X-Git-Url: https://wimlib.net/git/?a=blobdiff_plain;f=src%2Fcompress.c;h=79fe4ddf54e320aeeac7da13e8110a6ada23944c;hb=7bac2f53ecca0fe8d6f9e55a9bf8f86512256660;hp=76043d632b98f6f3e670c948f5f09b8d62521d90;hpb=4ecf344e77e1f5891055881950a6e89e32b16008;p=wimlib

diff --git a/src/compress.c b/src/compress.c
index 76043d63..79fe4ddf 100644
--- a/src/compress.c
+++ b/src/compress.c
@@ -28,6 +28,7 @@
 #endif
 
 #include "wimlib/assert.h"
+#include "wimlib/endianness.h"
 #include "wimlib/compiler.h"
 #include "wimlib/compress.h"
 #include "wimlib/util.h"
@@ -121,12 +122,12 @@ init_output_bitstream(struct output_bitstream *ostream,
 	ostream->bit_output          = data;
 	ostream->next_bit_output     = data + 2;
 	ostream->output              = data + 4;
-	ostream->bytes_remaining     = num_bytes;
+	ostream->bytes_remaining     = num_bytes - 4;
 	ostream->overrun             = false;
 }
 
 typedef struct {
-	freq_t freq;
+	input_idx_t freq;
 	u16 sym;
 	union {
 		u16 path_len;
@@ -231,26 +232,26 @@ huffman_tree_compute_path_lengths(HuffmanNode *base_node, u16 cur_len)
  * @num_syms:  The number of symbols in the alphabet.
  *
  * @max_codeword_len:  The maximum allowed length of a codeword in the code.
- * 			Note that if the code being created runs up against
- * 			this restriction, the code ultimately created will be
- * 			suboptimal, although there are some advantages for
- * 			limiting the length of the codewords.
+ *			Note that if the code being created runs up against
+ *			this restriction, the code ultimately created will be
+ *			suboptimal, although there are some advantages for
+ *			limiting the length of the codewords.
  *
  * @freq_tab:  An array of length @num_syms that contains the frequencies
- * 			of each symbol in the uncompressed data.
+ *			of each symbol in the uncompressed data.
  *
  * @lens:	   An array of length @num_syms into which the lengths of the
- * 			codewords for each symbol will be written.
+ *			codewords for each symbol will be written.
  *
  * @codewords:	   An array of @num_syms short integers into which the
- * 			codewords for each symbol will be written.  The first
- * 			lens[i] bits of codewords[i] will contain the codeword
- * 			for symbol i.
+ *			codewords for each symbol will be written.  The first
+ *			lens[i] bits of codewords[i] will contain the codeword
+ *			for symbol i.
  */
 void
 make_canonical_huffman_code(unsigned num_syms,
 			    unsigned max_codeword_len,
-			    const freq_t freq_tab[restrict],
+			    const input_idx_t freq_tab[restrict],
 			    u8 lens[restrict],
 			    u16 codewords[restrict])
 {
@@ -427,8 +428,8 @@ try_building_tree_again:
 			 * log_2(num_used_symbols).
 			 * */
 			for (unsigned i = 0; i < num_used_symbols; i++)
-				if (leaves[i].freq > 1)
-					leaves[i].freq >>= 1;
+				leaves[i].freq = (leaves[i].freq + 1) >> 1;
+
 			goto try_building_tree_again;
 		}
 		next_inode++;