1//===-- NSString.cpp ------------------------------------------------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8
9#include "NSString.h"
10
11#include "lldb/Core/ValueObject.h"
12#include "lldb/Core/ValueObjectConstResult.h"
13#include "lldb/DataFormatters/FormattersHelpers.h"
14#include "lldb/DataFormatters/StringPrinter.h"
15#include "lldb/Target/Language.h"
16#include "lldb/Target/Target.h"
17#include "lldb/Utility/ConstString.h"
18#include "lldb/Utility/DataBufferHeap.h"
19#include "lldb/Utility/Endian.h"
20#include "lldb/Utility/Status.h"
21#include "lldb/Utility/Stream.h"
22
23using namespace lldb;
24using namespace lldb_private;
25using namespace lldb_private::formatters;
26
27std::map<ConstString, CXXFunctionSummaryFormat::Callback> &
28NSString_Additionals::GetAdditionalSummaries() {
29  static std::map<ConstString, CXXFunctionSummaryFormat::Callback> g_map;
30  return g_map;
31}
32
33bool lldb_private::formatters::NSStringSummaryProvider(
34    ValueObject &valobj, Stream &stream,
35    const TypeSummaryOptions &summary_options) {
36  static constexpr llvm::StringLiteral g_TypeHint("NSString");
37
38  ProcessSP process_sp = valobj.GetProcessSP();
39  if (!process_sp)
40    return false;
41
42  ObjCLanguageRuntime *runtime = ObjCLanguageRuntime::Get(*process_sp);
43
44  if (!runtime)
45    return false;
46
47  ObjCLanguageRuntime::ClassDescriptorSP descriptor(
48      runtime->GetClassDescriptor(valobj));
49
50  if (!descriptor.get() || !descriptor->IsValid())
51    return false;
52
53  uint32_t ptr_size = process_sp->GetAddressByteSize();
54
55  lldb::addr_t valobj_addr = valobj.GetValueAsUnsigned(0);
56
57  if (!valobj_addr)
58    return false;
59
60  ConstString class_name_cs = descriptor->GetClassName();
61  llvm::StringRef class_name = class_name_cs.GetStringRef();
62
63  if (class_name.empty())
64    return false;
65
66  bool is_tagged_ptr = class_name == "NSTaggedPointerString" &&
67                       descriptor->GetTaggedPointerInfo();
68  // for a tagged pointer, the descriptor has everything we need
69  if (is_tagged_ptr)
70    return NSTaggedString_SummaryProvider(valobj, descriptor, stream,
71                                          summary_options);
72
73  auto &additionals_map(NSString_Additionals::GetAdditionalSummaries());
74  auto iter = additionals_map.find(class_name_cs), end = additionals_map.end();
75  if (iter != end)
76    return iter->second(valobj, stream, summary_options);
77
78  // if not a tagged pointer that we know about, try the normal route
79  uint64_t info_bits_location = valobj_addr + ptr_size;
80  if (process_sp->GetByteOrder() != lldb::eByteOrderLittle)
81    info_bits_location += 3;
82
83  Status error;
84
85  uint8_t info_bits = process_sp->ReadUnsignedIntegerFromMemory(
86      info_bits_location, 1, 0, error);
87  if (error.Fail())
88    return false;
89
90  bool is_mutable = (info_bits & 1) == 1;
91  bool is_inline = (info_bits & 0x60) == 0;
92  bool has_explicit_length = (info_bits & (1 | 4)) != 4;
93  bool is_unicode = (info_bits & 0x10) == 0x10;
94  bool is_path_store = class_name == "NSPathStore2";
95  bool has_null = (info_bits & 8) == 8;
96
97  size_t explicit_length = 0;
98  if (!has_null && has_explicit_length && !is_path_store) {
99    lldb::addr_t explicit_length_offset = 2 * ptr_size;
100    if (is_mutable && !is_inline)
101      explicit_length_offset =
102          explicit_length_offset + ptr_size; //  notInlineMutable.length;
103    else if (is_inline)
104      explicit_length = explicit_length + 0; // inline1.length;
105    else if (!is_inline && !is_mutable)
106      explicit_length_offset =
107          explicit_length_offset + ptr_size; // notInlineImmutable1.length;
108    else
109      explicit_length_offset = 0;
110
111    if (explicit_length_offset) {
112      explicit_length_offset = valobj_addr + explicit_length_offset;
113      explicit_length = process_sp->ReadUnsignedIntegerFromMemory(
114          explicit_length_offset, 4, 0, error);
115    }
116  }
117
118  const llvm::StringSet<> supported_string_classes = {
119      "NSString",     "CFMutableStringRef",
120      "CFStringRef",  "__NSCFConstantString",
121      "__NSCFString", "NSCFConstantString",
122      "NSCFString",   "NSPathStore2"};
123  if (supported_string_classes.count(class_name) == 0) {
124    // not one of us - but tell me class name
125    stream.Printf("class name = %s", class_name_cs.GetCString());
126    return true;
127  }
128
129  llvm::StringRef prefix, suffix;
130  if (Language *language = Language::FindPlugin(summary_options.GetLanguage()))
131    std::tie(prefix, suffix) = language->GetFormatterPrefixSuffix(g_TypeHint);
132
133  StringPrinter::ReadStringAndDumpToStreamOptions options(valobj);
134  options.SetPrefixToken(prefix.str());
135  options.SetSuffixToken(suffix.str());
136
137  if (is_mutable) {
138    uint64_t location = 2 * ptr_size + valobj_addr;
139    location = process_sp->ReadPointerFromMemory(location, error);
140    if (error.Fail())
141      return false;
142    if (has_explicit_length && is_unicode) {
143      options.SetLocation(location);
144      options.SetTargetSP(valobj.GetTargetSP());
145      options.SetStream(&stream);
146      options.SetQuote('"');
147      options.SetSourceSize(explicit_length);
148      options.SetHasSourceSize(has_explicit_length);
149      options.SetNeedsZeroTermination(false);
150      options.SetIgnoreMaxLength(summary_options.GetCapping() ==
151                                 TypeSummaryCapping::eTypeSummaryUncapped);
152      options.SetBinaryZeroIsTerminator(false);
153      return StringPrinter::ReadStringAndDumpToStream<
154          StringPrinter::StringElementType::UTF16>(options);
155    } else {
156      options.SetLocation(location + 1);
157      options.SetTargetSP(valobj.GetTargetSP());
158      options.SetStream(&stream);
159      options.SetSourceSize(explicit_length);
160      options.SetHasSourceSize(has_explicit_length);
161      options.SetNeedsZeroTermination(false);
162      options.SetIgnoreMaxLength(summary_options.GetCapping() ==
163                                 TypeSummaryCapping::eTypeSummaryUncapped);
164      options.SetBinaryZeroIsTerminator(false);
165      return StringPrinter::ReadStringAndDumpToStream<
166          StringPrinter::StringElementType::ASCII>(options);
167    }
168  } else if (is_inline && has_explicit_length && !is_unicode &&
169             !is_path_store && !is_mutable) {
170    uint64_t location = 3 * ptr_size + valobj_addr;
171
172    options.SetLocation(location);
173    options.SetTargetSP(valobj.GetTargetSP());
174    options.SetStream(&stream);
175    options.SetQuote('"');
176    options.SetSourceSize(explicit_length);
177    options.SetHasSourceSize(has_explicit_length);
178    options.SetIgnoreMaxLength(summary_options.GetCapping() ==
179                               TypeSummaryCapping::eTypeSummaryUncapped);
180    return StringPrinter::ReadStringAndDumpToStream<
181        StringPrinter::StringElementType::ASCII>(options);
182  } else if (is_unicode) {
183    uint64_t location = valobj_addr + 2 * ptr_size;
184    if (is_inline) {
185      if (!has_explicit_length) {
186        return false;
187      } else
188        location += ptr_size;
189    } else {
190      location = process_sp->ReadPointerFromMemory(location, error);
191      if (error.Fail())
192        return false;
193    }
194    options.SetLocation(location);
195    options.SetTargetSP(valobj.GetTargetSP());
196    options.SetStream(&stream);
197    options.SetQuote('"');
198    options.SetSourceSize(explicit_length);
199    options.SetHasSourceSize(has_explicit_length);
200    options.SetNeedsZeroTermination(!has_explicit_length);
201    options.SetIgnoreMaxLength(summary_options.GetCapping() ==
202                               TypeSummaryCapping::eTypeSummaryUncapped);
203    options.SetBinaryZeroIsTerminator(!has_explicit_length);
204    return StringPrinter::ReadStringAndDumpToStream<
205        StringPrinter::StringElementType::UTF16>(options);
206  } else if (is_path_store) {
207    // _lengthAndRefCount is the first ivar of NSPathStore2 (after the isa).
208    uint64_t length_ivar_offset = 1 * ptr_size;
209    CompilerType length_type = valobj.GetCompilerType().GetBasicTypeFromAST(
210        lldb::eBasicTypeUnsignedInt);
211    ValueObjectSP length_valobj_sp =
212        valobj.GetSyntheticChildAtOffset(length_ivar_offset, length_type, true,
213                                         ConstString("_lengthAndRefCount"));
214    if (!length_valobj_sp)
215      return false;
216    // Get the length out of _lengthAndRefCount.
217    explicit_length = length_valobj_sp->GetValueAsUnsigned(0) >> 20;
218    lldb::addr_t location = valobj.GetValueAsUnsigned(0) + ptr_size + 4;
219
220    options.SetLocation(location);
221    options.SetTargetSP(valobj.GetTargetSP());
222    options.SetStream(&stream);
223    options.SetQuote('"');
224    options.SetSourceSize(explicit_length);
225    options.SetHasSourceSize(has_explicit_length);
226    options.SetNeedsZeroTermination(!has_explicit_length);
227    options.SetIgnoreMaxLength(summary_options.GetCapping() ==
228                               TypeSummaryCapping::eTypeSummaryUncapped);
229    options.SetBinaryZeroIsTerminator(!has_explicit_length);
230    return StringPrinter::ReadStringAndDumpToStream<
231        StringPrinter::StringElementType::UTF16>(options);
232  } else if (is_inline) {
233    uint64_t location = valobj_addr + 2 * ptr_size;
234    if (!has_explicit_length) {
235      // in this kind of string, the byte before the string content is a length
236      // byte so let's try and use it to handle the embedded NUL case
237      Status error;
238      explicit_length =
239          process_sp->ReadUnsignedIntegerFromMemory(location, 1, 0, error);
240      has_explicit_length = !(error.Fail() || explicit_length == 0);
241      location++;
242    }
243    options.SetLocation(location);
244    options.SetTargetSP(valobj.GetTargetSP());
245    options.SetStream(&stream);
246    options.SetSourceSize(explicit_length);
247    options.SetHasSourceSize(has_explicit_length);
248    options.SetNeedsZeroTermination(!has_explicit_length);
249    options.SetIgnoreMaxLength(summary_options.GetCapping() ==
250                               TypeSummaryCapping::eTypeSummaryUncapped);
251    options.SetBinaryZeroIsTerminator(!has_explicit_length);
252    if (has_explicit_length)
253      return StringPrinter::ReadStringAndDumpToStream<
254          StringPrinter::StringElementType::UTF8>(options);
255    else
256      return StringPrinter::ReadStringAndDumpToStream<
257          StringPrinter::StringElementType::ASCII>(options);
258  } else {
259    uint64_t location = valobj_addr + 2 * ptr_size;
260    location = process_sp->ReadPointerFromMemory(location, error);
261    if (error.Fail())
262      return false;
263    if (has_explicit_length && !has_null)
264      explicit_length++; // account for the fact that there is no NULL and we
265                         // need to have one added
266    options.SetLocation(location);
267    options.SetTargetSP(valobj.GetTargetSP());
268    options.SetStream(&stream);
269    options.SetSourceSize(explicit_length);
270    options.SetHasSourceSize(has_explicit_length);
271    options.SetIgnoreMaxLength(summary_options.GetCapping() ==
272                               TypeSummaryCapping::eTypeSummaryUncapped);
273    return StringPrinter::ReadStringAndDumpToStream<
274        StringPrinter::StringElementType::ASCII>(options);
275  }
276}
277
278bool lldb_private::formatters::NSAttributedStringSummaryProvider(
279    ValueObject &valobj, Stream &stream, const TypeSummaryOptions &options) {
280  TargetSP target_sp(valobj.GetTargetSP());
281  if (!target_sp)
282    return false;
283  uint32_t addr_size = target_sp->GetArchitecture().GetAddressByteSize();
284  uint64_t pointer_value = valobj.GetValueAsUnsigned(0);
285  if (!pointer_value)
286    return false;
287  pointer_value += addr_size;
288  CompilerType type(valobj.GetCompilerType());
289  ExecutionContext exe_ctx(target_sp, false);
290  ValueObjectSP child_ptr_sp(valobj.CreateValueObjectFromAddress(
291      "string_ptr", pointer_value, exe_ctx, type));
292  if (!child_ptr_sp)
293    return false;
294  DataExtractor data;
295  Status error;
296  child_ptr_sp->GetData(data, error);
297  if (error.Fail())
298    return false;
299  ValueObjectSP child_sp(child_ptr_sp->CreateValueObjectFromData(
300      "string_data", data, exe_ctx, type));
301  child_sp->GetValueAsUnsigned(0);
302  if (child_sp)
303    return NSStringSummaryProvider(*child_sp, stream, options);
304  return false;
305}
306
307bool lldb_private::formatters::NSMutableAttributedStringSummaryProvider(
308    ValueObject &valobj, Stream &stream, const TypeSummaryOptions &options) {
309  return NSAttributedStringSummaryProvider(valobj, stream, options);
310}
311
312bool lldb_private::formatters::NSTaggedString_SummaryProvider(
313    ValueObject &valobj, ObjCLanguageRuntime::ClassDescriptorSP descriptor,
314    Stream &stream, const TypeSummaryOptions &summary_options) {
315  static constexpr llvm::StringLiteral g_TypeHint("NSString");
316
317  if (!descriptor)
318    return false;
319  uint64_t len_bits = 0, data_bits = 0;
320  if (!descriptor->GetTaggedPointerInfo(&len_bits, &data_bits, nullptr))
321    return false;
322
323  static const int g_MaxNonBitmaskedLen = 7; // TAGGED_STRING_UNPACKED_MAXLEN
324  static const int g_SixbitMaxLen = 9;
325  static const int g_fiveBitMaxLen = 11;
326
327  static const char *sixBitToCharLookup = "eilotrm.apdnsIc ufkMShjTRxgC4013"
328                                          "bDNvwyUL2O856P-B79AFKEWV_zGJ/HYX";
329
330  if (len_bits > g_fiveBitMaxLen)
331    return false;
332
333  llvm::StringRef prefix, suffix;
334  if (Language *language = Language::FindPlugin(summary_options.GetLanguage()))
335    std::tie(prefix, suffix) = language->GetFormatterPrefixSuffix(g_TypeHint);
336
337  // this is a fairly ugly trick - pretend that the numeric value is actually a
338  // char* this works under a few assumptions: little endian architecture
339  // sizeof(uint64_t) > g_MaxNonBitmaskedLen
340  if (len_bits <= g_MaxNonBitmaskedLen) {
341    stream << prefix;
342    stream.Printf("\"%s\"", (const char *)&data_bits);
343    stream << suffix;
344    return true;
345  }
346
347  // if the data is bitmasked, we need to actually process the bytes
348  uint8_t bitmask = 0;
349  uint8_t shift_offset = 0;
350
351  if (len_bits <= g_SixbitMaxLen) {
352    bitmask = 0x03f;
353    shift_offset = 6;
354  } else {
355    bitmask = 0x01f;
356    shift_offset = 5;
357  }
358
359  std::vector<uint8_t> bytes;
360  bytes.resize(len_bits);
361  for (; len_bits > 0; data_bits >>= shift_offset, --len_bits) {
362    uint8_t packed = data_bits & bitmask;
363    bytes.insert(bytes.begin(), sixBitToCharLookup[packed]);
364  }
365
366  stream << prefix;
367  stream.Printf("\"%s\"", &bytes[0]);
368  stream << suffix;
369  return true;
370}
371