(offset || limit)
1.289 moko 204: throw Exception(PARSER_RUNTIME, 0, "offset and load options are not supported for HTTP:// file load");
1.241 misha 205:
206: // fail on read problem
1.310 ! moko 207: File_read_http_result http = pa_internal_file_read_http(r, file_spec, as_text, options, transcode_text_result);
! 208:
! 209: File_read_result result = {true, http.str, http.length, http.headers};
! 210: return result;
1.241 misha 211: } else
1.310 ! moko 212: return file_read(r.charsets, file_spec, as_text, options, fail_on_read_problem, offset, limit, transcode_text_result);
! 213: }
1.126 paf 214:
1.310 ! moko 215: char* file_read_text(Request_charsets& charsets, const String& file_spec, bool fail_on_read_problem) {
! 216: File_read_result file = file_read(charsets, file_spec, true, 0, fail_on_read_problem);
! 217: return file.success ? file.str : 0;
1.123 paf 218: }
219:
1.310 ! moko 220: char* file_load_text(Request& r, const String& file_spec, bool fail_on_read_problem, HashStringValue* options, bool transcode_result) {
! 221: File_read_result file = file_load(r, file_spec, true, options, fail_on_read_problem, transcode_result);
! 222: return file.success ? file.str : 0;
! 223: }
1.257 pretende 224:
1.154 paf 225: #ifdef PA_SAFE_MODE
1.259 misha 226: void check_safe_mode(struct stat finfo, const String& file_spec, const char* fname) {
1.154 paf 227: if(finfo.st_uid/*foreign?*/!=geteuid()
228: && finfo.st_gid/*foreign?*/!=getegid())
1.289 moko 229: throw Exception(PARSER_RUNTIME,
230: &file_spec,
231: "parser is in safe mode: reading files of foreign group and user disabled "
232: "[recompile parser with --disable-safe-mode configure option], "
233: "actual filename '%s', fuid(%d)!=euid(%d) or fgid(%d)!=egid(%d)",
234: fname, finfo.st_uid, geteuid(), finfo.st_gid, getegid()
235: );
1.259 misha 236: }
237: #else
238: void check_safe_mode(struct stat, const String&, const char*) {
239: }
1.257 pretende 240: #endif
1.259 misha 241:
1.257 pretende 242:
1.149 paf 243:
1.154 paf 244: bool file_read_action_under_lock(const String& file_spec,
1.126 paf 245: const char* action_name, File_read_action action, void *context,
246: bool as_text,
1.123 paf 247: bool fail_on_read_problem) {
1.247 misha 248: const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC);
1.33 paf 249: int f;
250:
251: // first open, next stat:
1.45 paf 252: // directory update of NTFS hard links performed on open.
1.33 paf 253: // ex:
254: // a.html:^test[] and b.html hardlink to a.html
255: // user inserts ! before ^test in a.html
1.126 paf 256: // directory entry of b.html in NTFS not updated at once,
1.35 paf 257: // they delay update till open, so we would receive "!^test[" string
258: // if would do stat, next open.
1.123 paf 259: // later: it seems, even this does not help sometimes
1.301 moko 260: if((f=pa_open(fname, O_RDONLY|(as_text?_O_TEXT:_O_BINARY)))>=0) {
1.123 paf 261: try {
1.162 paf 262: if(pa_lock_shared_blocking(f)!=0)
1.289 moko 263: throw Exception("file.lock", &file_spec, "shared lock failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname);
1.123 paf 264:
1.124 paf 265: struct stat finfo;
1.298 moko 266: if(pa_fstat(f, &finfo)!=0)
1.124 paf 267: throw Exception("file.missing", // hardly possible: we just opened it OK
1.289 moko 268: &file_spec, "stat failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname);
1.124 paf 269:
1.149 paf 270: check_safe_mode(finfo, file_spec, fname);
1.32 paf 271:
1.289 moko 272: action(finfo, f, file_spec, context);
1.123 paf 273: } catch(...) {
1.162 paf 274: pa_unlock(f);close(f);
1.123 paf 275: if(fail_on_read_problem)
1.154 paf 276: rethrow;
1.289 moko 277: return false;
1.123 paf 278: }
1.87 paf 279:
1.162 paf 280: pa_unlock(f);close(f);
1.72 parser 281: return true;
1.229 misha 282: } else {
1.118 paf 283: if(fail_on_read_problem)
1.289 moko 284: throw Exception(errno==EACCES ? "file.access" : (errno==ENOENT || errno==ENOTDIR || errno==ENODEV) ? "file.missing" : 0,
285: &file_spec, "%s failed: %s (%d), actual filename '%s'", action_name, strerror(errno), errno, fname);
1.118 paf 286: return false;
287: }
1.8 paf 288: }
289:
1.202 paf 290: void create_dir_for_file(const String& file_spec) {
1.63 parser 291: size_t pos_after=1;
1.154 paf 292: size_t pos_before;
293: while((pos_before=file_spec.pos('/', pos_after))!=STRING_NOT_FOUND) {
1.277 moko 294: pa_mkdir(file_spec.mid(0, pos_before).taint_cstr(String::L_FILE_SPEC), 0775);
1.63 parser 295: pos_after=pos_before+1;
296: }
297: }
298:
1.98 paf 299: bool file_write_action_under_lock(
1.28 paf 300: const String& file_spec,
1.225 misha 301: const char* action_name,
302: File_write_action action,
303: void *context,
1.126 paf 304: bool as_text,
305: bool do_append,
306: bool do_block,
1.110 paf 307: bool fail_on_lock_problem) {
1.247 misha 308: const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC);
1.28 paf 309: int f;
1.80 paf 310: if(access(fname, W_OK)!=0) // no
1.126 paf 311: create_dir_for_file(file_spec);
1.50 paf 312:
1.301 moko 313: if((f=pa_open(fname,
1.80 paf 314: O_CREAT|O_RDWR
315: |(as_text?_O_TEXT:_O_BINARY)
1.138 paf 316: |(do_append?O_APPEND:PA_O_TRUNC), 0664))>=0) {
1.162 paf 317: if((do_block?pa_lock_exclusive_blocking(f):pa_lock_exclusive_nonblocking(f))!=0) {
1.289 moko 318: Exception e("file.lock", &file_spec, "shared lock failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname);
1.126 paf 319: close(f);
1.110 paf 320: if(fail_on_lock_problem)
321: throw e;
1.98 paf 322: return false;
323: }
1.96 paf 324:
1.158 paf 325: try {
1.254 misha 326: #if (defined(HAVE_FCHMOD) && defined(PA_SAFE_MODE))
327: struct stat finfo;
1.298 moko 328: if(pa_fstat(f, &finfo)==0 && finfo.st_mode & 0111)
1.254 misha 329: fchmod(f, finfo.st_mode & 0666/*clear executable bits*/); // backward: ignore errors if any
330: #endif
331: action(f, context);
1.158 paf 332: } catch(...) {
1.138 paf 333: #ifdef HAVE_FTRUNCATE
1.104 paf 334: if(!do_append)
1.125 paf 335: ftruncate(f, lseek(f, 0, SEEK_CUR)); // one can not use O_TRUNC, read lower
1.138 paf 336: #endif
1.162 paf 337: pa_unlock(f);close(f);
1.154 paf 338: rethrow;
1.158 paf 339: }
1.80 paf 340:
1.138 paf 341: #ifdef HAVE_FTRUNCATE
1.104 paf 342: if(!do_append)
1.125 paf 343: ftruncate(f, lseek(f, 0, SEEK_CUR)); // O_TRUNC truncates even exclusevely write-locked file [thanks to Igor Milyakov <virtan@rotabanner.com> for discovering]
1.138 paf 344: #endif
1.162 paf 345: pa_unlock(f);close(f);
1.98 paf 346: return true;
1.80 paf 347: } else
1.289 moko 348: throw Exception(errno==EACCES ? "file.access" : 0, &file_spec, "%s failed: %s (%d), actual filename '%s'", action_name, strerror(errno), errno, fname);
1.96 paf 349: // here should be nothing, see rethrow above
350: }
351:
352: #ifndef DOXYGEN
353: struct File_write_action_info {
1.250 misha 354: const char* str;
355: size_t length;
1.126 paf 356: };
1.96 paf 357: #endif
1.271 moko 358:
1.96 paf 359: static void file_write_action(int f, void *context) {
1.126 paf 360: File_write_action_info& info=*static_cast<File_write_action_info *>(context);
1.154 paf 361: if(info.length) {
1.271 moko 362: ssize_t written=write(f, info.str, info.length);
1.116 paf 363: if(written<0)
1.271 moko 364: throw Exception("file.write", 0, "write failed: %s (%d)", strerror(errno), errno);
1.275 moko 365: if((size_t)written!=info.length)
1.271 moko 366: throw Exception("file.write", 0, "write failed: %u of %u bytes written", written, info.length);
1.113 paf 367: }
1.96 paf 368: }
1.271 moko 369:
1.96 paf 370: void file_write(
1.250 misha 371: Request_charsets& charsets,
372: const String& file_spec,
373: const char* data,
374: size_t size,
1.126 paf 375: bool as_text,
1.250 misha 376: bool do_append,
377: Charset* asked_charset) {
378:
379: if(as_text && asked_charset){
380: String::C body=String::C(data, size);
381: body=Charset::transcode(body, charsets.source(), *asked_charset);
382: data=body.str;
383: size=body.length;
384: };
385:
1.126 paf 386: File_write_action_info info={data, size};
1.225 misha 387:
1.98 paf 388: file_write_action_under_lock(
1.154 paf 389: file_spec,
1.225 misha 390: "write",
391: file_write_action,
392: &info,
1.154 paf 393: as_text,
394: do_append);
1.30 paf 395: }
396:
1.261 misha 397: static size_t get_dir(char* fname, size_t helper_length){
398: bool dir=false;
399: size_t pos=0;
400: for(pos=helper_length; pos; pos--){
401: char c=fname[pos-1];
402: if(c=='/' || c=='\\'){
403: fname[pos-1]=0;
404: dir=true;
405: } else if(dir) break;
406: }
407: return pos;
408: }
409:
410: static bool entry_readable(char* fname, bool need_dir) {
411: if(need_dir){
412: size_t size=strlen(fname);
413: while(size) {
414: char c=fname[size-1];
415: if(c=='/' || c=='\\')
416: fname[--size]=0;
417: else
418: break;
419: }
420: }
421:
422: struct stat finfo;
423: if(access(fname, R_OK)==0 && entry_exists(fname, &finfo)) {
424: bool is_dir=(finfo.st_mode&S_IFDIR) != 0;
425: return is_dir==need_dir;
426: }
427: return false;
428: }
429:
430: static bool entry_readable(const String& file_spec, bool need_dir) {
431: return entry_readable(file_spec.taint_cstrm(String::L_FILE_SPEC), need_dir);
432: }
433:
1.63 parser 434: // throws nothing! [this is required in file_move & file_delete]
1.277 moko 435: static void rmdir(const String& file_spec, size_t pos_after) {
1.261 misha 436: char* dir_spec=file_spec.taint_cstrm(String::L_FILE_SPEC);
437: size_t length=strlen(dir_spec);
438: while( (length=get_dir(dir_spec, length)) && (length > pos_after) ){
1.274 moko 439: #ifdef _MSC_VER
1.261 misha 440: if(!entry_readable(dir_spec, true))
441: break;
442: DWORD attrs=GetFileAttributes(dir_spec);
443: if(
444: (attrs==INVALID_FILE_ATTRIBUTES)
445: || !(attrs & FILE_ATTRIBUTE_DIRECTORY)
446: || (attrs & FILE_ATTRIBUTE_REPARSE_POINT)
447: )
448: break;
449: #endif
450: if( rmdir(dir_spec) )
451: break;
452: };
1.50 paf 453: }
1.239 misha 454:
1.269 misha 455: bool file_delete(const String& file_spec, bool fail_on_problem, bool keep_empty_dirs) {
1.247 misha 456: const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC);
1.282 moko 457: if(unlink(fname)!=0) {
1.164 paf 458: if(fail_on_problem)
1.289 moko 459: throw Exception(errno==EACCES?"file.access":errno==ENOENT?"file.missing":0,
460: &file_spec, "unlink failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname);
1.93 paf 461: else
462: return false;
1.282 moko 463: }
1.50 paf 464:
1.269 misha 465: if(!keep_empty_dirs)
466: rmdir(file_spec, 1);
467:
1.93 paf 468: return true;
1.60 parser 469: }
1.239 misha 470:
1.269 misha 471: void file_move(const String& old_spec, const String& new_spec, bool keep_empty_dirs) {
1.247 misha 472: const char* old_spec_cstr=old_spec.taint_cstr(String::L_FILE_SPEC);
473: const char* new_spec_cstr=new_spec.taint_cstr(String::L_FILE_SPEC);
1.63 parser 474:
1.126 paf 475: create_dir_for_file(new_spec);
1.63 parser 476:
1.60 parser 477: if(rename(old_spec_cstr, new_spec_cstr)!=0)
1.289 moko 478: throw Exception(errno==EACCES ? "file.access" : errno==ENOENT ? "file.missing" : 0,
479: &old_spec, "rename failed: %s (%d), actual filename '%s' to '%s'", strerror(errno), errno, old_spec_cstr, new_spec_cstr);
1.63 parser 480:
1.269 misha 481: if(!keep_empty_dirs)
482: rmdir(old_spec, 1);
1.31 paf 483: }
484:
1.51 paf 485:
1.126 paf 486: bool entry_exists(const char* fname, struct stat *afinfo) {
1.118 paf 487: struct stat lfinfo;
1.298 moko 488: bool result=pa_stat(fname, &lfinfo)==0;
1.118 paf 489: if(afinfo)
490: *afinfo=lfinfo;
491: return result;
1.119 paf 492: }
493:
494: bool entry_exists(const String& file_spec) {
1.247 misha 495: const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC);
1.126 paf 496: return entry_exists(fname, 0);
1.118 paf 497: }
498:
1.215 paf 499: bool file_exist(const String& file_spec) {
1.126 paf 500: return entry_readable(file_spec, false);
1.51 paf 501: }
1.239 misha 502:
1.215 paf 503: bool dir_exists(const String& file_spec) {
1.126 paf 504: return entry_readable(file_spec, true);
1.65 parser 505: }
1.239 misha 506:
1.215 paf 507: const String* file_exist(const String& path, const String& name) {
1.154 paf 508: String& result=*new String(path);
1.270 moko 509: if(path.last_char() != '/')
510: result << "/";
1.154 paf 511: result << name;
1.215 paf 512: return file_exist(result)?&result:0;
1.43 paf 513: }
1.239 misha 514:
1.43 paf 515: bool file_executable(const String& file_spec) {
1.247 misha 516: return access(file_spec.taint_cstr(String::L_FILE_SPEC), X_OK)==0;
1.44 paf 517: }
518:
1.296 moko 519: bool file_stat(const String& file_spec, uint64_t& rsize, time_t& ratime, time_t& rmtime, time_t& rctime, bool fail_on_read_problem) {
1.247 misha 520: const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC);
1.154 paf 521: struct stat finfo;
1.298 moko 522: if(pa_stat(fname, &finfo)!=0) {
1.64 parser 523: if(fail_on_read_problem)
1.289 moko 524: throw Exception("file.missing", &file_spec, "getting file size failed: %s (%d), real filename '%s'", strerror(errno), errno, fname);
1.64 parser 525: else
526: return false;
1.282 moko 527: }
1.58 parser 528: rsize=finfo.st_size;
1.299 moko 529: ratime=(time_t)finfo.st_atime;
530: rmtime=(time_t)finfo.st_mtime;
531: rctime=(time_t)finfo.st_ctime;
1.64 parser 532: return true;
1.18 paf 533: }
534:
1.300 moko 535: size_t check_file_size(uint64_t size, const String& file_spec){
536: if(size > pa_file_size_limit)
537: throw Exception(PARSER_RUNTIME, &file_spec, "content size of %.15g bytes exceeds the limit (%.15g bytes)", (double)size, (double)pa_file_size_limit);
538: return (size_t)size;
539: }
540:
1.278 moko 541: /**
542: String related functions
543: */
544:
545: bool capitalized(const char* s){
546: bool upper=true;
547: for(const char* c=s; *c; c++){
548: if(*c != (upper ? toupper((unsigned char)*c) : tolower((unsigned char)*c)))
549: return false;
550: upper=strchr("-_ ", *c) != 0;
551: }
552: return true;
553: }
554:
555: const char* capitalize(const char* s){
556: if(!s || capitalized(s))
557: return s;
558:
559: char* result=pa_strdup(s);
560: if(result){
561: bool upper=true;
562: for(char* c=result; *c; c++){
563: *c=upper ? (char)toupper((unsigned char)*c) : (char)tolower((unsigned char)*c);
564: upper=strchr("-_ ", *c) != 0;
565: }
566: }
567: return (const char*)result;
568: }
569:
1.290 moko 570: char *str_lower(const char *s, size_t helper_length){
571: char *result=pa_strdup(s, helper_length);
572: for(char* c=result; *c; c++)
573: *c=(char)tolower((unsigned char)*c);
574: return result;
575: }
576:
577: char *str_upper(const char *s, size_t helper_length){
578: char *result=pa_strdup(s, helper_length);
579: for(char* c=result; *c; c++)
580: *c=(char)toupper((unsigned char)*c);
581: return result;
582: }
583:
1.278 moko 584: void fix_line_breaks(char *str, size_t& length) {
585: //_asm int 3;
586: const char* const eob=str+length;
587: char* dest=str;
588: // fix DOS: \r\n -> \n
589: // fix Macintosh: \r -> \n
590: char* bol=str;
591: while(char* eol=(char*)memchr(bol, '\r', eob -bol)) {
592: size_t len=eol-bol;
593: if(dest!=bol)
594: memmove(dest, bol, len);
595: dest+=len;
596: *dest++='\n';
597:
598: if(&eol[1]<eob && eol[1]=='\n') { // \r, \n = DOS
599: bol=eol+2;
600: length--;
601: } else // \r, not \n = Macintosh
602: bol=eol+1;
603: }
604: // last piece without \r
605: if(dest!=bol)
606: memmove(dest, bol, eob-bol);
607: str[length]=0; // terminating
608: }
609:
610: /**
611: scans for @a delim[default \n] in @a *row_ref,
612: @return piece of line before it or end of string, if no @a delim found
613: assigns @a *row_ref to point right after delimiter if there were one
614: or to zero if no @a delim were found.
615: */
616:
1.126 paf 617: char* getrow(char* *row_ref, char delim) {
1.229 misha 618: char* result=*row_ref;
619: if(result) {
1.126 paf 620: *row_ref=strchr(result, delim);
1.8 paf 621: if(*row_ref)
622: *((*row_ref)++)=0;
623: else if(!*result)
624: return 0;
1.229 misha 625: }
626: return result;
1.8 paf 627: }
628:
1.126 paf 629: char* lsplit(char* string, char delim) {
1.229 misha 630: if(string) {
1.126 paf 631: char* v=strchr(string, delim);
1.8 paf 632: if(v) {
633: *v=0;
634: return v+1;
635: }
1.229 misha 636: }
637: return 0;
1.8 paf 638: }
639:
1.126 paf 640: char* lsplit(char* *string_ref, char delim) {
1.229 misha 641: char* result=*string_ref;
1.126 paf 642: char* next=lsplit(*string_ref, delim);
1.229 misha 643: *string_ref=next;
644: return result;
1.9 paf 645: }
646:
1.126 paf 647: char* rsplit(char* string, char delim) {
1.229 misha 648: if(string) {
1.126 paf 649: char* v=strrchr(string, delim);
1.18 paf 650: if(v) {
1.9 paf 651: *v=0;
652: return v+1;
653: }
1.229 misha 654: }
655: return NULL;
1.10 paf 656: }
657:
1.229 misha 658:
659: // format: %[flags][width][.precision]type http://msdn.microsoft.com/ru-ru/library/56e442dc(en-us,VS.80).aspx
660: // flags: '-', '+', ' ', '#', '0' http://msdn.microsoft.com/ru-ru/library/8aky45ct(en-us,VS.80).aspx
661: // width, precision: non negative decimal number
662: enum FormatType {
663: FormatInvalid,
664: FormatInt,
665: FormatUInt,
666: FormatDouble
667: };
1.272 moko 668: FormatType format_type(const char* fmt){
1.229 misha 669: enum FormatState {
670: Percent,
671: Flags,
672: Width,
673: Precision,
674: Done
675: } state=Percent;
676:
677: FormatType result=FormatInvalid;
678:
1.272 moko 679: const char* pos=fmt;
1.229 misha 680: while(char c=*(pos++)){
681: switch(state){
682: case Percent:
683: if(c=='%'){
684: state=Flags;
685: } else {
686: return FormatInvalid; // 1st char must be '%' only
687: }
688: break;
689: case Flags:
690: if(strchr("-+ #0", c)!=0){
691: break;
692: }
693: // go to the next step
694: case Width:
695: if(c=='.'){
696: state=Precision;
697: break;
698: }
699: // go to the next step
700: case Precision:
701: if(c>='0' && c<='9'){
702: if(state == Flags) state=Width; // no more flags
703: break;
704: } else if(c=='d' || c=='i'){
705: result=FormatInt;
706: } else if(strchr("feEgG", c)!=0){
707: result=FormatDouble;
708: } else if(strchr("uoxX", c)!=0){
709: result=FormatUInt;
710: } else {
711: return FormatInvalid; // invalid char
712: }
713: state=Done;
714: break;
715: case Done:
716: return FormatInvalid; // no chars allowed after 'type'
717: }
718: }
719: return result;
720: }
721:
722:
1.272 moko 723: const char* format(double value, const char* fmt) {
1.229 misha 724: char local_buf[MAX_NUMBER];
1.235 misha 725: int size=-1;
1.229 misha 726:
727: if(fmt && strlen(fmt)){
728: switch(format_type(fmt)){
729: case FormatDouble:
730: size=snprintf(local_buf, sizeof(local_buf), fmt, value);
731: break;
732: case FormatInt:
733: size=snprintf(local_buf, sizeof(local_buf), fmt, (int)value);
734: break;
735: case FormatUInt:
1.126 paf 736: size=snprintf(local_buf, sizeof(local_buf), fmt, (uint)value);
1.229 misha 737: break;
738: case FormatInvalid:
1.289 moko 739: throw Exception(PARSER_RUNTIME, 0, "Incorrect format string '%s' was specified.", fmt);
1.229 misha 740: }
741: } else
742: size=snprintf(local_buf, sizeof(local_buf), "%d", (int)value);
743:
744: if(size < 0 || size >= MAX_NUMBER-1){ // on win32 we manually reduce max size while printing
1.304 moko 745: throw Exception(PARSER_RUNTIME, 0, "Error occurred white executing snprintf with format string '%s'.", fmt);
1.229 misha 746: }
747:
1.235 misha 748: return pa_strdup(local_buf, (size_t)size);
1.12 paf 749: }
750:
1.36 paf 751: size_t stdout_write(const void *buf, size_t size) {
1.12 paf 752: #ifdef WIN32
1.187 paf 753: size_t to_write = size;
1.12 paf 754: do{
1.154 paf 755: int chunk_written=fwrite(buf, 1, min((size_t)8*0x400, size), stdout);
1.12 paf 756: if(chunk_written<=0)
757: break;
758: size-=chunk_written;
1.36 paf 759: buf=((const char*)buf)+chunk_written;
1.126 paf 760: } while(size>0);
1.12 paf 761:
1.187 paf 762: return to_write-size;
1.12 paf 763: #else
1.126 paf 764: return fwrite(buf, 1, size, stdout);
1.12 paf 765: #endif
1.2 paf 766: }
1.14 paf 767:
1.229 misha 768: enum EscapeState {
769: EscapeRest,
770: EscapeFirst,
771: EscapeSecond,
772: EscapeUnicode
773: };
774:
1.236 misha 775: // @todo prescan for reduce required size (unescaped sting in 1 byte charset requires less memory usually)
1.258 misha 776: char* unescape_chars(const char* cp, int len, Charset* charset, bool js){
1.236 misha 777: char* s=new(PointerFreeGC) char[len+1]; // must be enough (%uXXXX==6 bytes, max utf-8 char length==6 bytes)
1.230 misha 778: char* dst=s;
1.229 misha 779: EscapeState escapeState=EscapeRest;
780: uint escapedValue=0;
781: int srcPos=0;
1.230 misha 782: short int jsCnt=0;
1.236 misha 783: while(srcPos<len){
1.229 misha 784: uchar c=(uchar)cp[srcPos];
1.258 misha 785: if(c=='%' || (c=='\\' && js)){
1.229 misha 786: escapeState=EscapeFirst;
787: } else {
788: switch(escapeState) {
789: case EscapeRest:
1.286 moko 790: if(c=='+' && !js){
1.230 misha 791: *dst++=' ';
1.229 misha 792: } else {
1.230 misha 793: *dst++=c;
1.229 misha 794: }
795: break;
796: case EscapeFirst:
1.232 misha 797: if(charset && c=='u'){
1.229 misha 798: // escaped unicode value: %u0430
799: jsCnt=0;
800: escapedValue=0;
801: escapeState=EscapeUnicode;
802: } else {
1.231 misha 803: if(isxdigit(c)){
1.229 misha 804: escapedValue=hex_value[c] << 4;
805: escapeState=EscapeSecond;
806: } else {
1.230 misha 807: *dst++=c;
1.229 misha 808: escapeState=EscapeRest;
809: }
810: }
811: break;
812: case EscapeSecond:
1.231 misha 813: if(isxdigit(c)){
1.229 misha 814: escapedValue+=hex_value[c];
1.230 misha 815: *dst++=(char)escapedValue;
1.229 misha 816: }
817: escapeState=EscapeRest;
818: break;
819: case EscapeUnicode:
1.231 misha 820: if(isxdigit(c)){
1.229 misha 821: escapedValue=(escapedValue << 4) + hex_value[c];
822: if(++jsCnt==4){
1.230 misha 823: // transcode utf8 char to client charset (we can lost some chars here)
1.232 misha 824: charset->store_Char((XMLByte*&)dst, (XMLCh)escapedValue, '?');
1.229 misha 825: escapeState=EscapeRest;
826: }
827: } else {
828: // not full unicode value
829: escapeState=EscapeRest;
830: }
831: break;
832: }
833: }
834:
835: srcPos++;
836: }
837:
1.230 misha 838: *dst=0; // zero-termination
1.229 misha 839: return s;
840: }
1.24 paf 841:
1.268 misha 842: char *search_stop(char*& current, char cstop_at) {
843: // sanity check
844: if(!current)
845: return 0;
846:
847: // skip leading WS
848: while(*current==' ' || *current=='\t')
849: current++;
850: if(!*current)
851: return current=0;
852:
853: char *result=current;
854: if(char *pstop_at=strchr(current, cstop_at)) {
855: *pstop_at=0;
856: current=pstop_at+1;
857: } else
858: current=0;
859: return result;
860: }
861:
1.24 paf 862: #ifdef WIN32
1.126 paf 863: void back_slashes_to_slashes(char* s) {
1.24 paf 864: if(s)
865: for(; *s; s++)
866: if(*s=='\\')
1.126 paf 867: *s='/';
1.24 paf 868: }
869: #endif
1.41 paf 870:
1.232 misha 871: size_t strpos(const char *str, const char *substr) {
872: const char *p = strstr(str, substr);
873: return (p==0)?STRING_NOT_FOUND:p-str;
874: }
875:
1.226 misha 876: int remove_crlf(char* start, char* end) {
877: char* from=start;
878: char* to=start;
879: bool skip=false;
880: while(from < end){
881: switch(*from){
882: case '\n':
883: case '\r':
884: case '\t':
885: case ' ':
886: if(!skip){
887: *to=' ';
888: to++;
889: skip=true;
890: }
891: break;
892: default:
893: if(from != to)
894: *to=*from;
895: to++;
896: skip=false;
1.69 parser 897: }
1.226 misha 898: from++;
899: }
900: return to-start;
1.91 paf 901: }
902:
1.279 moko 903: const char* hex_digits="0123456789ABCDEF";
904:
1.278 moko 905: const char* hex_string(unsigned char* bytes, size_t size, bool upcase) {
906: char *bytes_hex=new(PointerFreeGC) char [size*2/*byte->hh*/+1/*for zero-teminator*/];
907: unsigned char *src=bytes;
908: unsigned char *end=bytes+size;
909: char *dest=bytes_hex;
910:
1.279 moko 911: const char *hex=upcase? hex_digits : "0123456789abcdef";
1.278 moko 912:
913: for(; src<end; src++) {
914: *dest++=hex[*src/0x10];
915: *dest++=hex[*src%0x10];
916: }
917: *dest=0;
918:
919: return bytes_hex;
920: }
1.91 paf 921:
1.221 misha 922: int file_block_read(const int f, unsigned char* buffer, const size_t size){
923: int nCount = read(f, buffer, size);
924: if (nCount < 0)
1.289 moko 925: throw Exception("file.read", 0, "read failed: %s (%d)", strerror(errno), errno);
1.221 misha 926: return nCount;
927: }
928:
1.278 moko 929: static unsigned long crc32Table[256];
930: static void InitCrc32Table()
931: {
932: if(crc32Table[1] == 0){
933: // This is the official polynomial used by CRC32 in PKZip.
934: // Often times the polynomial shown reversed as 0x04C11DB7.
935: static const unsigned long dwPolynomial = 0xEDB88320;
936:
937: for(int i = 0; i < 256; i++)
938: {
939: unsigned long dwCrc = i;
940: for(int j = 8; j > 0; j--)
941: {
942: if(dwCrc & 1)
943: dwCrc = (dwCrc >> 1) ^ dwPolynomial;
944: else
945: dwCrc >>= 1;
946: }
947: crc32Table[i] = dwCrc;
948: }
949: }
950: }
951:
952: inline void CalcCrc32(const unsigned char byte, unsigned long &crc32)
953: {
954: crc32 = ((crc32) >> 8) ^ crc32Table[(byte) ^ ((crc32) & 0x000000FF)];
955: }
956:
957:
1.297 moko 958: unsigned long pa_crc32(const char *in, size_t in_size){
1.218 misha 959: unsigned long crc32=0xFFFFFFFF;
1.220 misha 960:
1.240 misha 961: InitCrc32Table();
1.239 misha 962: for(size_t i = 0; i<in_size; i++)
963: CalcCrc32(in[i], crc32);
1.220 misha 964:
1.218 misha 965: return ~crc32;
966: }
967:
1.289 moko 968: static void file_crc32_file_action(struct stat& finfo, int f, const String&, void *context) {
1.218 misha 969: unsigned long& crc32=*static_cast<unsigned long *>(context);
970: if(finfo.st_size) {
971: InitCrc32Table();
1.220 misha 972: int nCount=0;
1.218 misha 973: do {
1.221 misha 974: unsigned char buffer[FILE_BUFFER_SIZE];
975: nCount = file_block_read(f, buffer, sizeof(buffer));
1.220 misha 976: for(int i = 0; i < nCount; i++) CalcCrc32(buffer[i], crc32);
977: } while(nCount > 0);
1.218 misha 978: }
979: }
980:
1.297 moko 981: unsigned long pa_crc32(const String& file_spec){
1.278 moko 982: unsigned long crc32=0xFFFFFFFF;
983: file_read_action_under_lock(file_spec, "crc32", file_crc32_file_action, &crc32);
984: return ~crc32;
985: }
986:
987: // content-type: xxx; charset=WE-NEED-THIS
988: // content-type: xxx; charset="WE-NEED-THIS"
989: // content-type: xxx; charset="WE-NEED-THIS";
990: Charset* detect_charset(const char* content_type){
991: if(content_type){
1.291 moko 992: char* CONTENT_TYPE=str_upper(content_type);
1.278 moko 993:
994: if(const char* begin=strstr(CONTENT_TYPE, "CHARSET=")){
995: begin+=8; // skip "CHARSET="
996: char* end=0;
997: if(*begin && (*begin=='"' || *begin =='\'')){
998: char quote=*begin;
999: begin++;
1000: end=(char*)strchr(begin, quote);
1001: }
1002: if(!end)
1003: end=(char*)strchr(begin, ';');
1004:
1005: if(end)
1006: *end=0; // terminator
1007:
1.295 moko 1008: return *begin ? &pa_charsets.get_direct(begin) : 0;
1.278 moko 1009: }
1010: }
1011: return 0;
1012: }
1013:
1.301 moko 1014: const UTF16* pa_utf16_encode(const char* in, Charset& source_charset){
1.302 moko 1015: if(!in)
1016: return 0;
1.301 moko 1017:
1.302 moko 1018: String::C sIn(in,strlen(in));
1.301 moko 1019:
1.302 moko 1020: UTF16* utf16=(UTF16*)pa_malloc_atomic(sIn.length*2+2);
1021: UTF16* utf16_end=utf16;
1.301 moko 1022:
1.302 moko 1023: if(!source_charset.isUTF8())
1024: sIn=Charset::transcode(sIn, source_charset, pa_UTF8_charset);
1.301 moko 1025:
1.302 moko 1026: int status=pa_convertUTF8toUTF16((const UTF8**)&sIn.str, (const UTF8*)(sIn.str+sIn.length), &utf16_end, utf16+sIn.length, strictConversion);
1027: if(status != conversionOK)
1028: throw Exception("utf-16 encode", new String(in), "utf-16 conversion failed (%d)", status);
1.301 moko 1029:
1.302 moko 1030: *utf16_end=0;
1.301 moko 1031:
1.302 moko 1032: return utf16;
1.301 moko 1033: }
1034:
1035: const char* pa_utf16_decode(const UTF16* in, Charset& asked_charset){
1036: if(!in)
1037: return 0;
1038:
1039: const UTF16* utf16_start=in;
1040: const UTF16* utf16_end;
1041:
1042: for(utf16_end=in; *utf16_end; utf16_end++);
1043:
1044: char *result = (char *)pa_malloc_atomic((utf16_end-in)*6+1);
1045: char *result_end = result;
1046:
1047: int status=pa_convertUTF16toUTF8(&utf16_start, utf16_end, (UTF8**)&result_end, (UTF8*)(result+(utf16_end-in)*6), strictConversion);
1048:
1049: if(status != conversionOK)
1050: throw Exception("utf-16 decode", 0, "utf conversion failed (%d)", status);
1051:
1052: *result_end='\0';
1053:
1.302 moko 1054: if(asked_charset.isUTF8())
1055: return result;
1.301 moko 1056:
1.302 moko 1057: return Charset::transcode(result, pa_UTF8_charset, asked_charset).cstr();
1.301 moko 1058: }
1.278 moko 1059:
1.283 moko 1060: static bool is_latin(const char *in){
1061: for(; *in; in++){
1062: if ((unsigned char)(*in) > 0x7F)
1063: return false;
1064: }
1065: return true;
1066: }
1067:
1068: #define MAX_IDNA_LENGTH 256
1069:
1.301 moko 1070: const char *pa_idna_encode(const char *in, Charset& source_charset){
1.283 moko 1071: if(!in || is_latin(in))
1072: return in;
1073:
1074: uint32_t utf32[MAX_IDNA_LENGTH];
1075: uint32_t *utf32_end=utf32;
1076:
1077: String::C sIn(in,strlen(in));
1078:
1079: if(!source_charset.isUTF8())
1.295 moko 1080: sIn=Charset::transcode(sIn, source_charset, pa_UTF8_charset);
1.283 moko 1081:
1082: int status=pa_convertUTF8toUTF32((const UTF8**)&sIn.str, (const UTF8*)(sIn.str+sIn.length), &utf32_end, utf32+MAX_IDNA_LENGTH-1, strictConversion);
1083: if(status != conversionOK)
1084: throw Exception("idna encode", new String(in), "utf conversion failed (%d)", status);
1085:
1086: *utf32_end=0;
1087:
1088: char *result = (char *)pa_malloc(MAX_IDNA_LENGTH);
1089: status=pa_idna_to_ascii_4z(utf32, result, MAX_IDNA_LENGTH, 0);
1090: if(status != IDNA_SUCCESS)
1091: throw Exception("idna encode", new String(in), "encode failed: %s", pa_idna_strerror(status));
1092:
1093: return result;
1094: }
1095:
1096: const char *pa_idna_decode(const char *in, Charset &asked_charset){
1097: if(!in || !(*in))
1098: return in;
1099:
1100: uint32_t utf32[MAX_IDNA_LENGTH];
1101: const uint32_t *utf32_start=utf32;
1102: uint32_t *utf32_end;
1103:
1104: int status=pa_idna_to_unicode_4z(in, utf32, MAX_IDNA_LENGTH, 0);
1105: if(status != IDNA_SUCCESS)
1106: throw Exception("idna decode", new String(in), "decode failed: %s", pa_idna_strerror(status));
1107:
1108: for(utf32_end=utf32; *utf32_end; utf32_end++);
1109:
1110: char *result = (char *)pa_malloc(MAX_IDNA_LENGTH);
1111: char *result_end = result;
1112:
1113: status=pa_convertUTF32toUTF8(&utf32_start, utf32_end, (UTF8**)&result_end, (UTF8*)(result+MAX_IDNA_LENGTH-1), strictConversion);
1114:
1115: if(status != conversionOK)
1116: throw Exception("idna decode", new String(in), "utf conversion failed (%d)", status);
1117:
1118: *result_end='\0';
1119:
1120: if(!asked_charset.isUTF8())
1.295 moko 1121: result = (char *)Charset::transcode(result, pa_UTF8_charset, asked_charset).cstr();
1.283 moko 1122:
1123: return result;
1124: }
1.292 moko 1125: /// must be last in this file
1126: #undef vsnprintf
1127: int pa_vsnprintf(char* b, size_t s, const char* f, va_list l) {
1128: if(!s)
1129: return 0;
1130:
1131: int r;
1132: // note: on win32 & maybe somewhere else
1133: // vsnprintf do not writes terminating 0 in 'buffer full' case, reducing
1134: // http://stackoverflow.com/questions/2915672/snprintf-and-visual-studio-2010
1135: --s;
1136:
1137: // clients do not check for negative 's', feature: ignore such prints
1138: if((ssize_t)s<0)
1139: return 0;
1140:
1141: #ifdef _MSC_VER
1142: // win32: if the number of bytes to write exceeds buffer, then count bytes are written and -1 is returned
1143: r=_vsnprintf(b, s, f, l);
1144: if(r<0)
1145: r=s;
1146: #else
1147: r=vsnprintf(b, s, f, l);
1148: /*
1149: solaris: man vsnprintf
1150:
1151: The snprintf() function returns the number of characters
1152: formatted, that is, the number of characters that would have
1153: been written to the buffer if it were large enough. If the
1154: value of n is 0 on a call to snprintf(), an unspecified
1155: value less than 1 is returned.
1156: */
1157:
1158: if(r<0)
1159: r=0;
1160: else if((size_t)r>s)
1161: r=s;
1162: #endif
1163: b[r]=0;
1164: return r;
1165: }
1166:
1167: int pa_snprintf(char* b, size_t s, const char* f, ...) {
1168: va_list l;
1169: va_start(l, f);
1170: int r=pa_vsnprintf(b, s, f, l);
1171: va_end(l);
1172: return r;
1173: }
1174:
E-mail: