PrintfFormatString.cpp revision e729acbba75903f42e79e72e46cdebe3f4c35521 
1//= PrintfFormatStrings.cpp - Analysis of printf format strings --*- C++ -*-==//
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// Handling of format string in printf and friends.  The structure of format
11// strings for fprintf() are described in C99 7.19.6.1.
12//
13//===----------------------------------------------------------------------===//
14
15#include "clang/Analysis/Analyses/PrintfFormatString.h"
16
17using namespace clang;
18using namespace analyze_printf;
19
20namespace {
21class FormatSpecifierResult {
22  FormatSpecifier FS;
23  const char *Start;
24  bool HasError;
25public:
26  FormatSpecifierResult(bool err = false)
27    : Start(0), HasError(err) {}
28  FormatSpecifierResult(const char *start,
29                        const FormatSpecifier &fs)
30    : FS(fs), Start(start), HasError(false) {}
31
32
33  const char *getStart() const { return Start; }
34  bool hasError() const { return HasError; }
35  bool hasValue() const { return Start != 0; }
36  const FormatSpecifier &getValue() const {
37    assert(hasValue());
38    return FS;
39  }
40  const FormatSpecifier &getValue() { return FS; }
41};
42} // end anonymous namespace
43
44template <typename T>
45class UpdateOnReturn {
46  T &ValueToUpdate;
47  const T &ValueToCopy;
48public:
49  UpdateOnReturn(T &valueToUpdate, const T &valueToCopy)
50    : ValueToUpdate(valueToUpdate), ValueToCopy(valueToCopy) {}
51
52  ~UpdateOnReturn() {
53    ValueToUpdate = ValueToCopy;
54  }
55};
56
57static OptionalAmount ParseAmount(const char *&Beg, const char *E) {
58  const char *I = Beg;
59  UpdateOnReturn <const char*> UpdateBeg(Beg, I);
60
61  bool foundDigits = false;
62  unsigned accumulator = 0;
63
64  for ( ; I != E; ++I) {
65    char c = *I;
66    if (c >= '0' && c <= '9') {
67      foundDigits = true;
68      accumulator += (accumulator * 10) + (c - '0');
69      continue;
70    }
71
72    if (foundDigits)
73      return OptionalAmount(accumulator);
74
75    if (c == '*')
76      return OptionalAmount(OptionalAmount::Arg);
77
78    break;
79  }
80
81  return OptionalAmount();
82}
83
84static FormatSpecifierResult ParseFormatSpecifier(FormatStringHandler &H,
85                                                  const char *&Beg, const char *E) {
86
87  const char *I = Beg;
88  const char *Start = 0;
89  UpdateOnReturn <const char*> UpdateBeg(Beg, I);
90
91  // Look for a '%' character that indicates the start of a format specifier.
92  for ( ; I != E ; ++I) {
93    char c = *I;
94    if (c == '\0') {
95      // Detect spurious null characters, which are likely errors.
96      H.HandleNullChar(I);
97      return true;
98    }
99    if (c == '%') {
100      Start = I++;  // Record the start of the format specifier.
101      break;
102    }
103  }
104
105  // No format specifier found?
106  if (!Start)
107    return false;
108
109  if (I == E) {
110    // No more characters left?
111    H.HandleIncompleteFormatSpecifier(Start, E);
112    return true;
113  }
114
115  FormatSpecifier FS;
116
117  // Look for flags (if any).
118  bool hasMore = true;
119  for ( ; I != E; ++I) {
120    switch (*I) {
121      default: hasMore = false; break;
122      case '-': FS.setIsLeftJustified(); break;
123      case '+': FS.setHasPlusPrefix(); break;
124      case ' ': FS.setHasSpacePrefix(); break;
125      case '#': FS.setHasAlternativeForm(); break;
126      case '0': FS.setHasLeadingZeros(); break;
127    }
128    if (!hasMore)
129      break;
130  }
131
132  if (I == E) {
133    // No more characters left?
134    H.HandleIncompleteFormatSpecifier(Start, E);
135    return true;
136  }
137
138  // Look for the field width (if any).
139  FS.setFieldWidth(ParseAmount(I, E));
140
141  if (I == E) {
142    // No more characters left?
143    H.HandleIncompleteFormatSpecifier(Start, E);
144    return true;
145  }
146
147  // Look for the precision (if any).
148  if (*I == '.') {
149    const char *startPrecision = I++;
150    if (I == E) {
151      H.HandleIncompletePrecision(I - 1);
152      return true;
153    }
154
155    FS.setPrecision(ParseAmount(I, E));
156
157    if (I == E) {
158      // No more characters left?
159      H.HandleIncompletePrecision(startPrecision);
160      return true;
161    }
162  }
163
164  // Look for the length modifier.
165  LengthModifier lm = None;
166  switch (*I) {
167    default:
168      break;
169    case 'h':
170      ++I;
171      lm = (I != E && *I == 'h') ? ++I, AsChar : AsShort;
172      break;
173    case 'l':
174      ++I;
175      lm = (I != E && *I == 'l') ? ++I, AsLongLong : AsLong;
176      break;
177    case 'j': lm = AsIntMax;     ++I; break;
178    case 'z': lm = AsSizeT;      ++I; break;
179    case 't': lm = AsPtrDiff;    ++I; break;
180    case 'L': lm = AsLongDouble; ++I; break;
181  }
182  FS.setLengthModifier(lm);
183
184  if (I == E) {
185    // No more characters left?
186    H.HandleIncompleteFormatSpecifier(Start, E);
187    return true;
188  }
189
190  // Finally, look for the conversion specifier.
191  const char *conversionPosition = I++;
192  ConversionSpecifier::Kind k;
193  switch (*conversionPosition) {
194    default:
195      H.HandleInvalidConversionSpecifier(conversionPosition);
196      return true;
197    // C99: 7.19.6.1 (section 8).
198    case 'd': k = ConversionSpecifier::dArg; break;
199    case 'i': k = ConversionSpecifier::iArg; break;
200    case 'o': k = ConversionSpecifier::oArg; break;
201    case 'u': k = ConversionSpecifier::uArg; break;
202    case 'x': k = ConversionSpecifier::xArg; break;
203    case 'X': k = ConversionSpecifier::XArg; break;
204    case 'f': k = ConversionSpecifier::fArg; break;
205    case 'F': k = ConversionSpecifier::FArg; break;
206    case 'e': k = ConversionSpecifier::eArg; break;
207    case 'E': k = ConversionSpecifier::EArg; break;
208    case 'g': k = ConversionSpecifier::gArg; break;
209    case 'G': k = ConversionSpecifier::GArg; break;
210    case 'a': k = ConversionSpecifier::aArg; break;
211    case 'A': k = ConversionSpecifier::AArg; break;
212    case 'c': k = ConversionSpecifier::IntAsCharArg; break;
213    case 's': k = ConversionSpecifier::CStrArg;      break;
214    case 'p': k = ConversionSpecifier::VoidPtrArg;   break;
215    case 'n': k = ConversionSpecifier::OutIntPtrArg; break;
216    case '%': k = ConversionSpecifier::PercentArg;   break;
217    // Objective-C.
218    case '@': k = ConversionSpecifier::ObjCObjArg; break;
219  }
220  FS.setConversionSpecifier(ConversionSpecifier(conversionPosition, k));
221  return FormatSpecifierResult(Start, FS);
222}
223
224namespace clang { namespace analyze_printf {
225bool ParseFormatString(FormatStringHandler &H,
226                       const char *I, const char *E) {
227  // Keep looking for a format specifier until we have exhausted the string.
228  while (I != E) {
229    const FormatSpecifierResult &FSR = ParseFormatSpecifier(H, I, E);
230    // Did an error of any kind occur when parsing the specifier?  If so,
231    // don't do any more processing.
232    if (FSR.hasError())
233      return true;;
234    // Done processing the string?
235    if (!FSR.hasValue())
236      break;
237    // We have a format specifier.  Pass it to the callback.
238    if (!H.HandleFormatSpecifier(FSR.getValue(), FSR.getStart(),
239                                 I - FSR.getStart()))
240      return false;
241  }
242  assert(I == E && "Format string not exhausted");
243  return false;
244}
245}}
246
247FormatStringHandler::~FormatStringHandler() {}
248