aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--.gitignore1
-rw-r--r--CMakeLists.txt2
-rw-r--r--src/liblzma/common/index.c1
-rw-r--r--src/liblzma/common/index.h12
-rw-r--r--src/liblzma/common/index_decoder.c2
-rw-r--r--src/liblzma/common/index_decoder.h1
-rw-r--r--src/liblzma/common/index_encoder.c2
-rw-r--r--src/liblzma/common/index_hash.c4
-rw-r--r--src/liblzma/common/stream_buffer_encoder.c1
-rw-r--r--src/liblzma/common/stream_decoder.c3
-rw-r--r--src/liblzma/common/stream_decoder_mt.c2
-rw-r--r--tests/Makefile.am3
-rw-r--r--tests/test_index_hash.c388
13 files changed, 414 insertions, 8 deletions
diff --git a/.gitignore b/.gitignore
index 44f269cc..7f2dec23 100644
--- a/.gitignore
+++ b/.gitignore
@@ -62,6 +62,7 @@ coverage
/tests/test_filter_flags
/tests/test_hardware
/tests/test_index
+/tests/test_index_hash
/test/test_lzip_decoder
/tests/test_memlimit
/tests/test_stream_flags
diff --git a/CMakeLists.txt b/CMakeLists.txt
index c4f10594..3a58d740 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -801,6 +801,7 @@ if(BUILD_TESTING)
test_filter_flags
test_hardware
test_index
+ test_index_hash
test_memlimit
test_stream_flags
test_vli
@@ -812,6 +813,7 @@ if(BUILD_TESTING)
target_include_directories("${TEST}" PRIVATE
src/common
src/liblzma/api
+ src/liblzma
lib
)
diff --git a/src/liblzma/common/index.c b/src/liblzma/common/index.c
index 24ec3c10..97cc9f95 100644
--- a/src/liblzma/common/index.c
+++ b/src/liblzma/common/index.c
@@ -10,6 +10,7 @@
//
///////////////////////////////////////////////////////////////////////////////
+#include "common.h"
#include "index.h"
#include "stream_flags_common.h"
diff --git a/src/liblzma/common/index.h b/src/liblzma/common/index.h
index 64e97247..031efcc7 100644
--- a/src/liblzma/common/index.h
+++ b/src/liblzma/common/index.h
@@ -2,6 +2,13 @@
//
/// \file index.h
/// \brief Handling of Index
+/// \note This header file does not include common.h or lzma.h because
+/// this file is needed by both liblzma internally and by the
+/// tests. Including common.h will include and define many things
+/// the tests do not need and prevents issues with header file
+/// include order. This way, if lzma.h or common.h are not
+/// included before this file it will break on every OS instead
+/// of causing more subtle errors.
//
// Author: Lasse Collin
//
@@ -13,8 +20,6 @@
#ifndef LZMA_INDEX_H
#define LZMA_INDEX_H
-#include "common.h"
-
/// Minimum Unpadded Size
#define UNPADDED_SIZE_MIN LZMA_VLI_C(5)
@@ -22,6 +27,9 @@
/// Maximum Unpadded Size
#define UNPADDED_SIZE_MAX (LZMA_VLI_MAX & ~LZMA_VLI_C(3))
+/// Index Indicator based on xz specification
+#define INDEX_INDICATOR 0
+
/// Get the size of the Index Padding field. This is needed by Index encoder
/// and decoder, but applications should have no use for this.
diff --git a/src/liblzma/common/index_decoder.c b/src/liblzma/common/index_decoder.c
index b2689885..8622b2f0 100644
--- a/src/liblzma/common/index_decoder.c
+++ b/src/liblzma/common/index_decoder.c
@@ -80,7 +80,7 @@ index_decode(void *coder_ptr, const lzma_allocator *allocator,
// format". One could argue that the application should
// verify the Index Indicator before trying to decode the
// Index, but well, I suppose it is simpler this way.
- if (in[(*in_pos)++] != 0x00)
+ if (in[(*in_pos)++] != INDEX_INDICATOR)
return LZMA_DATA_ERROR;
coder->sequence = SEQ_COUNT;
diff --git a/src/liblzma/common/index_decoder.h b/src/liblzma/common/index_decoder.h
index 1af433b5..3fec4833 100644
--- a/src/liblzma/common/index_decoder.h
+++ b/src/liblzma/common/index_decoder.h
@@ -13,6 +13,7 @@
#ifndef LZMA_INDEX_DECODER_H
#define LZMA_INDEX_DECODER_H
+#include "common.h"
#include "index.h"
diff --git a/src/liblzma/common/index_encoder.c b/src/liblzma/common/index_encoder.c
index ac97d0ce..c7cafb72 100644
--- a/src/liblzma/common/index_encoder.c
+++ b/src/liblzma/common/index_encoder.c
@@ -65,7 +65,7 @@ index_encode(void *coder_ptr,
while (*out_pos < out_size)
switch (coder->sequence) {
case SEQ_INDICATOR:
- out[*out_pos] = 0x00;
+ out[*out_pos] = INDEX_INDICATOR;
++*out_pos;
coder->sequence = SEQ_COUNT;
break;
diff --git a/src/liblzma/common/index_hash.c b/src/liblzma/common/index_hash.c
index 34df85d7..f55f7bc8 100644
--- a/src/liblzma/common/index_hash.c
+++ b/src/liblzma/common/index_hash.c
@@ -145,7 +145,7 @@ lzma_index_hash_append(lzma_index_hash *index_hash, lzma_vli unpadded_size,
lzma_vli uncompressed_size)
{
// Validate the arguments.
- if (index_hash->sequence != SEQ_BLOCK
+ if (index_hash == NULL || index_hash->sequence != SEQ_BLOCK
|| unpadded_size < UNPADDED_SIZE_MIN
|| unpadded_size > UNPADDED_SIZE_MAX
|| uncompressed_size > LZMA_VLI_MAX)
@@ -190,7 +190,7 @@ lzma_index_hash_decode(lzma_index_hash *index_hash, const uint8_t *in,
switch (index_hash->sequence) {
case SEQ_BLOCK:
// Check the Index Indicator is present.
- if (in[(*in_pos)++] != 0x00)
+ if (in[(*in_pos)++] != INDEX_INDICATOR)
return LZMA_DATA_ERROR;
index_hash->sequence = SEQ_COUNT;
diff --git a/src/liblzma/common/stream_buffer_encoder.c b/src/liblzma/common/stream_buffer_encoder.c
index af49554a..73157590 100644
--- a/src/liblzma/common/stream_buffer_encoder.c
+++ b/src/liblzma/common/stream_buffer_encoder.c
@@ -10,6 +10,7 @@
//
///////////////////////////////////////////////////////////////////////////////
+#include "common.h"
#include "index.h"
diff --git a/src/liblzma/common/stream_decoder.c b/src/liblzma/common/stream_decoder.c
index dcf7c149..64283812 100644
--- a/src/liblzma/common/stream_decoder.c
+++ b/src/liblzma/common/stream_decoder.c
@@ -12,6 +12,7 @@
#include "stream_decoder.h"
#include "block_decoder.h"
+#include "index.h"
typedef struct {
@@ -164,7 +165,7 @@ stream_decode(void *coder_ptr, const lzma_allocator *allocator,
if (coder->pos == 0) {
// Detect if it's Index.
- if (in[*in_pos] == 0x00) {
+ if (in[*in_pos] == INDEX_INDICATOR) {
coder->sequence = SEQ_INDEX;
break;
}
diff --git a/src/liblzma/common/stream_decoder_mt.c b/src/liblzma/common/stream_decoder_mt.c
index 5733c764..fd5cd7fd 100644
--- a/src/liblzma/common/stream_decoder_mt.c
+++ b/src/liblzma/common/stream_decoder_mt.c
@@ -887,7 +887,7 @@ decode_block_header(struct lzma_stream_coder *coder,
if (coder->pos == 0) {
// Detect if it's Index.
- if (in[*in_pos] == 0x00)
+ if (in[*in_pos] == INDEX_INDICATOR)
return LZMA_INDEX_DETECTED;
// Calculate the size of the Block Header. Note that
diff --git a/tests/Makefile.am b/tests/Makefile.am
index f84c1749..b1a2378a 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -26,6 +26,7 @@ EXTRA_DIST = \
AM_CPPFLAGS = \
-I$(top_srcdir)/src/common \
-I$(top_srcdir)/src/liblzma/api \
+ -I$(top_srcdir)/src/liblzma \
-I$(top_builddir)/lib
LDADD = $(top_builddir)/src/liblzma/liblzma.la
@@ -44,6 +45,7 @@ check_PROGRAMS = \
test_filter_flags \
test_block_header \
test_index \
+ test_index_hash \
test_bcj_exact_size \
test_memlimit \
test_lzip_decoder \
@@ -56,6 +58,7 @@ TESTS = \
test_filter_flags \
test_block_header \
test_index \
+ test_index_hash \
test_bcj_exact_size \
test_memlimit \
test_lzip_decoder \
diff --git a/tests/test_index_hash.c b/tests/test_index_hash.c
new file mode 100644
index 00000000..2cf91138
--- /dev/null
+++ b/tests/test_index_hash.c
@@ -0,0 +1,388 @@
+///////////////////////////////////////////////////////////////////////////////
+//
+/// \file test_index_hash.c
+/// \brief Tests src/liblzma/common/index_hash.c API functions
+///
+/// \note No test included for lzma_index_hash_end since it
+/// would be trivial unless tested for memory leaks
+/// with something like valgrind
+//
+// Author: Jia Tan
+//
+// This file has been put into the public domain.
+// You can do whatever you want with this file.
+//
+///////////////////////////////////////////////////////////////////////////////
+
+#include "tests.h"
+
+// Needed for UNPADDED_SIZE_MIN and UNPADDED_SIZE_MAX macro definitions
+// and index_size and vli_ceil4 helper functions
+#include "common/index.h"
+
+
+static void
+test_lzma_index_hash_init(void)
+{
+#ifndef HAVE_DECODERS
+ assert_skip("Decoder support disabled");
+#else
+ // First test with NULL index_hash.
+ // This should create a fresh index_hash.
+ lzma_index_hash *index_hash = lzma_index_hash_init(NULL, NULL);
+ assert_true(index_hash != NULL);
+
+ // Next test with non-NULL index_hash.
+ lzma_index_hash *second_hash = lzma_index_hash_init(index_hash, NULL);
+
+ // It should not create a new index_hash pointer.
+ // Instead it must just re-init the first index_hash.
+ assert_true(index_hash == second_hash);
+
+ lzma_index_hash_end(index_hash, NULL);
+#endif
+}
+
+
+static void
+test_lzma_index_hash_append(void)
+{
+#ifndef HAVE_DECODERS
+ assert_skip("Decoder support disabled");
+#else
+ // Test all invalid parameters
+ assert_lzma_ret(lzma_index_hash_append(NULL, 0, 0),
+ LZMA_PROG_ERROR);
+
+ // Test NULL index_hash
+ assert_lzma_ret(lzma_index_hash_append(NULL, UNPADDED_SIZE_MIN,
+ LZMA_VLI_MAX), LZMA_PROG_ERROR);
+
+ // Test with invalid Unpadded Size
+ lzma_index_hash *index_hash = lzma_index_hash_init(NULL, NULL);
+ assert_true(index_hash != NULL);
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ UNPADDED_SIZE_MIN - 1, LZMA_VLI_MAX),
+ LZMA_PROG_ERROR);
+
+ // Test with invalid Uncompressed Size
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ UNPADDED_SIZE_MIN, LZMA_VLI_MAX + 1),
+ LZMA_PROG_ERROR);
+
+ // First append a Record describing a small Block.
+ // This should succeed.
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ UNPADDED_SIZE_MIN, 1), LZMA_OK);
+
+ // Append another small Record.
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ UNPADDED_SIZE_MIN, 1), LZMA_OK);
+
+ // Append a Record that would cause the compressed size to grow
+ // too big
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ UNPADDED_SIZE_MAX, 1), LZMA_DATA_ERROR);
+
+ lzma_index_hash_end(index_hash, NULL);
+#endif
+}
+
+
+#ifdef HAVE_DECODERS
+// Fill an index_hash with unpadded and uncompressed VLIs
+// by calling lzma_index_hash_append
+static void
+fill_index_hash(lzma_index_hash *index_hash, const lzma_vli *unpadded_sizes,
+ const lzma_vli *uncomp_sizes, uint32_t block_count)
+{
+ for (uint32_t i = 0; i < block_count; ++i)
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ unpadded_sizes[i], uncomp_sizes[i]), LZMA_OK);
+}
+
+
+#ifdef HAVE_ENCODERS
+// Set the contents of buf to the expected Index based on the
+// .xz specification. This needs the unpadded and uncompressed VLIs
+// to correctly create the Index.
+static void
+generate_index(uint8_t *buf, const lzma_vli *unpadded_sizes,
+ const lzma_vli *uncomp_sizes, uint32_t block_count,
+ size_t index_max_size)
+{
+ size_t in_pos = 0;
+ size_t out_pos = 0;
+
+ // First set Index Indicator
+ buf[out_pos++] = INDEX_INDICATOR;
+
+ // Next write out Number of Records
+ assert_lzma_ret(lzma_vli_encode(block_count, &in_pos, buf,
+ &out_pos, index_max_size), LZMA_STREAM_END);
+
+ // Next write out each Record.
+ // A Record consists of Unpadded Size and Uncompressed Size
+ // written next to each other as VLIs.
+ for (uint32_t i = 0; i < block_count; ++i) {
+ in_pos = 0;
+ assert_lzma_ret(lzma_vli_encode(unpadded_sizes[i], &in_pos,
+ buf, &out_pos, index_max_size), LZMA_STREAM_END);
+ in_pos = 0;
+ assert_lzma_ret(lzma_vli_encode(uncomp_sizes[i], &in_pos,
+ buf, &out_pos, index_max_size), LZMA_STREAM_END);
+ }
+
+ // Add Index Padding
+ lzma_vli rounded_out_pos = vli_ceil4(out_pos);
+ memzero(buf + out_pos, rounded_out_pos - out_pos);
+ out_pos = rounded_out_pos;
+
+ // Add the CRC32
+ write32le(buf + out_pos, lzma_crc32(buf, out_pos, 0));
+ out_pos += 4;
+
+ assert_uint_eq(out_pos, index_max_size);
+}
+#endif
+#endif
+
+
+static void
+test_lzma_index_hash_decode(void)
+{
+#if !defined(HAVE_ENCODERS) || !defined(HAVE_DECODERS)
+ assert_skip("Encoder or decoder support disabled");
+#else
+ lzma_index_hash *index_hash = lzma_index_hash_init(NULL, NULL);
+ assert_true(index_hash != NULL);
+
+ size_t in_pos = 0;
+
+ // Six valid values for the Unpadded Size fields in an Index
+ const lzma_vli unpadded_sizes[6] = {
+ UNPADDED_SIZE_MIN,
+ 1000,
+ 4000,
+ 8000,
+ 16000,
+ 32000
+ };
+
+ // Six valid values for the Uncompressed Size fields in an Index
+ const lzma_vli uncomp_sizes[6] = {
+ 1,
+ 500,
+ 8000,
+ 20,
+ 1,
+ 500
+ };
+
+ // Add two Records to an index_hash
+ fill_index_hash(index_hash, unpadded_sizes, uncomp_sizes, 2);
+
+ const lzma_vli size_two_records = lzma_index_hash_size(index_hash);
+ assert_uint(size_two_records, >, 0);
+ uint8_t *index_two_records = tuktest_malloc(size_two_records);
+
+ generate_index(index_two_records, unpadded_sizes, uncomp_sizes, 2,
+ size_two_records);
+
+ // First test for basic buffer size error
+ in_pos = size_two_records + 1;
+ assert_lzma_ret(lzma_index_hash_decode(index_hash,
+ index_two_records, &in_pos,
+ size_two_records), LZMA_BUF_ERROR);
+
+ // Next test for invalid Index Indicator
+ in_pos = 0;
+ index_two_records[0] ^= 1;
+ assert_lzma_ret(lzma_index_hash_decode(index_hash,
+ index_two_records, &in_pos,
+ size_two_records), LZMA_DATA_ERROR);
+ index_two_records[0] ^= 1;
+
+ // Next verify the index_hash as expected
+ in_pos = 0;
+ assert_lzma_ret(lzma_index_hash_decode(index_hash,
+ index_two_records, &in_pos,
+ size_two_records), LZMA_STREAM_END);
+
+ // Next test an index_hash with three Records
+ index_hash = lzma_index_hash_init(index_hash, NULL);
+ fill_index_hash(index_hash, unpadded_sizes, uncomp_sizes, 3);
+
+ const lzma_vli size_three_records = lzma_index_hash_size(
+ index_hash);
+ assert_uint(size_three_records, >, 0);
+ uint8_t *index_three_records = tuktest_malloc(size_three_records);
+
+ generate_index(index_three_records, unpadded_sizes, uncomp_sizes,
+ 3, size_three_records);
+
+ in_pos = 0;
+ assert_lzma_ret(lzma_index_hash_decode(index_hash,
+ index_three_records, &in_pos,
+ size_three_records), LZMA_STREAM_END);
+
+ // Next test an index_hash with five Records
+ index_hash = lzma_index_hash_init(index_hash, NULL);
+ fill_index_hash(index_hash, unpadded_sizes, uncomp_sizes, 5);
+
+ const lzma_vli size_five_records = lzma_index_hash_size(
+ index_hash);
+ assert_uint(size_five_records, >, 0);
+ uint8_t *index_five_records = tuktest_malloc(size_five_records);
+
+ generate_index(index_five_records, unpadded_sizes, uncomp_sizes, 5,
+ size_five_records);
+
+ // Instead of testing all input at once, give input
+ // one byte at a time
+ in_pos = 0;
+ for (lzma_vli i = 0; i < size_five_records - 1; ++i) {
+ assert_lzma_ret(lzma_index_hash_decode(index_hash,
+ index_five_records, &in_pos, in_pos + 1),
+ LZMA_OK);
+ }
+
+ // Last byte should return LZMA_STREAM_END
+ assert_lzma_ret(lzma_index_hash_decode(index_hash,
+ index_five_records, &in_pos,
+ in_pos + 1), LZMA_STREAM_END);
+
+ // Next test if the index_hash is given an incorrect Unpadded
+ // Size. Should detect and report LZMA_DATA_ERROR
+ index_hash = lzma_index_hash_init(index_hash, NULL);
+ fill_index_hash(index_hash, unpadded_sizes, uncomp_sizes, 5);
+ // The sixth Record will have an invalid Unpadded Size
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ unpadded_sizes[5] + 1,
+ uncomp_sizes[5]), LZMA_OK);
+
+ const lzma_vli size_six_records = lzma_index_hash_size(
+ index_hash);
+
+ assert_uint(size_six_records, >, 0);
+ uint8_t *index_six_records = tuktest_malloc(size_six_records);
+
+ generate_index(index_six_records, unpadded_sizes, uncomp_sizes, 6,
+ size_six_records);
+ in_pos = 0;
+ assert_lzma_ret(lzma_index_hash_decode(index_hash,
+ index_six_records, &in_pos,
+ size_six_records), LZMA_DATA_ERROR);
+
+ // Next test if the Index is corrupt (invalid CRC32).
+ // Should detect and report LZMA_DATA_ERROR
+ index_hash = lzma_index_hash_init(index_hash, NULL);
+ fill_index_hash(index_hash, unpadded_sizes, uncomp_sizes, 2);
+
+ index_two_records[size_two_records - 1] ^= 1;
+
+ in_pos = 0;
+ assert_lzma_ret(lzma_index_hash_decode(index_hash,
+ index_two_records, &in_pos,
+ size_two_records), LZMA_DATA_ERROR);
+
+ // Next test with Index and index_hash struct not matching
+ // a Record
+ index_hash = lzma_index_hash_init(index_hash, NULL);
+ fill_index_hash(index_hash, unpadded_sizes, uncomp_sizes, 2);
+ // Recalculate Index with invalid Unpadded Size
+ const lzma_vli unpadded_sizes_invalid[2] = {
+ unpadded_sizes[0],
+ unpadded_sizes[1] + 1
+ };
+
+ generate_index(index_two_records, unpadded_sizes_invalid,
+ uncomp_sizes, 2, size_two_records);
+
+ in_pos = 0;
+ assert_lzma_ret(lzma_index_hash_decode(index_hash,
+ index_two_records, &in_pos,
+ size_two_records), LZMA_DATA_ERROR);
+
+ lzma_index_hash_end(index_hash, NULL);
+#endif
+}
+
+
+static void
+test_lzma_index_hash_size(void)
+{
+#ifndef HAVE_DECODERS
+ assert_skip("Decoder support disabled");
+#else
+ lzma_index_hash *index_hash = lzma_index_hash_init(NULL, NULL);
+ assert_true(index_hash != NULL);
+
+ // First test empty index_hash
+ // Expected size should be:
+ // Index Indicator - 1 byte
+ // Number of Records - 1 byte
+ // List of Records - 0 bytes
+ // Index Padding - 2 bytes
+ // CRC32 - 4 bytes
+ // Total - 8 bytes
+ assert_uint_eq(lzma_index_hash_size(index_hash), 8);
+
+ // Append a Record describing a small Block to the index_hash
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ UNPADDED_SIZE_MIN, 1), LZMA_OK);
+
+ // Expected size should be:
+ // Index Indicator - 1 byte
+ // Number of Records - 1 byte
+ // List of Records - 2 bytes
+ // Index Padding - 0 bytes
+ // CRC32 - 4 bytes
+ // Total - 8 bytes
+ lzma_vli expected_size = 8;
+ assert_uint_eq(lzma_index_hash_size(index_hash), expected_size);
+
+ // Append additional small Record
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ UNPADDED_SIZE_MIN, 1), LZMA_OK);
+
+ // Expected size should be:
+ // Index Indicator - 1 byte
+ // Number of Records - 1 byte
+ // List of Records - 4 bytes
+ // Index Padding - 2 bytes
+ // CRC32 - 4 bytes
+ // Total - 12 bytes
+ expected_size = 12;
+ assert_uint_eq(lzma_index_hash_size(index_hash), expected_size);
+
+ // Append a larger Record to the index_hash (3 bytes for each VLI)
+ const lzma_vli three_byte_vli = 0x10000;
+ assert_lzma_ret(lzma_index_hash_append(index_hash,
+ three_byte_vli, three_byte_vli), LZMA_OK);
+
+ // Expected size should be:
+ // Index Indicator - 1 byte
+ // Number of Records - 1 byte
+ // List of Records - 10 bytes
+ // Index Padding - 0 bytes
+ // CRC32 - 4 bytes
+ // Total - 16 bytes
+ expected_size = 16;
+ assert_uint_eq(lzma_index_hash_size(index_hash), expected_size);
+
+ lzma_index_hash_end(index_hash, NULL);
+#endif
+}
+
+
+extern int
+main(int argc, char **argv)
+{
+ tuktest_start(argc, argv);
+ tuktest_run(test_lzma_index_hash_init);
+ tuktest_run(test_lzma_index_hash_append);
+ tuktest_run(test_lzma_index_hash_decode);
+ tuktest_run(test_lzma_index_hash_size);
+ return tuktest_end();
+}