Sort of garbage collection commit. :-| Many things are still

broken. API has changed a lot and it will still change a
little more here and there. The command line tool doesn't
have all the required changes to reflect the API changes, so
it's easy to get "internal error" or trigger assertions.

1 files changed, 60 insertions, 97 deletions

diff --git a/src/lzmadec/lzmadec.c b/src/lzmadec/lzmadec.c
index 1fc561b7..ed5947ad 100644
--- a/src/lzmadec/lzmadec.c
+++ b/src/lzmadec/lzmadec.c
@@ -65,14 +65,7 @@ static uint8_t out_buf[BUFSIZ];
 static lzma_stream strm = LZMA_STREAM_INIT;
 
 /// Number of bytes to use memory at maximum
-static size_t mem_limit;
-
-/// Memory allocation hooks
-static lzma_allocator allocator = {
-	.alloc = (void *(*)(void *, size_t, size_t))(&lzma_memlimit_alloc),
-	.free = (void (*)(void *, void *))(&lzma_memlimit_free),
-	.opaque = NULL,
-};
+static size_t memlimit;
 
 /// Program name to be shown in error messages
 static const char *argv0;
@@ -116,7 +109,7 @@ help(void)
 		" MiB of memory at maximum.\n"
 "\n"
 "Report bugs to <" PACKAGE_BUGREPORT "> (in English or Finnish).\n",
-		argv0, ((uint64_t)(mem_limit) + 512 * 1024) / (1024 * 1024));
+		argv0, ((uint64_t)(memlimit) + 512 * 1024) / (1024 * 1024));
 		// Using PRIu64 above instead of %zu to support pre-C99 libc.
 	exit(0);
 }
@@ -148,7 +141,7 @@ version(void)
 /// Finds out the amount of physical memory in the system, and sets
 /// a default memory usage limit.
 static void
-set_default_mem_limit(void)
+set_default_memlimit(void)
 {
 	uint64_t mem = physmem();
 	if (mem != 0) {
@@ -159,10 +152,10 @@ set_default_mem_limit(void)
 			mem = SIZE_MAX;
 #endif
 
-		mem_limit = mem / 3;
+		memlimit = mem / 3;
 	} else {
 		// Cannot autodetect, use 10 MiB as the default limit.
-		mem_limit = (1U << 23) + (1U << 21);
+		memlimit = (1U << 23) + (1U << 21);
 	}
 
 	return;
@@ -272,7 +265,7 @@ parse_options(int argc, char **argv)
 			break;
 
 		case 'M':
-			mem_limit = str_to_size(optarg);
+			memlimit = str_to_size(optarg);
 			break;
 
 		case 'h':
@@ -309,19 +302,20 @@ parse_options(int argc, char **argv)
 static void
 init(void)
 {
+	const uint32_t flags = LZMA_WARN_UNSUPPORTED_CHECK | LZMA_CONCATENATED;
 	lzma_ret ret;
 
 	switch (format_type) {
 	case FORMAT_AUTO:
-		ret = lzma_auto_decoder(&strm);
+		ret = lzma_auto_decoder(&strm, memlimit, flags);
 		break;
 
 	case FORMAT_NATIVE:
-		ret = lzma_stream_decoder(&strm);
+		ret = lzma_stream_decoder(&strm, memlimit, flags);
 		break;
 
 	case FORMAT_ALONE:
-		ret = lzma_alone_decoder(&strm);
+		ret = lzma_alone_decoder(&strm, memlimit);
 		break;
 
 	default:
@@ -345,50 +339,6 @@ init(void)
 
 
 static void
-read_input(void)
-{
-	strm.next_in = in_buf;
-	strm.avail_in = fread(in_buf, 1, BUFSIZ, file);
-
-	if (ferror(file)) {
-		// POSIX says that fread() sets errno if an error occurred.
-		// ferror() doesn't touch errno.
-		fprintf(stderr, "%s: %s: Error reading input file: %s\n",
-				argv0, filename, strerror(errno));
-		exit(ERROR);
-	}
-
-	return;
-}
-
-
-static bool
-skip_padding(void)
-{
-	// Handle concatenated Streams. There can be arbitrary number of
-	// nul-byte padding between the Streams, which must be ignored.
-	//
-	// NOTE: Concatenating LZMA_Alone files works only if at least
-	// one of lc, lp, and pb is non-zero. Using the concatenation
-	// on LZMA_Alone files is strongly discouraged.
-	while (true) {
-		while (strm.avail_in > 0) {
-			if (*strm.next_in != '\0')
-				return true;
-
-			++strm.next_in;
-			--strm.avail_in;
-		}
-
-		if (feof(file))
-			return false;
-
-		read_input();
-	}
-}
-
-
-static void
 uncompress(void)
 {
 	if (file == stdin && !force && isatty(STDIN_FILENO)) {
@@ -400,44 +350,65 @@ uncompress(void)
 	}
 
 	init();
+
 	strm.avail_in = 0;
+	strm.next_out = out_buf;
+	strm.avail_out = BUFSIZ;
+
+	lzma_action action = LZMA_RUN;
 
 	while (true) {
-		if (strm.avail_in == 0)
-			read_input();
+		if (strm.avail_in == 0) {
+			strm.next_in = in_buf;
+			strm.avail_in = fread(in_buf, 1, BUFSIZ, file);
+
+			if (ferror(file)) {
+				// POSIX says that fread() sets errno if
+				// an error occurred. ferror() doesn't
+				// touch errno.
+				fprintf(stderr, "%s: %s: Error reading "
+						"input file: %s\n",
+						argv0, filename,
+						strerror(errno));
+				exit(ERROR);
+			}
 
-		strm.next_out = out_buf;
-		strm.avail_out = BUFSIZ;
+			if (feof(file))
+				action = LZMA_FINISH;
+		}
 
-		const lzma_ret ret = lzma_code(&strm, LZMA_RUN);
+		const lzma_ret ret = lzma_code(&strm, action);
 
 		// Write and check write error before checking decoder error.
 		// This way as much data as possible gets written to output
-		// even if decoder detected an error. Checking write error
-		// needs to be done before checking decoder error due to
-		// how concatenated Streams are handled a few lines later.
-		const size_t write_size = BUFSIZ - strm.avail_out;
-		if (fwrite(out_buf, 1, write_size, stdout) != write_size) {
-			// Wouldn't be a surprise if writing to stderr would
-			// fail too but at least try to show an error message.
-			fprintf(stderr, "%s: Cannot write to "
-					"standard output: %s\n", argv0,
-					strerror(errno));
-			exit(ERROR);
+		// even if decoder detected an error.
+		if (strm.avail_out == 0 || ret != LZMA_OK) {
+			const size_t write_size = BUFSIZ - strm.avail_out;
+
+			if (fwrite(out_buf, 1, write_size, stdout)
+					!= write_size) {
+				// Wouldn't be a surprise if writing to stderr
+				// would fail too but at least try to show an
+				// error message.
+				fprintf(stderr, "%s: Cannot write to "
+						"standard output: %s\n", argv0,
+						strerror(errno));
+				exit(ERROR);
+			}
+
+			strm.next_out = out_buf;
+			strm.avail_out = BUFSIZ;
 		}
 
 		if (ret != LZMA_OK) {
-			if (ret == LZMA_STREAM_END) {
-				if (skip_padding()) {
-					init();
-					continue;
-				}
-
+			// FIXME !!! Doesn't work with LZMA_Alone for the
+			// same reason as in process.c.
+			if (ret == LZMA_STREAM_END)
 				return;
-			}
 
 			fprintf(stderr, "%s: %s: ", argv0, filename);
 
+			// FIXME Add LZMA_*_CHECK and LZMA_FORMAT_ERROR.
 			switch (ret) {
 			case LZMA_DATA_ERROR:
 				fprintf(stderr, "File is corrupt\n");
@@ -452,6 +423,11 @@ uncompress(void)
 				fprintf(stderr, "%s\n", strerror(ENOMEM));
 				exit(ERROR);
 
+			case LZMA_MEMLIMIT_ERROR:
+				fprintf(stderr, "Memory usage limit "
+						"reached\n");
+				exit(ERROR);
+
 			case LZMA_BUF_ERROR:
 				fprintf(stderr, "Unexpected end of input\n");
 				exit(ERROR);
@@ -479,23 +455,12 @@ main(int argc, char **argv)
 {
 	argv0 = argv[0];
 
-	set_default_mem_limit();
+	set_default_memlimit();
 
 	parse_options(argc, argv);
 
 	lzma_init_decoder();
 
-	lzma_memlimit *mem_limiter = lzma_memlimit_create(mem_limit);
-	if (mem_limiter == NULL) {
-		fprintf(stderr, "%s: %s\n", argv0, strerror(ENOMEM));
-		exit(ERROR);
-	}
-
-	assert(lzma_memlimit_count(mem_limiter) == 0);
-
-	allocator.opaque = mem_limiter;
-	strm.allocator = &allocator;
-
 #ifdef WIN32
 	setmode(fileno(stdin), O_BINARY);
 	setmode(fileno(stdout), O_BINARY);
@@ -531,8 +496,6 @@ main(int argc, char **argv)
 	// Free the memory only when debugging. Freeing wastes some time,
 	// but allows detecting possible memory leaks with Valgrind.
 	lzma_end(&strm);
-	assert(lzma_memlimit_count(mem_limiter) == 0);
-	lzma_memlimit_end(mem_limiter, false);
 #endif
 
 	return exit_status;