1 //===- Support/FileUtilities.cpp - File System Utilities ------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file was developed by the LLVM research group and is distributed under
6 // the University of Illinois Open Source License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements a family of utility functions which are useful for doing
11 // various things with files.
13 //===----------------------------------------------------------------------===//
15 #include "llvm/Support/FileUtilities.h"
16 #include "llvm/System/Path.h"
17 #include "llvm/System/MappedFile.h"
18 #include "llvm/ADT/StringExtras.h"
23 static bool isNumberChar(char C) {
25 case '0': case '1': case '2': case '3': case '4':
26 case '5': case '6': case '7': case '8': case '9':
27 case '.': case '+': case '-':
28 case 'D': // Strange exponential notation.
29 case 'd': // Strange exponential notation.
31 case 'E': return true;
32 default: return false;
36 static char *BackupNumber(char *Pos, char *FirstChar) {
37 // If we didn't stop in the middle of a number, don't backup.
38 if (!isNumberChar(*Pos)) return Pos;
40 // Otherwise, return to the start of the number.
41 while (Pos > FirstChar && isNumberChar(Pos[-1]))
46 /// CompareNumbers - compare two numbers, returning true if they are different.
47 static bool CompareNumbers(char *&F1P, char *&F2P, char *F1End, char *F2End,
48 double AbsTolerance, double RelTolerance,
49 std::string *ErrorMsg) {
50 char *F1NumEnd, *F2NumEnd;
51 double V1 = 0.0, V2 = 0.0;
53 // If one of the positions is at a space and the other isn't, chomp up 'til
54 // the end of the space.
55 while (isspace(*F1P) && F1P != F1End)
57 while (isspace(*F2P) && F2P != F2End)
60 // If we stop on numbers, compare their difference. Note that some ugliness
61 // is built into this to permit support for numbers that use "D" or "d" as
62 // their exponential marker, e.g. "1.234D45". This occurs in 200.sixtrack in
64 if (isNumberChar(*F1P) && isNumberChar(*F2P)) {
68 V1 = strtod(F1P, &F1NumEnd);
69 V2 = strtod(F2P, &F2NumEnd);
71 if (*F1NumEnd == 'D' || *F1NumEnd == 'd') {
72 *F1NumEnd = 'e'; // Strange exponential notation!
75 if (*F2NumEnd == 'D' || *F2NumEnd == 'd') {
76 *F2NumEnd = 'e'; // Strange exponential notation!
79 } while (isDNotation);
81 // Otherwise, the diff failed.
86 if (F1NumEnd == F1P || F2NumEnd == F2P) {
88 *ErrorMsg = "FP Comparison failed, not a numeric difference between '";
90 *ErrorMsg += "' and '";
97 // Check to see if these are inside the absolute tolerance
98 if (AbsTolerance < std::abs(V1-V2)) {
99 // Nope, check the relative tolerance...
102 Diff = std::abs(V1/V2 - 1.0);
104 Diff = std::abs(V2/V1 - 1.0);
106 Diff = 0; // Both zero.
107 if (Diff > RelTolerance) {
109 *ErrorMsg = "Compared: " + ftostr(V1) + " and " + ftostr(V2) +
110 ": diff = " + ftostr(Diff) + "\n";
111 *ErrorMsg += "Out of tolerance: rel/abs: " + ftostr(RelTolerance) +
112 "/" + ftostr(AbsTolerance);
118 // Otherwise, advance our read pointers to the end of the numbers.
119 F1P = F1NumEnd; F2P = F2NumEnd;
123 // PadFileIfNeeded - If the files are not identical, we will have to be doing
124 // numeric comparisons in here. There are bad cases involved where we (i.e.,
125 // strtod) might run off the beginning or end of the file if it starts or ends
126 // with a number. Because of this, if needed, we pad the file so that it starts
127 // and ends with a null character.
128 static void PadFileIfNeeded(char *&FileStart, char *&FileEnd, char *&FP) {
129 if (FileStart-FileEnd < 2 ||
130 isNumberChar(FileStart[0]) || isNumberChar(FileEnd[-1])) {
131 unsigned FileLen = FileEnd-FileStart;
132 char *NewFile = new char[FileLen+2];
133 NewFile[0] = 0; // Add null padding
134 NewFile[FileLen+1] = 0; // Add null padding
135 memcpy(NewFile+1, FileStart, FileLen);
136 FP = NewFile+(FP-FileStart)+1;
137 FileStart = NewFile+1;
138 FileEnd = FileStart+FileLen;
142 /// DiffFilesWithTolerance - Compare the two files specified, returning 0 if the
143 /// files match, 1 if they are different, and 2 if there is a file error. This
144 /// function differs from DiffFiles in that you can specify an absolete and
145 /// relative FP error that is allowed to exist. If you specify a string to fill
146 /// in for the error option, it will set the string to an error message if an
147 /// error occurs, allowing the caller to distinguish between a failed diff and a
148 /// file system error.
150 int llvm::DiffFilesWithTolerance(const sys::Path &FileA,
151 const sys::Path &FileB,
152 double AbsTol, double RelTol,
153 std::string *Error) {
154 sys::FileStatus FileAStat, FileBStat;
155 if (FileA.getFileStatus(FileAStat, Error))
157 if (FileB.getFileStatus(FileBStat, Error))
160 // Check for zero length files because some systems croak when you try to
161 // mmap an empty file.
162 size_t A_size = FileAStat.getSize();
163 size_t B_size = FileBStat.getSize();
165 // If they are both zero sized then they're the same
166 if (A_size == 0 && B_size == 0)
169 // If only one of them is zero sized then they can't be the same
170 if ((A_size == 0 || B_size == 0)) {
172 *Error = "Files differ: one is zero-sized, the other isn't";
176 // Now its safe to mmap the files into memory becasue both files
177 // have a non-zero size.
179 if (F1.open(FileA, sys::MappedFile::READ_ACCESS, Error))
182 if (F2.open(FileB, sys::MappedFile::READ_ACCESS, Error))
189 // Okay, now that we opened the files, scan them for the first difference.
190 char *File1Start = F1.charBase();
191 char *File2Start = F2.charBase();
192 char *File1End = File1Start+A_size;
193 char *File2End = File2Start+B_size;
194 char *F1P = File1Start;
195 char *F2P = File2Start;
197 if (A_size == B_size) {
198 // Are the buffers identical?
199 if (std::memcmp(File1Start, File2Start, A_size) == 0)
202 if (AbsTol == 0 && RelTol == 0) {
204 *Error = "Files differ without tolerance allowance";
205 return 1; // Files different!
209 char *OrigFile1Start = File1Start;
210 char *OrigFile2Start = File2Start;
212 // If the files need padding, do so now.
213 PadFileIfNeeded(File1Start, File1End, F1P);
214 PadFileIfNeeded(File2Start, File2End, F2P);
216 bool CompareFailed = false;
218 // Scan for the end of file or next difference.
219 while (F1P < File1End && F2P < File2End && *F1P == *F2P)
222 if (F1P >= File1End || F2P >= File2End) break;
224 // Okay, we must have found a difference. Backup to the start of the
225 // current number each stream is at so that we can compare from the
227 F1P = BackupNumber(F1P, File1Start);
228 F2P = BackupNumber(F2P, File2Start);
230 // Now that we are at the start of the numbers, compare them, exiting if
232 if (CompareNumbers(F1P, F2P, File1End, File2End, AbsTol, RelTol, Error)) {
233 CompareFailed = true;
238 // Okay, we reached the end of file. If both files are at the end, we
240 bool F1AtEnd = F1P >= File1End;
241 bool F2AtEnd = F2P >= File2End;
242 if (!CompareFailed && (!F1AtEnd || !F2AtEnd)) {
243 // Else, we might have run off the end due to a number: backup and retry.
244 if (F1AtEnd && isNumberChar(F1P[-1])) --F1P;
245 if (F2AtEnd && isNumberChar(F2P[-1])) --F2P;
246 F1P = BackupNumber(F1P, File1Start);
247 F2P = BackupNumber(F2P, File2Start);
249 // Now that we are at the start of the numbers, compare them, exiting if
251 if (CompareNumbers(F1P, F2P, File1End, File2End, AbsTol, RelTol, Error))
252 CompareFailed = true;
254 // If we found the end, we succeeded.
255 if (F1P < File1End || F2P < File2End)
256 CompareFailed = true;
259 if (OrigFile1Start != File1Start)
260 delete[] (File1Start-1); // Back up past null byte
261 if (OrigFile2Start != File2Start)
262 delete[] (File2Start-1); // Back up past null byte
263 return CompareFailed;