1/* Read, sort and compare two directories.  Used for GNU DIFF.
2
3   Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1998, 2001, 2002
4   Free Software Foundation, Inc.
5
6   This file is part of GNU DIFF.
7
8   GNU DIFF is free software; you can redistribute it and/or modify
9   it under the terms of the GNU General Public License as published by
10   the Free Software Foundation; either version 2, or (at your option)
11   any later version.
12
13   GNU DIFF is distributed in the hope that it will be useful,
14   but WITHOUT ANY WARRANTY; without even the implied warranty of
15   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16   GNU General Public License for more details.
17
18   You should have received a copy of the GNU General Public License
19   along with this program; see the file COPYING.
20   If not, write to the Free Software Foundation,
21   59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.  */
22
23#include "diff.h"
24#include <error.h>
25#include <exclude.h>
26#include <setjmp.h>
27#include <xalloc.h>
28
29/* Read the directory named by DIR and store into DIRDATA a sorted vector
30   of filenames for its contents.  DIR->desc == -1 means this directory is
31   known to be nonexistent, so set DIRDATA to an empty vector.
32   Return -1 (setting errno) if error, 0 otherwise.  */
33
34struct dirdata
35{
36  size_t nnames;	/* Number of names.  */
37  char const **names;	/* Sorted names of files in dir, followed by 0.  */
38  char *data;	/* Allocated storage for file names.  */
39};
40
41/* Whether file names in directories should be compared with strcoll.  */
42static bool locale_specific_sorting;
43
44/* Where to go if strcoll fails.  */
45static jmp_buf failed_strcoll;
46
47static bool dir_loop (struct comparison const *, int);
48static int compare_names_for_qsort (void const *, void const *);
49
50
51/* Read a directory and get its vector of names.  */
52
53static bool
54dir_read (struct file_data const *dir, struct dirdata *dirdata)
55{
56  register struct dirent *next;
57  register size_t i;
58
59  /* Address of block containing the files that are described.  */
60  char const **names;
61
62  /* Number of files in directory.  */
63  size_t nnames;
64
65  /* Allocated and used storage for file name data.  */
66  char *data;
67  size_t data_alloc, data_used;
68
69  dirdata->names = 0;
70  dirdata->data = 0;
71  nnames = 0;
72  data = 0;
73
74  if (dir->desc != -1)
75    {
76      /* Open the directory and check for errors.  */
77      register DIR *reading = opendir (dir->name);
78      if (!reading)
79	return 0;
80
81      /* Initialize the table of filenames.  */
82
83      data_alloc = 512;
84      data_used = 0;
85      dirdata->data = data = xmalloc (data_alloc);
86
87      /* Read the directory entries, and insert the subfiles
88	 into the `data' table.  */
89
90      while ((errno = 0, (next = readdir (reading)) != 0))
91	{
92	  char *d_name = next->d_name;
93	  size_t d_size = NAMLEN (next) + 1;
94
95	  /* Ignore "." and "..".  */
96	  if (d_name[0] == '.'
97	      && (d_name[1] == 0 || (d_name[1] == '.' && d_name[2] == 0)))
98	    continue;
99
100	  if (excluded_filename (excluded, d_name))
101	    continue;
102
103	  while (data_alloc < data_used + d_size)
104	    {
105	      if (PTRDIFF_MAX / 2 <= data_alloc)
106		xalloc_die ();
107	      dirdata->data = data = xrealloc (data, data_alloc *= 2);
108	    }
109
110	  memcpy (data + data_used, d_name, d_size);
111	  data_used += d_size;
112	  nnames++;
113	}
114      if (errno)
115	{
116	  int e = errno;
117	  closedir (reading);
118	  errno = e;
119	  return 0;
120	}
121#if CLOSEDIR_VOID
122      closedir (reading);
123#else
124      if (closedir (reading) != 0)
125	return 0;
126#endif
127    }
128
129  /* Create the `names' table from the `data' table.  */
130  if (PTRDIFF_MAX / sizeof *names - 1 <= nnames)
131    xalloc_die ();
132  dirdata->names = names = xmalloc ((nnames + 1) * sizeof *names);
133  dirdata->nnames = nnames;
134  for (i = 0;  i < nnames;  i++)
135    {
136      names[i] = data;
137      data += strlen (data) + 1;
138    }
139  names[nnames] = 0;
140  return 1;
141}
142
143/* Compare file names, returning a value compatible with strcmp.  */
144
145static int
146compare_names (char const *name1, char const *name2)
147{
148  if (ignore_file_name_case)
149    {
150      int r = strcasecmp (name1, name2);
151      if (r)
152	return r;
153    }
154
155  if (locale_specific_sorting)
156    {
157      int r;
158      errno = 0;
159      r = strcoll (name1, name2);
160      if (errno)
161	{
162	  error (0, errno, _("cannot compare file names `%s' and `%s'"),
163		 name1, name2);
164	  longjmp (failed_strcoll, 1);
165	}
166      if (r)
167	return r;
168    }
169
170  return file_name_cmp (name1, name2);
171}
172
173/* A wrapper for compare_names suitable as an argument for qsort.  */
174
175static int
176compare_names_for_qsort (void const *file1, void const *file2)
177{
178  char const *const *f1 = file1;
179  char const *const *f2 = file2;
180  return compare_names (*f1, *f2);
181}
182
183/* Compare the contents of two directories named in CMP.
184   This is a top-level routine; it does everything necessary for diff
185   on two directories.
186
187   CMP->file[0].desc == -1 says directory CMP->file[0] doesn't exist,
188   but pretend it is empty.  Likewise for CMP->file[1].
189
190   HANDLE_FILE is a caller-provided subroutine called to handle each file.
191   It gets three operands: CMP, name of file in dir 0, name of file in dir 1.
192   These names are relative to the original working directory.
193
194   For a file that appears in only one of the dirs, one of the name-args
195   to HANDLE_FILE is zero.
196
197   Returns the maximum of all the values returned by HANDLE_FILE,
198   or EXIT_TROUBLE if trouble is encountered in opening files.  */
199
200int
201diff_dirs (struct comparison const *cmp,
202	   int (*handle_file) (struct comparison const *,
203			       char const *, char const *))
204{
205  struct dirdata dirdata[2];
206  int volatile val = EXIT_SUCCESS;
207  int i;
208
209  if ((cmp->file[0].desc == -1 || dir_loop (cmp, 0))
210      && (cmp->file[1].desc == -1 || dir_loop (cmp, 1)))
211    {
212      error (0, 0, "%s: recursive directory loop",
213	     cmp->file[cmp->file[0].desc == -1].name);
214      return EXIT_TROUBLE;
215    }
216
217  /* Get contents of both dirs.  */
218  for (i = 0; i < 2; i++)
219    if (! dir_read (&cmp->file[i], &dirdata[i]))
220      {
221	perror_with_name (cmp->file[i].name);
222	val = EXIT_TROUBLE;
223      }
224
225  if (val == EXIT_SUCCESS)
226    {
227      char const **volatile names[2];
228      names[0] = dirdata[0].names;
229      names[1] = dirdata[1].names;
230
231      /* Use locale-specific sorting if possible, else native byte order.  */
232      locale_specific_sorting = 1;
233      if (setjmp (failed_strcoll))
234	locale_specific_sorting = 0;
235
236      /* Sort the directories.  */
237      for (i = 0; i < 2; i++)
238	qsort (names[i], dirdata[i].nnames, sizeof *dirdata[i].names,
239	       compare_names_for_qsort);
240
241      /* If `-S name' was given, and this is the topmost level of comparison,
242	 ignore all file names less than the specified starting name.  */
243
244      if (starting_file && ! cmp->parent)
245	{
246	  while (*names[0] && compare_names (*names[0], starting_file) < 0)
247	    names[0]++;
248	  while (*names[1] && compare_names (*names[1], starting_file) < 0)
249	    names[1]++;
250	}
251
252      /* Loop while files remain in one or both dirs.  */
253      while (*names[0] || *names[1])
254	{
255	  /* Compare next name in dir 0 with next name in dir 1.
256	     At the end of a dir,
257	     pretend the "next name" in that dir is very large.  */
258	  int nameorder = (!*names[0] ? 1 : !*names[1] ? -1
259			   : compare_names (*names[0], *names[1]));
260	  int v1 = (*handle_file) (cmp,
261				   0 < nameorder ? 0 : *names[0]++,
262				   nameorder < 0 ? 0 : *names[1]++);
263	  if (val < v1)
264	    val = v1;
265	}
266    }
267
268  for (i = 0; i < 2; i++)
269    {
270      if (dirdata[i].names)
271	free (dirdata[i].names);
272      if (dirdata[i].data)
273	free (dirdata[i].data);
274    }
275
276  return val;
277}
278
279/* Return nonzero if CMP is looping recursively in argument I.  */
280
281static bool
282dir_loop (struct comparison const *cmp, int i)
283{
284  struct comparison const *p = cmp;
285  while ((p = p->parent))
286    if (0 < same_file (&p->file[i].stat, &cmp->file[i].stat))
287      return 1;
288  return 0;
289}
290