Line data Source code
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/ADT/SmallString.h"
17 : #include "llvm/Support/ErrorOr.h"
18 : #include "llvm/Support/MemoryBuffer.h"
19 : #include "llvm/Support/raw_ostream.h"
20 : #include <cctype>
21 : #include <cmath>
22 : #include <cstdint>
23 : #include <cstdlib>
24 : #include <cstring>
25 : #include <memory>
26 : #include <system_error>
27 :
28 : using namespace llvm;
29 :
30 : static bool isSignedChar(char C) {
31 0 : return (C == '+' || C == '-');
32 : }
33 :
34 : static bool isExponentChar(char C) {
35 : switch (C) {
36 : case 'D': // Strange exponential notation.
37 : case 'd': // Strange exponential notation.
38 : case 'e':
39 : case 'E': return true;
40 : default: return false;
41 : }
42 : }
43 :
44 0 : static bool isNumberChar(char C) {
45 : switch (C) {
46 : case '0': case '1': case '2': case '3': case '4':
47 : case '5': case '6': case '7': case '8': case '9':
48 : case '.': return true;
49 0 : default: return isSignedChar(C) || isExponentChar(C);
50 : }
51 : }
52 :
53 0 : static const char *BackupNumber(const char *Pos, const char *FirstChar) {
54 : // If we didn't stop in the middle of a number, don't backup.
55 0 : if (!isNumberChar(*Pos)) return Pos;
56 :
57 : // Otherwise, return to the start of the number.
58 : bool HasPeriod = false;
59 0 : while (Pos > FirstChar && isNumberChar(Pos[-1])) {
60 : // Backup over at most one period.
61 0 : if (Pos[-1] == '.') {
62 0 : if (HasPeriod)
63 : break;
64 : HasPeriod = true;
65 : }
66 :
67 0 : --Pos;
68 0 : if (Pos > FirstChar && isSignedChar(Pos[0]) && !isExponentChar(Pos[-1]))
69 : break;
70 : }
71 : return Pos;
72 : }
73 :
74 : /// EndOfNumber - Return the first character that is not part of the specified
75 : /// number. This assumes that the buffer is null terminated, so it won't fall
76 : /// off the end.
77 : static const char *EndOfNumber(const char *Pos) {
78 0 : while (isNumberChar(*Pos))
79 0 : ++Pos;
80 : return Pos;
81 : }
82 :
83 : /// CompareNumbers - compare two numbers, returning true if they are different.
84 0 : static bool CompareNumbers(const char *&F1P, const char *&F2P,
85 : const char *F1End, const char *F2End,
86 : double AbsTolerance, double RelTolerance,
87 : std::string *ErrorMsg) {
88 : const char *F1NumEnd, *F2NumEnd;
89 : double V1 = 0.0, V2 = 0.0;
90 :
91 : // If one of the positions is at a space and the other isn't, chomp up 'til
92 : // the end of the space.
93 0 : while (isspace(static_cast<unsigned char>(*F1P)) && F1P != F1End)
94 0 : ++F1P;
95 0 : while (isspace(static_cast<unsigned char>(*F2P)) && F2P != F2End)
96 0 : ++F2P;
97 :
98 : // If we stop on numbers, compare their difference.
99 0 : if (!isNumberChar(*F1P) || !isNumberChar(*F2P)) {
100 : // The diff failed.
101 0 : F1NumEnd = F1P;
102 0 : F2NumEnd = F2P;
103 : } else {
104 : // Note that some ugliness is built into this to permit support for numbers
105 : // that use "D" or "d" as their exponential marker, e.g. "1.234D45". This
106 : // occurs in 200.sixtrack in spec2k.
107 0 : V1 = strtod(F1P, const_cast<char**>(&F1NumEnd));
108 0 : V2 = strtod(F2P, const_cast<char**>(&F2NumEnd));
109 :
110 0 : if (*F1NumEnd == 'D' || *F1NumEnd == 'd') {
111 : // Copy string into tmp buffer to replace the 'D' with an 'e'.
112 0 : SmallString<200> StrTmp(F1P, EndOfNumber(F1NumEnd)+1);
113 : // Strange exponential notation!
114 0 : StrTmp[static_cast<unsigned>(F1NumEnd-F1P)] = 'e';
115 :
116 0 : V1 = strtod(&StrTmp[0], const_cast<char**>(&F1NumEnd));
117 0 : F1NumEnd = F1P + (F1NumEnd-&StrTmp[0]);
118 : }
119 :
120 0 : if (*F2NumEnd == 'D' || *F2NumEnd == 'd') {
121 : // Copy string into tmp buffer to replace the 'D' with an 'e'.
122 0 : SmallString<200> StrTmp(F2P, EndOfNumber(F2NumEnd)+1);
123 : // Strange exponential notation!
124 0 : StrTmp[static_cast<unsigned>(F2NumEnd-F2P)] = 'e';
125 :
126 0 : V2 = strtod(&StrTmp[0], const_cast<char**>(&F2NumEnd));
127 0 : F2NumEnd = F2P + (F2NumEnd-&StrTmp[0]);
128 : }
129 : }
130 :
131 0 : if (F1NumEnd == F1P || F2NumEnd == F2P) {
132 0 : if (ErrorMsg) {
133 : *ErrorMsg = "FP Comparison failed, not a numeric difference between '";
134 0 : *ErrorMsg += F1P[0];
135 : *ErrorMsg += "' and '";
136 0 : *ErrorMsg += F2P[0];
137 : *ErrorMsg += "'";
138 : }
139 0 : return true;
140 : }
141 :
142 : // Check to see if these are inside the absolute tolerance
143 0 : if (AbsTolerance < std::abs(V1-V2)) {
144 : // Nope, check the relative tolerance...
145 : double Diff;
146 0 : if (V2)
147 0 : Diff = std::abs(V1/V2 - 1.0);
148 0 : else if (V1)
149 0 : Diff = std::abs(V2/V1 - 1.0);
150 : else
151 : Diff = 0; // Both zero.
152 0 : if (Diff > RelTolerance) {
153 0 : if (ErrorMsg) {
154 0 : raw_string_ostream(*ErrorMsg)
155 0 : << "Compared: " << V1 << " and " << V2 << '\n'
156 0 : << "abs. diff = " << std::abs(V1-V2) << " rel.diff = " << Diff << '\n'
157 0 : << "Out of tolerance: rel/abs: " << RelTolerance << '/'
158 0 : << AbsTolerance;
159 : }
160 0 : return true;
161 : }
162 : }
163 :
164 : // Otherwise, advance our read pointers to the end of the numbers.
165 0 : F1P = F1NumEnd; F2P = F2NumEnd;
166 0 : return false;
167 : }
168 :
169 : /// DiffFilesWithTolerance - Compare the two files specified, returning 0 if the
170 : /// files match, 1 if they are different, and 2 if there is a file error. This
171 : /// function differs from DiffFiles in that you can specify an absolete and
172 : /// relative FP error that is allowed to exist. If you specify a string to fill
173 : /// in for the error option, it will set the string to an error message if an
174 : /// error occurs, allowing the caller to distinguish between a failed diff and a
175 : /// file system error.
176 : ///
177 0 : int llvm::DiffFilesWithTolerance(StringRef NameA,
178 : StringRef NameB,
179 : double AbsTol, double RelTol,
180 : std::string *Error) {
181 : // Now its safe to mmap the files into memory because both files
182 : // have a non-zero size.
183 0 : ErrorOr<std::unique_ptr<MemoryBuffer>> F1OrErr = MemoryBuffer::getFile(NameA);
184 0 : if (std::error_code EC = F1OrErr.getError()) {
185 0 : if (Error)
186 0 : *Error = EC.message();
187 : return 2;
188 : }
189 : MemoryBuffer &F1 = *F1OrErr.get();
190 :
191 0 : ErrorOr<std::unique_ptr<MemoryBuffer>> F2OrErr = MemoryBuffer::getFile(NameB);
192 0 : if (std::error_code EC = F2OrErr.getError()) {
193 0 : if (Error)
194 0 : *Error = EC.message();
195 : return 2;
196 : }
197 : MemoryBuffer &F2 = *F2OrErr.get();
198 :
199 : // Okay, now that we opened the files, scan them for the first difference.
200 0 : const char *File1Start = F1.getBufferStart();
201 0 : const char *File2Start = F2.getBufferStart();
202 0 : const char *File1End = F1.getBufferEnd();
203 0 : const char *File2End = F2.getBufferEnd();
204 0 : const char *F1P = File1Start;
205 0 : const char *F2P = File2Start;
206 : uint64_t A_size = F1.getBufferSize();
207 : uint64_t B_size = F2.getBufferSize();
208 :
209 : // Are the buffers identical? Common case: Handle this efficiently.
210 0 : if (A_size == B_size &&
211 0 : std::memcmp(File1Start, File2Start, A_size) == 0)
212 : return 0;
213 :
214 : // Otherwise, we are done a tolerances are set.
215 0 : if (AbsTol == 0 && RelTol == 0) {
216 0 : if (Error)
217 : *Error = "Files differ without tolerance allowance";
218 0 : return 1; // Files different!
219 : }
220 :
221 : bool CompareFailed = false;
222 : while (true) {
223 : // Scan for the end of file or next difference.
224 0 : while (F1P < File1End && F2P < File2End && *F1P == *F2P) {
225 0 : ++F1P;
226 0 : ++F2P;
227 : }
228 :
229 0 : if (F1P >= File1End || F2P >= File2End) break;
230 :
231 : // Okay, we must have found a difference. Backup to the start of the
232 : // current number each stream is at so that we can compare from the
233 : // beginning.
234 0 : F1P = BackupNumber(F1P, File1Start);
235 0 : F2P = BackupNumber(F2P, File2Start);
236 :
237 : // Now that we are at the start of the numbers, compare them, exiting if
238 : // they don't match.
239 0 : if (CompareNumbers(F1P, F2P, File1End, File2End, AbsTol, RelTol, Error)) {
240 : CompareFailed = true;
241 : break;
242 : }
243 : }
244 :
245 : // Okay, we reached the end of file. If both files are at the end, we
246 : // succeeded.
247 0 : bool F1AtEnd = F1P >= File1End;
248 0 : bool F2AtEnd = F2P >= File2End;
249 0 : if (!CompareFailed && (!F1AtEnd || !F2AtEnd)) {
250 : // Else, we might have run off the end due to a number: backup and retry.
251 0 : if (F1AtEnd && isNumberChar(F1P[-1])) --F1P;
252 0 : if (F2AtEnd && isNumberChar(F2P[-1])) --F2P;
253 0 : F1P = BackupNumber(F1P, File1Start);
254 0 : F2P = BackupNumber(F2P, File2Start);
255 :
256 : // Now that we are at the start of the numbers, compare them, exiting if
257 : // they don't match.
258 0 : if (CompareNumbers(F1P, F2P, File1End, File2End, AbsTol, RelTol, Error))
259 : CompareFailed = true;
260 :
261 : // If we found the end, we succeeded.
262 0 : if (F1P < File1End || F2P < File2End)
263 : CompareFailed = true;
264 : }
265 :
266 0 : return CompareFailed;
267 : }
|
