Added initial support for preset dictionary for raw LZMA1

and LZMA2. It is not supported by the .xz format or the xz
command line tool yet.

src/liblzma/lz/lz_decoder.c

@@ -210,7 +210,7 @@ lzma_lz_decoder_init(lzma_next_coder *next, lzma_allocator *allocator,
 		const lzma_filter_info *filters,
 		lzma_ret (*lz_init)(lzma_lz_decoder *lz,
 			lzma_allocator *allocator, const void *options,
-			size_t *dict_size))
+			lzma_lz_options *lz_options))
 {
 	// Allocate the base structure if it isn't already allocated.
 	if (next->coder == NULL) {
@@ -229,17 +229,17 @@ lzma_lz_decoder_init(lzma_next_coder *next, lzma_allocator *allocator,
 
 	// Allocate and initialize the LZ-based decoder. It will also give
 	// us the dictionary size.
-	size_t dict_size;
+	lzma_lz_options lz_options;
 	return_if_error(lz_init(&next->coder->lz, allocator,
-			filters[0].options, &dict_size));
+			filters[0].options, &lz_options));
 
 	// If the dictionary size is very small, increase it to 4096 bytes.
 	// This is to prevent constant wrapping of the dictionary, which
 	// would slow things down. The downside is that since we don't check
 	// separately for the real dictionary size, we may happily accept
 	// corrupt files.
-	if (dict_size < 4096)
+	if (lz_options.dict_size < 4096)
-		dict_size = 4096;
+		lz_options.dict_size = 4096;
 
 	// Make dictionary size a multipe of 16. Some LZ-based decoders like
 	// LZMA use the lowest bits lzma_dict.pos to know the alignment of the
@@ -248,23 +248,38 @@ lzma_lz_decoder_init(lzma_next_coder *next, lzma_allocator *allocator,
 	// recommended to give aligned buffers to liblzma.
 	//
 	// Avoid integer overflow.
-	if (dict_size > SIZE_MAX - 15)
+	if (lz_options.dict_size > SIZE_MAX - 15)
 		return LZMA_MEM_ERROR;
 
-	dict_size = (dict_size + 15) & ~((size_t)(15));
+	lz_options.dict_size = (lz_options.dict_size + 15) & ~((size_t)(15));
 
 	// Allocate and initialize the dictionary.
-	if (next->coder->dict.size != dict_size) {
+	if (next->coder->dict.size != lz_options.dict_size) {
 		lzma_free(next->coder->dict.buf, allocator);
-		next->coder->dict.buf = lzma_alloc(dict_size, allocator);
+		next->coder->dict.buf
+				= lzma_alloc(lz_options.dict_size, allocator);
 		if (next->coder->dict.buf == NULL)
 			return LZMA_MEM_ERROR;
 
-		next->coder->dict.size = dict_size;
+		next->coder->dict.size = lz_options.dict_size;
 	}
 
 	lz_decoder_reset(next->coder);
 
+	// Use the preset dictionary if it was given to us.
+	if (lz_options.preset_dict != NULL
+			&& lz_options.preset_dict_size > 0) {
+		// If the preset dictionary is bigger than the actual
+		// dictionary, copy only the tail.
+		const size_t copy_size = MIN(lz_options.preset_dict_size,
+				lz_options.dict_size);
+		const size_t offset = lz_options.preset_dict_size - copy_size;
+		memcpy(next->coder->dict.buf, lz_options.preset_dict + offset,
+				copy_size);
+		next->coder->dict.pos = copy_size;
+		next->coder->dict.full = copy_size;
+	}
+
 	// Miscellaneous initializations
 	next->coder->next_finished = false;
 	next->coder->this_finished = false;

src/liblzma/lz/lz_decoder.h

@@ -51,6 +51,13 @@ typedef struct {
 } lzma_dict;
 
 
+typedef struct {
+	size_t dict_size;
+	const uint8_t *preset_dict;
+	size_t preset_dict_size;
+} lzma_lz_options;
+
+
 typedef struct {
 	/// Data specific to the LZ-based decoder
 	lzma_coder *coder;
@@ -86,7 +93,7 @@ extern lzma_ret lzma_lz_decoder_init(lzma_next_coder *next,
 		lzma_allocator *allocator, const lzma_filter_info *filters,
 		lzma_ret (*lz_init)(lzma_lz_decoder *lz,
 			lzma_allocator *allocator, const void *options,
-			size_t *dict_size));
+			lzma_lz_options *lz_options));
 
 extern uint64_t lzma_lz_decoder_memusage(size_t dictionary_size);
 

src/liblzma/lz/lz_encoder.c

@@ -363,7 +363,8 @@ lz_encoder_prepare(lzma_mf *mf, lzma_allocator *allocator,
 
 
 static bool
-lz_encoder_init(lzma_mf *mf, lzma_allocator *allocator)
+lz_encoder_init(lzma_mf *mf, lzma_allocator *allocator,
+		const lzma_lz_options *lz_options)
 {
 	// Allocate the history buffer.
 	if (mf->buffer == NULL) {
@@ -421,6 +422,19 @@ lz_encoder_init(lzma_mf *mf, lzma_allocator *allocator)
 	// we avoid wasting RAM and improve initialization speed a lot.
 	//memzero(mf->son, (size_t)(mf->sons_count) * sizeof(uint32_t));
 
+	// Handle preset dictionary.
+	if (lz_options->preset_dict != NULL
+			&& lz_options->preset_dict_size > 0) {
+		// If the preset dictionary is bigger than the actual
+		// dictionary, use only the tail.
+		mf->write_pos = MIN(lz_options->preset_dict_size, mf->size);
+		memcpy(mf->buffer, lz_options->preset_dict
+				+ lz_options->preset_dict_size - mf->write_pos,
+				mf->write_pos);
+		mf->action = LZMA_SYNC_FLUSH;
+		mf->skip(mf, mf->write_pos);
+	}
+
 	mf->action = LZMA_RUN;
 
 	return false;
@@ -509,7 +523,7 @@ lzma_lz_encoder_init(lzma_next_coder *next, lzma_allocator *allocator,
 
 	// Allocate new buffers if needed, and do the rest of
 	// the initialization.
-	if (lz_encoder_init(&next->coder->mf, allocator))
+	if (lz_encoder_init(&next->coder->mf, allocator, &lz_options))
 		return LZMA_MEM_ERROR;
 
 	// Initialize the next filter in the chain, if any.

src/liblzma/lzma/lzma2_decoder.c

@@ -230,7 +230,7 @@ lzma2_decoder_end(lzma_coder *coder, lzma_allocator *allocator)
 
 static lzma_ret
 lzma2_decoder_init(lzma_lz_decoder *lz, lzma_allocator *allocator,
-		const void *options, size_t *dict_size)
+		const void *opt, lzma_lz_options *lz_options)
 {
 	if (lz->coder == NULL) {
 		lz->coder = lzma_alloc(sizeof(lzma_coder), allocator);
@@ -243,12 +243,15 @@ lzma2_decoder_init(lzma_lz_decoder *lz, lzma_allocator *allocator,
 		lz->coder->lzma = LZMA_LZ_DECODER_INIT;
 	}
 
+	const lzma_options_lzma *options = opt;
+
 	lz->coder->sequence = SEQ_CONTROL;
 	lz->coder->need_properties = true;
-	lz->coder->need_dictionary_reset = true;
+	lz->coder->need_dictionary_reset = options->preset_dict == NULL
+			|| options->preset_dict_size == 0;
 
 	return lzma_lzma_decoder_create(&lz->coder->lzma,
-			allocator, options, dict_size);
+			allocator, options, lz_options);
 }
 
 

src/liblzma/lzma/lzma2_encoder.c

@@ -318,15 +318,17 @@ lzma2_encoder_init(lzma_lz_encoder *lz, lzma_allocator *allocator,
 		lz->coder->lzma = NULL;
 	}
 
-	lz->coder->sequence = SEQ_INIT;
-	lz->coder->need_properties = true;
-	lz->coder->need_state_reset = false;
-	lz->coder->need_dictionary_reset = true;
-
 	lz->coder->opt_cur = *(const lzma_options_lzma *)(options);
 	lz->coder->opt_new = lz->coder->opt_cur.persistent
 				? options : NULL;
 
+	lz->coder->sequence = SEQ_INIT;
+	lz->coder->need_properties = true;
+	lz->coder->need_state_reset = false;
+	lz->coder->need_dictionary_reset
+			= lz->coder->opt_cur.preset_dict == NULL
+			|| lz->coder->opt_cur.preset_dict_size == 0;
+
 	// Initialize LZMA encoder
 	return_if_error(lzma_lzma_encoder_create(&lz->coder->lzma, allocator,
 			&lz->coder->opt_cur, lz_options));

src/liblzma/lzma/lzma_decoder.c

@@ -941,7 +941,7 @@ lzma_decoder_reset(lzma_coder *coder, const void *opt)
 
 extern lzma_ret
 lzma_lzma_decoder_create(lzma_lz_decoder *lz, lzma_allocator *allocator,
-		const void *opt, size_t *dict_size)
+		const void *opt, lzma_lz_options *lz_options)
 {
 	if (lz->coder == NULL) {
 		lz->coder = lzma_alloc(sizeof(lzma_coder), allocator);
@@ -956,7 +956,9 @@ lzma_lzma_decoder_create(lzma_lz_decoder *lz, lzma_allocator *allocator,
 	// All dictionary sizes are OK here. LZ decoder will take care of
 	// the special cases.
 	const lzma_options_lzma *options = opt;
-	*dict_size = options->dict_size;
+	lz_options->dict_size = options->dict_size;
+	lz_options->preset_dict = options->preset_dict;
+	lz_options->preset_dict_size = options->preset_dict_size;
 
 	return LZMA_OK;
 }
@@ -967,13 +969,13 @@ lzma_lzma_decoder_create(lzma_lz_decoder *lz, lzma_allocator *allocator,
 /// the LZ initialization).
 static lzma_ret
 lzma_decoder_init(lzma_lz_decoder *lz, lzma_allocator *allocator,
-		const void *options, size_t *dict_size)
+		const void *options, lzma_lz_options *lz_options)
 {
 	if (!is_lclppb_valid(options))
 		return LZMA_PROG_ERROR;
 
 	return_if_error(lzma_lzma_decoder_create(
-			lz, allocator, options, dict_size));
+			lz, allocator, options, lz_options));
 
 	lzma_decoder_reset(lz->coder, options);
 	lzma_decoder_uncompressed(lz->coder, LZMA_VLI_UNKNOWN);

src/liblzma/lzma/lzma_decoder.h

@@ -48,7 +48,7 @@ extern bool lzma_lzma_lclppb_decode(
 /// LZMA2 decoders.
 extern lzma_ret lzma_lzma_decoder_create(
 		lzma_lz_decoder *lz, lzma_allocator *allocator,
-		const void *opt, size_t *dict_size);
+		const void *opt, lzma_lz_options *lz_options);
 
 /// Gets memory usage without validating lc/lp/pb. This is used by LZMA2
 /// decoder, because raw LZMA2 decoding doesn't need lc/lp/pb.

src/liblzma/lzma/lzma_encoder.c

@@ -274,6 +274,8 @@ encode_symbol(lzma_coder *coder, lzma_mf *mf,
 static bool
 encode_init(lzma_coder *coder, lzma_mf *mf)
 {
+	assert(mf_position(mf) == 0);
+
 	if (mf->read_pos == mf->read_limit) {
 		if (mf->action == LZMA_RUN)
 			return false; // We cannot do anything.
@@ -594,7 +596,12 @@ lzma_lzma_encoder_create(lzma_coder **coder_ptr, lzma_allocator *allocator,
 			return LZMA_OPTIONS_ERROR;
 	}
 
-	coder->is_initialized = false;
+	// We don't need to write the first byte as literal if there is
+	// a non-empty preset dictionary. encode_init() wouldn't even work
+	// if there is a non-empty preset dictionary, because encode_init()
+	// assumes that position is zero and previous byte is also zero.
+	coder->is_initialized = options->preset_dict != NULL
+			&& options->preset_dict_size > 0;
 	coder->is_flushed = false;
 
 	set_lz_options(lz_options, options);