1 //===- Support/FileUtilities.cpp - File System Utilities ------------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements a family of utility functions which are useful for doing
11 // various things with files.
12 //
13 //===----------------------------------------------------------------------===//
14
15 #include "llvm/Support/FileUtilities.h"
16 #include "llvm/Support/MemoryBuffer.h"
17 #include "llvm/Support/raw_ostream.h"
18 #include "llvm/Support/Path.h"
19 #include "llvm/Support/system_error.h"
20 #include "llvm/ADT/OwningPtr.h"
21 #include "llvm/ADT/SmallString.h"
22 #include <cstdlib>
23 #include <cstring>
24 #include <cctype>
25 using namespace llvm;
26
isSignedChar(char C)27 static bool isSignedChar(char C) {
28 return (C == '+' || C == '-');
29 }
30
isExponentChar(char C)31 static bool isExponentChar(char C) {
32 switch (C) {
33 case 'D': // Strange exponential notation.
34 case 'd': // Strange exponential notation.
35 case 'e':
36 case 'E': return true;
37 default: return false;
38 }
39 }
40
isNumberChar(char C)41 static bool isNumberChar(char C) {
42 switch (C) {
43 case '0': case '1': case '2': case '3': case '4':
44 case '5': case '6': case '7': case '8': case '9':
45 case '.': return true;
46 default: return isSignedChar(C) || isExponentChar(C);
47 }
48 }
49
BackupNumber(const char * Pos,const char * FirstChar)50 static const char *BackupNumber(const char *Pos, const char *FirstChar) {
51 // If we didn't stop in the middle of a number, don't backup.
52 if (!isNumberChar(*Pos)) return Pos;
53
54 // Otherwise, return to the start of the number.
55 bool HasPeriod = false;
56 while (Pos > FirstChar && isNumberChar(Pos[-1])) {
57 // Backup over at most one period.
58 if (Pos[-1] == '.') {
59 if (HasPeriod)
60 break;
61 HasPeriod = true;
62 }
63
64 --Pos;
65 if (Pos > FirstChar && isSignedChar(Pos[0]) && !isExponentChar(Pos[-1]))
66 break;
67 }
68 return Pos;
69 }
70
71 /// EndOfNumber - Return the first character that is not part of the specified
72 /// number. This assumes that the buffer is null terminated, so it won't fall
73 /// off the end.
EndOfNumber(const char * Pos)74 static const char *EndOfNumber(const char *Pos) {
75 while (isNumberChar(*Pos))
76 ++Pos;
77 return Pos;
78 }
79
80 /// CompareNumbers - compare two numbers, returning true if they are different.
CompareNumbers(const char * & F1P,const char * & F2P,const char * F1End,const char * F2End,double AbsTolerance,double RelTolerance,std::string * ErrorMsg)81 static bool CompareNumbers(const char *&F1P, const char *&F2P,
82 const char *F1End, const char *F2End,
83 double AbsTolerance, double RelTolerance,
84 std::string *ErrorMsg) {
85 const char *F1NumEnd, *F2NumEnd;
86 double V1 = 0.0, V2 = 0.0;
87
88 // If one of the positions is at a space and the other isn't, chomp up 'til
89 // the end of the space.
90 while (isspace(*F1P) && F1P != F1End)
91 ++F1P;
92 while (isspace(*F2P) && F2P != F2End)
93 ++F2P;
94
95 // If we stop on numbers, compare their difference.
96 if (!isNumberChar(*F1P) || !isNumberChar(*F2P)) {
97 // The diff failed.
98 F1NumEnd = F1P;
99 F2NumEnd = F2P;
100 } else {
101 // Note that some ugliness is built into this to permit support for numbers
102 // that use "D" or "d" as their exponential marker, e.g. "1.234D45". This
103 // occurs in 200.sixtrack in spec2k.
104 V1 = strtod(F1P, const_cast<char**>(&F1NumEnd));
105 V2 = strtod(F2P, const_cast<char**>(&F2NumEnd));
106
107 if (*F1NumEnd == 'D' || *F1NumEnd == 'd') {
108 // Copy string into tmp buffer to replace the 'D' with an 'e'.
109 SmallString<200> StrTmp(F1P, EndOfNumber(F1NumEnd)+1);
110 // Strange exponential notation!
111 StrTmp[static_cast<unsigned>(F1NumEnd-F1P)] = 'e';
112
113 V1 = strtod(&StrTmp[0], const_cast<char**>(&F1NumEnd));
114 F1NumEnd = F1P + (F1NumEnd-&StrTmp[0]);
115 }
116
117 if (*F2NumEnd == 'D' || *F2NumEnd == 'd') {
118 // Copy string into tmp buffer to replace the 'D' with an 'e'.
119 SmallString<200> StrTmp(F2P, EndOfNumber(F2NumEnd)+1);
120 // Strange exponential notation!
121 StrTmp[static_cast<unsigned>(F2NumEnd-F2P)] = 'e';
122
123 V2 = strtod(&StrTmp[0], const_cast<char**>(&F2NumEnd));
124 F2NumEnd = F2P + (F2NumEnd-&StrTmp[0]);
125 }
126 }
127
128 if (F1NumEnd == F1P || F2NumEnd == F2P) {
129 if (ErrorMsg) {
130 *ErrorMsg = "FP Comparison failed, not a numeric difference between '";
131 *ErrorMsg += F1P[0];
132 *ErrorMsg += "' and '";
133 *ErrorMsg += F2P[0];
134 *ErrorMsg += "'";
135 }
136 return true;
137 }
138
139 // Check to see if these are inside the absolute tolerance
140 if (AbsTolerance < std::abs(V1-V2)) {
141 // Nope, check the relative tolerance...
142 double Diff;
143 if (V2)
144 Diff = std::abs(V1/V2 - 1.0);
145 else if (V1)
146 Diff = std::abs(V2/V1 - 1.0);
147 else
148 Diff = 0; // Both zero.
149 if (Diff > RelTolerance) {
150 if (ErrorMsg) {
151 raw_string_ostream(*ErrorMsg)
152 << "Compared: " << V1 << " and " << V2 << '\n'
153 << "abs. diff = " << std::abs(V1-V2) << " rel.diff = " << Diff << '\n'
154 << "Out of tolerance: rel/abs: " << RelTolerance << '/'
155 << AbsTolerance;
156 }
157 return true;
158 }
159 }
160
161 // Otherwise, advance our read pointers to the end of the numbers.
162 F1P = F1NumEnd; F2P = F2NumEnd;
163 return false;
164 }
165
166 /// DiffFilesWithTolerance - Compare the two files specified, returning 0 if the
167 /// files match, 1 if they are different, and 2 if there is a file error. This
168 /// function differs from DiffFiles in that you can specify an absolete and
169 /// relative FP error that is allowed to exist. If you specify a string to fill
170 /// in for the error option, it will set the string to an error message if an
171 /// error occurs, allowing the caller to distinguish between a failed diff and a
172 /// file system error.
173 ///
DiffFilesWithTolerance(const sys::PathWithStatus & FileA,const sys::PathWithStatus & FileB,double AbsTol,double RelTol,std::string * Error)174 int llvm::DiffFilesWithTolerance(const sys::PathWithStatus &FileA,
175 const sys::PathWithStatus &FileB,
176 double AbsTol, double RelTol,
177 std::string *Error) {
178 const sys::FileStatus *FileAStat = FileA.getFileStatus(false, Error);
179 if (!FileAStat)
180 return 2;
181 const sys::FileStatus *FileBStat = FileB.getFileStatus(false, Error);
182 if (!FileBStat)
183 return 2;
184
185 // Check for zero length files because some systems croak when you try to
186 // mmap an empty file.
187 size_t A_size = FileAStat->getSize();
188 size_t B_size = FileBStat->getSize();
189
190 // If they are both zero sized then they're the same
191 if (A_size == 0 && B_size == 0)
192 return 0;
193
194 // If only one of them is zero sized then they can't be the same
195 if ((A_size == 0 || B_size == 0)) {
196 if (Error)
197 *Error = "Files differ: one is zero-sized, the other isn't";
198 return 1;
199 }
200
201 // Now its safe to mmap the files into memory because both files
202 // have a non-zero size.
203 OwningPtr<MemoryBuffer> F1;
204 if (error_code ec = MemoryBuffer::getFile(FileA.c_str(), F1)) {
205 if (Error)
206 *Error = ec.message();
207 return 2;
208 }
209 OwningPtr<MemoryBuffer> F2;
210 if (error_code ec = MemoryBuffer::getFile(FileB.c_str(), F2)) {
211 if (Error)
212 *Error = ec.message();
213 return 2;
214 }
215
216 // Okay, now that we opened the files, scan them for the first difference.
217 const char *File1Start = F1->getBufferStart();
218 const char *File2Start = F2->getBufferStart();
219 const char *File1End = F1->getBufferEnd();
220 const char *File2End = F2->getBufferEnd();
221 const char *F1P = File1Start;
222 const char *F2P = File2Start;
223
224 // Are the buffers identical? Common case: Handle this efficiently.
225 if (A_size == B_size &&
226 std::memcmp(File1Start, File2Start, A_size) == 0)
227 return 0;
228
229 // Otherwise, we are done a tolerances are set.
230 if (AbsTol == 0 && RelTol == 0) {
231 if (Error)
232 *Error = "Files differ without tolerance allowance";
233 return 1; // Files different!
234 }
235
236 bool CompareFailed = false;
237 while (1) {
238 // Scan for the end of file or next difference.
239 while (F1P < File1End && F2P < File2End && *F1P == *F2P)
240 ++F1P, ++F2P;
241
242 if (F1P >= File1End || F2P >= File2End) break;
243
244 // Okay, we must have found a difference. Backup to the start of the
245 // current number each stream is at so that we can compare from the
246 // beginning.
247 F1P = BackupNumber(F1P, File1Start);
248 F2P = BackupNumber(F2P, File2Start);
249
250 // Now that we are at the start of the numbers, compare them, exiting if
251 // they don't match.
252 if (CompareNumbers(F1P, F2P, File1End, File2End, AbsTol, RelTol, Error)) {
253 CompareFailed = true;
254 break;
255 }
256 }
257
258 // Okay, we reached the end of file. If both files are at the end, we
259 // succeeded.
260 bool F1AtEnd = F1P >= File1End;
261 bool F2AtEnd = F2P >= File2End;
262 if (!CompareFailed && (!F1AtEnd || !F2AtEnd)) {
263 // Else, we might have run off the end due to a number: backup and retry.
264 if (F1AtEnd && isNumberChar(F1P[-1])) --F1P;
265 if (F2AtEnd && isNumberChar(F2P[-1])) --F2P;
266 F1P = BackupNumber(F1P, File1Start);
267 F2P = BackupNumber(F2P, File2Start);
268
269 // Now that we are at the start of the numbers, compare them, exiting if
270 // they don't match.
271 if (CompareNumbers(F1P, F2P, File1End, File2End, AbsTol, RelTol, Error))
272 CompareFailed = true;
273
274 // If we found the end, we succeeded.
275 if (F1P < File1End || F2P < File2End)
276 CompareFailed = true;
277 }
278
279 return CompareFailed;
280 }
281