LLVM API Documentation
00001 //===- Support/FileUtilities.cpp - File System Utilities ------------------===// 00002 // 00003 // The LLVM Compiler Infrastructure 00004 // 00005 // This file is distributed under the University of Illinois Open Source 00006 // License. See LICENSE.TXT for details. 00007 // 00008 //===----------------------------------------------------------------------===// 00009 // 00010 // This file implements a family of utility functions which are useful for doing 00011 // various things with files. 00012 // 00013 //===----------------------------------------------------------------------===// 00014 00015 #include "llvm/Support/FileUtilities.h" 00016 #include "llvm/System/Path.h" 00017 #include "llvm/Support/MemoryBuffer.h" 00018 #include "llvm/ADT/OwningPtr.h" 00019 #include "llvm/ADT/SmallString.h" 00020 #include "llvm/ADT/StringExtras.h" 00021 #include <cstdlib> 00022 #include <cstring> 00023 #include <cctype> 00024 using namespace llvm; 00025 00026 static bool isSignedChar(char C) { 00027 return (C == '+' || C == '-'); 00028 } 00029 00030 static bool isExponentChar(char C) { 00031 switch (C) { 00032 case 'D': // Strange exponential notation. 00033 case 'd': // Strange exponential notation. 00034 case 'e': 00035 case 'E': return true; 00036 default: return false; 00037 } 00038 } 00039 00040 static bool isNumberChar(char C) { 00041 switch (C) { 00042 case '0': case '1': case '2': case '3': case '4': 00043 case '5': case '6': case '7': case '8': case '9': 00044 case '.': return true; 00045 default: return isSignedChar(C) || isExponentChar(C); 00046 } 00047 } 00048 00049 static const char *BackupNumber(const char *Pos, const char *FirstChar) { 00050 // If we didn't stop in the middle of a number, don't backup. 00051 if (!isNumberChar(*Pos)) return Pos; 00052 00053 // Otherwise, return to the start of the number. 00054 while (Pos > FirstChar && isNumberChar(Pos[-1])) { 00055 --Pos; 00056 if (Pos > FirstChar && isSignedChar(Pos[0]) && !isExponentChar(Pos[-1])) 00057 break; 00058 } 00059 return Pos; 00060 } 00061 00062 /// EndOfNumber - Return the first character that is not part of the specified 00063 /// number. This assumes that the buffer is null terminated, so it won't fall 00064 /// off the end. 00065 static const char *EndOfNumber(const char *Pos) { 00066 while (isNumberChar(*Pos)) 00067 ++Pos; 00068 return Pos; 00069 } 00070 00071 /// CompareNumbers - compare two numbers, returning true if they are different. 00072 static bool CompareNumbers(const char *&F1P, const char *&F2P, 00073 const char *F1End, const char *F2End, 00074 double AbsTolerance, double RelTolerance, 00075 std::string *ErrorMsg) { 00076 const char *F1NumEnd, *F2NumEnd; 00077 double V1 = 0.0, V2 = 0.0; 00078 00079 // If one of the positions is at a space and the other isn't, chomp up 'til 00080 // the end of the space. 00081 while (isspace(*F1P) && F1P != F1End) 00082 ++F1P; 00083 while (isspace(*F2P) && F2P != F2End) 00084 ++F2P; 00085 00086 // If we stop on numbers, compare their difference. 00087 if (!isNumberChar(*F1P) || !isNumberChar(*F2P)) { 00088 // The diff failed. 00089 F1NumEnd = F1P; 00090 F2NumEnd = F2P; 00091 } else { 00092 // Note that some ugliness is built into this to permit support for numbers 00093 // that use "D" or "d" as their exponential marker, e.g. "1.234D45". This 00094 // occurs in 200.sixtrack in spec2k. 00095 V1 = strtod(F1P, const_cast<char**>(&F1NumEnd)); 00096 V2 = strtod(F2P, const_cast<char**>(&F2NumEnd)); 00097 00098 if (*F1NumEnd == 'D' || *F1NumEnd == 'd') { 00099 // Copy string into tmp buffer to replace the 'D' with an 'e'. 00100 SmallString<200> StrTmp(F1P, EndOfNumber(F1NumEnd)+1); 00101 // Strange exponential notation! 00102 StrTmp[static_cast<unsigned>(F1NumEnd-F1P)] = 'e'; 00103 00104 V1 = strtod(&StrTmp[0], const_cast<char**>(&F1NumEnd)); 00105 F1NumEnd = F1P + (F1NumEnd-&StrTmp[0]); 00106 } 00107 00108 if (*F2NumEnd == 'D' || *F2NumEnd == 'd') { 00109 // Copy string into tmp buffer to replace the 'D' with an 'e'. 00110 SmallString<200> StrTmp(F2P, EndOfNumber(F2NumEnd)+1); 00111 // Strange exponential notation! 00112 StrTmp[static_cast<unsigned>(F2NumEnd-F2P)] = 'e'; 00113 00114 V2 = strtod(&StrTmp[0], const_cast<char**>(&F2NumEnd)); 00115 F2NumEnd = F2P + (F2NumEnd-&StrTmp[0]); 00116 } 00117 } 00118 00119 if (F1NumEnd == F1P || F2NumEnd == F2P) { 00120 if (ErrorMsg) { 00121 *ErrorMsg = "FP Comparison failed, not a numeric difference between '"; 00122 *ErrorMsg += F1P[0]; 00123 *ErrorMsg += "' and '"; 00124 *ErrorMsg += F2P[0]; 00125 *ErrorMsg += "'"; 00126 } 00127 return true; 00128 } 00129 00130 // Check to see if these are inside the absolute tolerance 00131 if (AbsTolerance < std::abs(V1-V2)) { 00132 // Nope, check the relative tolerance... 00133 double Diff; 00134 if (V2) 00135 Diff = std::abs(V1/V2 - 1.0); 00136 else if (V1) 00137 Diff = std::abs(V2/V1 - 1.0); 00138 else 00139 Diff = 0; // Both zero. 00140 if (Diff > RelTolerance) { 00141 if (ErrorMsg) { 00142 *ErrorMsg = "Compared: " + ftostr(V1) + " and " + ftostr(V2) + "\n"; 00143 *ErrorMsg += "abs. diff = " + ftostr(std::abs(V1-V2)) + 00144 " rel.diff = " + ftostr(Diff) + "\n"; 00145 *ErrorMsg += "Out of tolerance: rel/abs: " + ftostr(RelTolerance) + 00146 "/" + ftostr(AbsTolerance); 00147 } 00148 return true; 00149 } 00150 } 00151 00152 // Otherwise, advance our read pointers to the end of the numbers. 00153 F1P = F1NumEnd; F2P = F2NumEnd; 00154 return false; 00155 } 00156 00157 /// DiffFilesWithTolerance - Compare the two files specified, returning 0 if the 00158 /// files match, 1 if they are different, and 2 if there is a file error. This 00159 /// function differs from DiffFiles in that you can specify an absolete and 00160 /// relative FP error that is allowed to exist. If you specify a string to fill 00161 /// in for the error option, it will set the string to an error message if an 00162 /// error occurs, allowing the caller to distinguish between a failed diff and a 00163 /// file system error. 00164 /// 00165 int llvm::DiffFilesWithTolerance(const sys::PathWithStatus &FileA, 00166 const sys::PathWithStatus &FileB, 00167 double AbsTol, double RelTol, 00168 std::string *Error) { 00169 const sys::FileStatus *FileAStat = FileA.getFileStatus(false, Error); 00170 if (!FileAStat) 00171 return 2; 00172 const sys::FileStatus *FileBStat = FileB.getFileStatus(false, Error); 00173 if (!FileBStat) 00174 return 2; 00175 00176 // Check for zero length files because some systems croak when you try to 00177 // mmap an empty file. 00178 size_t A_size = FileAStat->getSize(); 00179 size_t B_size = FileBStat->getSize(); 00180 00181 // If they are both zero sized then they're the same 00182 if (A_size == 0 && B_size == 0) 00183 return 0; 00184 00185 // If only one of them is zero sized then they can't be the same 00186 if ((A_size == 0 || B_size == 0)) { 00187 if (Error) 00188 *Error = "Files differ: one is zero-sized, the other isn't"; 00189 return 1; 00190 } 00191 00192 // Now its safe to mmap the files into memory becasue both files 00193 // have a non-zero size. 00194 OwningPtr<MemoryBuffer> F1(MemoryBuffer::getFile(FileA.c_str(), Error)); 00195 OwningPtr<MemoryBuffer> F2(MemoryBuffer::getFile(FileB.c_str(), Error)); 00196 if (F1 == 0 || F2 == 0) 00197 return 2; 00198 00199 // Okay, now that we opened the files, scan them for the first difference. 00200 const char *File1Start = F1->getBufferStart(); 00201 const char *File2Start = F2->getBufferStart(); 00202 const char *File1End = F1->getBufferEnd(); 00203 const char *File2End = F2->getBufferEnd(); 00204 const char *F1P = File1Start; 00205 const char *F2P = File2Start; 00206 00207 if (A_size == B_size) { 00208 // Are the buffers identical? Common case: Handle this efficiently. 00209 if (std::memcmp(File1Start, File2Start, A_size) == 0) 00210 return 0; 00211 00212 if (AbsTol == 0 && RelTol == 0) { 00213 if (Error) 00214 *Error = "Files differ without tolerance allowance"; 00215 return 1; // Files different! 00216 } 00217 } 00218 00219 bool CompareFailed = false; 00220 while (1) { 00221 // Scan for the end of file or next difference. 00222 while (F1P < File1End && F2P < File2End && *F1P == *F2P) 00223 ++F1P, ++F2P; 00224 00225 if (F1P >= File1End || F2P >= File2End) break; 00226 00227 // Okay, we must have found a difference. Backup to the start of the 00228 // current number each stream is at so that we can compare from the 00229 // beginning. 00230 F1P = BackupNumber(F1P, File1Start); 00231 F2P = BackupNumber(F2P, File2Start); 00232 00233 // Now that we are at the start of the numbers, compare them, exiting if 00234 // they don't match. 00235 if (CompareNumbers(F1P, F2P, File1End, File2End, AbsTol, RelTol, Error)) { 00236 CompareFailed = true; 00237 break; 00238 } 00239 } 00240 00241 // Okay, we reached the end of file. If both files are at the end, we 00242 // succeeded. 00243 bool F1AtEnd = F1P >= File1End; 00244 bool F2AtEnd = F2P >= File2End; 00245 if (!CompareFailed && (!F1AtEnd || !F2AtEnd)) { 00246 // Else, we might have run off the end due to a number: backup and retry. 00247 if (F1AtEnd && isNumberChar(F1P[-1])) --F1P; 00248 if (F2AtEnd && isNumberChar(F2P[-1])) --F2P; 00249 F1P = BackupNumber(F1P, File1Start); 00250 F2P = BackupNumber(F2P, File2Start); 00251 00252 // Now that we are at the start of the numbers, compare them, exiting if 00253 // they don't match. 00254 if (CompareNumbers(F1P, F2P, File1End, File2End, AbsTol, RelTol, Error)) 00255 CompareFailed = true; 00256 00257 // If we found the end, we succeeded. 00258 if (F1P < File1End || F2P < File2End) 00259 CompareFailed = true; 00260 } 00261 00262 return CompareFailed; 00263 }
This web site is hosted by the Computer Science Department at the University of Illinois at Urbana-Champaign.