cpp

Coverage Report

Created: 2024-06-29 00:17

/home/uke/oil/cpp/libc.cc
Line
Count
Source (jump to first uncovered line)
1
// libc.cc: Replacement for pyext/libc.c
2
3
#include "cpp/libc.h"
4
5
#include <errno.h>
6
#include <fnmatch.h>
7
#include <glob.h>
8
#include <locale.h>
9
#include <regex.h>
10
#include <sys/ioctl.h>
11
#include <unistd.h>  // gethostname()
12
#include <wchar.h>
13
14
namespace libc {
15
16
2
BigStr* gethostname() {
17
  // Note: Fixed issue #1656 - OS X and FreeBSD don't have HOST_NAME_MAX
18
  // https://reviews.freebsd.org/D30062
19
2
  BigStr* result = OverAllocatedStr(_POSIX_HOST_NAME_MAX);
20
2
  int status = ::gethostname(result->data_, _POSIX_HOST_NAME_MAX);
21
2
  if (status != 0) {
22
0
    throw Alloc<OSError>(errno);
23
0
  }
24
  // Important: set the length of the string!
25
2
  result->MaybeShrink(strlen(result->data_));
26
2
  return result;
27
2
}
28
29
2
BigStr* realpath(BigStr* path) {
30
2
  BigStr* result = OverAllocatedStr(PATH_MAX);
31
2
  char* p = ::realpath(path->data_, result->data_);
32
2
  if (p == nullptr) {
33
1
    throw Alloc<OSError>(errno);
34
1
  }
35
1
  result->MaybeShrink(strlen(result->data_));
36
1
  return result;
37
2
}
38
39
5
int fnmatch(BigStr* pat, BigStr* str, int flags) {
40
5
#ifdef FNM_EXTMATCH
41
5
  flags |= FNM_EXTMATCH;
42
#else
43
  // TODO: We should detect this at ./configure time, and then maybe flag these
44
  // at parse time, not runtime
45
#endif
46
47
5
  int result = ::fnmatch(pat->data_, str->data_, flags);
48
5
  switch (result) {
49
3
  case 0:
50
3
    return 1;
51
2
  case FNM_NOMATCH:
52
2
    return 0;
53
0
  default:
54
    // Other error
55
0
    return -1;
56
5
  }
57
5
}
58
59
2
List<BigStr*>* glob(BigStr* pat) {
60
2
  glob_t results;
61
  // Hm, it's weird that the first one can't be called with GLOB_APPEND.  You
62
  // get a segfault.
63
2
  int flags = 0;
64
  // int flags = GLOB_APPEND;
65
  // flags |= GLOB_NOMAGIC;
66
2
  int ret = glob(pat->data_, flags, NULL, &results);
67
68
2
  const char* err_str = NULL;
69
2
  switch (ret) {
70
1
  case 0:  // no error
71
1
    break;
72
0
  case GLOB_ABORTED:
73
0
    err_str = "read error";
74
0
    break;
75
1
  case GLOB_NOMATCH:
76
    // No error, because not matching isn't necessarily a problem.
77
    // NOTE: This can be turned on to log overaggressive calls to glob().
78
    // err_str = "nothing matched";
79
1
    break;
80
0
  case GLOB_NOSPACE:
81
0
    err_str = "no dynamic memory";
82
0
    break;
83
0
  default:
84
0
    err_str = "unknown problem";
85
0
    break;
86
2
  }
87
2
  if (err_str) {
88
0
    throw Alloc<RuntimeError>(StrFromC(err_str));
89
0
  }
90
91
  // http://stackoverflow.com/questions/3512414/does-this-pylist-appendlist-py-buildvalue-leak
92
2
  size_t n = results.gl_pathc;
93
2
  auto matches = NewList<BigStr*>();
94
95
  // Print array of results
96
2
  size_t i;
97
5
  for (i = 0; i < n; i++) {
98
3
    const char* m = results.gl_pathv[i];
99
3
    matches->append(StrFromC(m));
100
3
  }
101
2
  globfree(&results);
102
103
2
  return matches;
104
2
}
105
106
// Raises RuntimeError if the pattern is invalid.  TODO: Use a different
107
// exception?
108
List<int>* regex_search(BigStr* pattern, int cflags, BigStr* str, int eflags,
109
4
                        int pos) {
110
4
  cflags |= REG_EXTENDED;
111
4
  regex_t pat;
112
4
  int status = regcomp(&pat, pattern->data_, cflags);
113
4
  if (status != 0) {
114
0
    char error_desc[50];
115
0
    regerror(status, &pat, error_desc, 50);
116
117
0
    char error_message[80];
118
0
    snprintf(error_message, 80, "Invalid regex %s (%s)", pattern->data_,
119
0
             error_desc);
120
121
0
    throw Alloc<ValueError>(StrFromC(error_message));
122
0
  }
123
  // log("pat = %d, str = %d", len(pattern), len(str));
124
125
4
  int num_groups = pat.re_nsub + 1;  // number of captures
126
127
4
  List<int>* indices = NewList<int>();
128
4
  indices->reserve(num_groups * 2);
129
130
4
  const char* s = str->data_;
131
4
  regmatch_t* pmatch =
132
4
      static_cast<regmatch_t*>(malloc(sizeof(regmatch_t) * num_groups));
133
4
  bool match = regexec(&pat, s + pos, num_groups, pmatch, eflags) == 0;
134
4
  if (match) {
135
3
    int i;
136
10
    for (i = 0; i < num_groups; i++) {
137
7
      int start = pmatch[i].rm_so;
138
7
      if (start != -1) {
139
6
        start += pos;
140
6
      }
141
7
      indices->append(start);
142
143
7
      int end = pmatch[i].rm_eo;
144
7
      if (end != -1) {
145
6
        end += pos;
146
6
      }
147
7
      indices->append(end);
148
7
    }
149
3
  }
150
151
4
  free(pmatch);
152
4
  regfree(&pat);
153
154
4
  if (!match) {
155
1
    return nullptr;
156
1
  }
157
158
3
  return indices;
159
4
}
160
161
// For ${//}, the number of groups is always 1, so we want 2 match position
162
// results -- the whole regex (which we ignore), and then first group.
163
//
164
// For [[ =~ ]], do we need to count how many matches the user gave?
165
166
const int NMATCH = 2;
167
168
// Odd: This a Tuple2* not Tuple2 because it's Optional[Tuple2]!
169
Tuple2<int, int>* regex_first_group_match(BigStr* pattern, BigStr* str,
170
3
                                          int pos) {
171
3
  regex_t pat;
172
3
  regmatch_t m[NMATCH];
173
174
  // Could have been checked by regex_parse for [[ =~ ]], but not for glob
175
  // patterns like ${foo/x*/y}.
176
177
3
  if (regcomp(&pat, pattern->data_, REG_EXTENDED) != 0) {
178
0
    throw Alloc<RuntimeError>(
179
0
        StrFromC("Invalid regex syntax (func_regex_first_group_match)"));
180
0
  }
181
182
  // Match at offset 'pos'
183
3
  int result = regexec(&pat, str->data_ + pos, NMATCH, m, 0 /*flags*/);
184
3
  regfree(&pat);
185
186
3
  if (result != 0) {
187
0
    return nullptr;
188
0
  }
189
190
  // Assume there is a match
191
3
  regoff_t start = m[1].rm_so;
192
3
  regoff_t end = m[1].rm_eo;
193
3
  Tuple2<int, int>* tup = Alloc<Tuple2<int, int>>(pos + start, pos + end);
194
195
3
  return tup;
196
3
}
197
198
1
int wcswidth(BigStr* s) {
199
  // Behavior of mbstowcs() depends on LC_CTYPE
200
201
  // Calculate length first
202
1
  int num_wide_chars = ::mbstowcs(NULL, s->data_, 0);
203
1
  if (num_wide_chars == -1) {
204
0
    throw Alloc<UnicodeError>(StrFromC("mbstowcs() 1"));
205
0
  }
206
207
  // Allocate buffer
208
1
  int buf_size = (num_wide_chars + 1) * sizeof(wchar_t);
209
1
  wchar_t* wide_chars = static_cast<wchar_t*>(malloc(buf_size));
210
1
  DCHECK(wide_chars != nullptr);
211
212
  // Convert to wide chars
213
0
  num_wide_chars = ::mbstowcs(wide_chars, s->data_, num_wide_chars);
214
1
  if (num_wide_chars == -1) {
215
0
    free(wide_chars);  // cleanup
216
217
0
    throw Alloc<UnicodeError>(StrFromC("mbstowcs() 2"));
218
0
  }
219
220
  // Find number of columns
221
1
  int width = ::wcswidth(wide_chars, num_wide_chars);
222
1
  if (width == -1) {
223
0
    free(wide_chars);  // cleanup
224
225
    // unprintable chars
226
0
    throw Alloc<UnicodeError>(StrFromC("wcswidth()"));
227
0
  }
228
229
1
  free(wide_chars);
230
1
  return width;
231
1
}
232
233
1
int get_terminal_width() {
234
1
  struct winsize w;
235
1
  if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &w) == -1) {
236
1
    throw Alloc<IOError>(errno);
237
1
  }
238
0
  return w.ws_col;
239
1
}
240
241
}  // namespace libc