Add SPDX license identifiers to files under tests/ossfuzz
[xz/debian.git] / doc / examples / 02_decompress.c
bloba87a5d3ece2ed6edbc6e01e77b6e48a112867e47
1 // SPDX-License-Identifier: 0BSD
3 ///////////////////////////////////////////////////////////////////////////////
4 //
5 /// \file 02_decompress.c
6 /// \brief Decompress .xz files to stdout
7 ///
8 /// Usage: ./02_decompress INPUT_FILES... > OUTFILE
9 ///
10 /// Example: ./02_decompress foo.xz bar.xz > foobar
12 // Author: Lasse Collin
14 ///////////////////////////////////////////////////////////////////////////////
16 #include <stdbool.h>
17 #include <stdlib.h>
18 #include <stdio.h>
19 #include <string.h>
20 #include <errno.h>
21 #include <lzma.h>
24 static bool
25 init_decoder(lzma_stream *strm)
27 // Initialize a .xz decoder. The decoder supports a memory usage limit
28 // and a set of flags.
30 // The memory usage of the decompressor depends on the settings used
31 // to compress a .xz file. It can vary from less than a megabyte to
32 // a few gigabytes, but in practice (at least for now) it rarely
33 // exceeds 65 MiB because that's how much memory is required to
34 // decompress files created with "xz -9". Settings requiring more
35 // memory take extra effort to use and don't (at least for now)
36 // provide significantly better compression in most cases.
38 // Memory usage limit is useful if it is important that the
39 // decompressor won't consume gigabytes of memory. The need
40 // for limiting depends on the application. In this example,
41 // no memory usage limiting is used. This is done by setting
42 // the limit to UINT64_MAX.
44 // The .xz format allows concatenating compressed files as is:
46 // echo foo | xz > foobar.xz
47 // echo bar | xz >> foobar.xz
49 // When decompressing normal standalone .xz files, LZMA_CONCATENATED
50 // should always be used to support decompression of concatenated
51 // .xz files. If LZMA_CONCATENATED isn't used, the decoder will stop
52 // after the first .xz stream. This can be useful when .xz data has
53 // been embedded inside another file format.
55 // Flags other than LZMA_CONCATENATED are supported too, and can
56 // be combined with bitwise-or. See lzma/container.h
57 // (src/liblzma/api/lzma/container.h in the source package or e.g.
58 // /usr/include/lzma/container.h depending on the install prefix)
59 // for details.
60 lzma_ret ret = lzma_stream_decoder(
61 strm, UINT64_MAX, LZMA_CONCATENATED);
63 // Return successfully if the initialization went fine.
64 if (ret == LZMA_OK)
65 return true;
67 // Something went wrong. The possible errors are documented in
68 // lzma/container.h (src/liblzma/api/lzma/container.h in the source
69 // package or e.g. /usr/include/lzma/container.h depending on the
70 // install prefix).
72 // Note that LZMA_MEMLIMIT_ERROR is never possible here. If you
73 // specify a very tiny limit, the error will be delayed until
74 // the first headers have been parsed by a call to lzma_code().
75 const char *msg;
76 switch (ret) {
77 case LZMA_MEM_ERROR:
78 msg = "Memory allocation failed";
79 break;
81 case LZMA_OPTIONS_ERROR:
82 msg = "Unsupported decompressor flags";
83 break;
85 default:
86 // This is most likely LZMA_PROG_ERROR indicating a bug in
87 // this program or in liblzma. It is inconvenient to have a
88 // separate error message for errors that should be impossible
89 // to occur, but knowing the error code is important for
90 // debugging. That's why it is good to print the error code
91 // at least when there is no good error message to show.
92 msg = "Unknown error, possibly a bug";
93 break;
96 fprintf(stderr, "Error initializing the decoder: %s (error code %u)\n",
97 msg, ret);
98 return false;
102 static bool
103 decompress(lzma_stream *strm, const char *inname, FILE *infile, FILE *outfile)
105 // When LZMA_CONCATENATED flag was used when initializing the decoder,
106 // we need to tell lzma_code() when there will be no more input.
107 // This is done by setting action to LZMA_FINISH instead of LZMA_RUN
108 // in the same way as it is done when encoding.
110 // When LZMA_CONCATENATED isn't used, there is no need to use
111 // LZMA_FINISH to tell when all the input has been read, but it
112 // is still OK to use it if you want. When LZMA_CONCATENATED isn't
113 // used, the decoder will stop after the first .xz stream. In that
114 // case some unused data may be left in strm->next_in.
115 lzma_action action = LZMA_RUN;
117 uint8_t inbuf[BUFSIZ];
118 uint8_t outbuf[BUFSIZ];
120 strm->next_in = NULL;
121 strm->avail_in = 0;
122 strm->next_out = outbuf;
123 strm->avail_out = sizeof(outbuf);
125 while (true) {
126 if (strm->avail_in == 0 && !feof(infile)) {
127 strm->next_in = inbuf;
128 strm->avail_in = fread(inbuf, 1, sizeof(inbuf),
129 infile);
131 if (ferror(infile)) {
132 fprintf(stderr, "%s: Read error: %s\n",
133 inname, strerror(errno));
134 return false;
137 // Once the end of the input file has been reached,
138 // we need to tell lzma_code() that no more input
139 // will be coming. As said before, this isn't required
140 // if the LZMA_CONCATENATED flag isn't used when
141 // initializing the decoder.
142 if (feof(infile))
143 action = LZMA_FINISH;
146 lzma_ret ret = lzma_code(strm, action);
148 if (strm->avail_out == 0 || ret == LZMA_STREAM_END) {
149 size_t write_size = sizeof(outbuf) - strm->avail_out;
151 if (fwrite(outbuf, 1, write_size, outfile)
152 != write_size) {
153 fprintf(stderr, "Write error: %s\n",
154 strerror(errno));
155 return false;
158 strm->next_out = outbuf;
159 strm->avail_out = sizeof(outbuf);
162 if (ret != LZMA_OK) {
163 // Once everything has been decoded successfully, the
164 // return value of lzma_code() will be LZMA_STREAM_END.
166 // It is important to check for LZMA_STREAM_END. Do not
167 // assume that getting ret != LZMA_OK would mean that
168 // everything has gone well or that when you aren't
169 // getting more output it must have successfully
170 // decoded everything.
171 if (ret == LZMA_STREAM_END)
172 return true;
174 // It's not LZMA_OK nor LZMA_STREAM_END,
175 // so it must be an error code. See lzma/base.h
176 // (src/liblzma/api/lzma/base.h in the source package
177 // or e.g. /usr/include/lzma/base.h depending on the
178 // install prefix) for the list and documentation of
179 // possible values. Many values listen in lzma_ret
180 // enumeration aren't possible in this example, but
181 // can be made possible by enabling memory usage limit
182 // or adding flags to the decoder initialization.
183 const char *msg;
184 switch (ret) {
185 case LZMA_MEM_ERROR:
186 msg = "Memory allocation failed";
187 break;
189 case LZMA_FORMAT_ERROR:
190 // .xz magic bytes weren't found.
191 msg = "The input is not in the .xz format";
192 break;
194 case LZMA_OPTIONS_ERROR:
195 // For example, the headers specify a filter
196 // that isn't supported by this liblzma
197 // version (or it hasn't been enabled when
198 // building liblzma, but no-one sane does
199 // that unless building liblzma for an
200 // embedded system). Upgrading to a newer
201 // liblzma might help.
203 // Note that it is unlikely that the file has
204 // accidentally became corrupt if you get this
205 // error. The integrity of the .xz headers is
206 // always verified with a CRC32, so
207 // unintentionally corrupt files can be
208 // distinguished from unsupported files.
209 msg = "Unsupported compression options";
210 break;
212 case LZMA_DATA_ERROR:
213 msg = "Compressed file is corrupt";
214 break;
216 case LZMA_BUF_ERROR:
217 // Typically this error means that a valid
218 // file has got truncated, but it might also
219 // be a damaged part in the file that makes
220 // the decoder think the file is truncated.
221 // If you prefer, you can use the same error
222 // message for this as for LZMA_DATA_ERROR.
223 msg = "Compressed file is truncated or "
224 "otherwise corrupt";
225 break;
227 default:
228 // This is most likely LZMA_PROG_ERROR.
229 msg = "Unknown error, possibly a bug";
230 break;
233 fprintf(stderr, "%s: Decoder error: "
234 "%s (error code %u)\n",
235 inname, msg, ret);
236 return false;
242 extern int
243 main(int argc, char **argv)
245 if (argc <= 1) {
246 fprintf(stderr, "Usage: %s FILES...\n", argv[0]);
247 return EXIT_FAILURE;
250 lzma_stream strm = LZMA_STREAM_INIT;
252 bool success = true;
254 // Try to decompress all files.
255 for (int i = 1; i < argc; ++i) {
256 if (!init_decoder(&strm)) {
257 // Decoder initialization failed. There's no point
258 // to retry it so we need to exit.
259 success = false;
260 break;
263 FILE *infile = fopen(argv[i], "rb");
265 if (infile == NULL) {
266 fprintf(stderr, "%s: Error opening the "
267 "input file: %s\n",
268 argv[i], strerror(errno));
269 success = false;
270 } else {
271 success &= decompress(&strm, argv[i], infile, stdout);
272 fclose(infile);
276 // Free the memory allocated for the decoder. This only needs to be
277 // done after the last file.
278 lzma_end(&strm);
280 if (fclose(stdout)) {
281 fprintf(stderr, "Write error: %s\n", strerror(errno));
282 success = false;
285 return success ? EXIT_SUCCESS : EXIT_FAILURE;