Update the code to mostly match the new simpler file format

specification. Simplify things by removing most of the
support for known uncompressed size in most places.
There are some miscellaneous changes here and there too.

The API of liblzma has got many changes and still some
more will be done soon. While most of the code has been
updated, some things are not fixed (the command line tool
will choke with invalid filter chain, if nothing else).

Subblock filter is somewhat broken for now. It will be
updated once the encoded format of the Subblock filter
has been decided.

8 files changed, 118 insertions, 59 deletions

diff --git a/src/liblzma/check/check_byteswap.h b/src/common/bswap.h
index 264def26..8f82a8f4 100644
--- a/src/liblzma/check/check_byteswap.h
+++ b/src/common/bswap.h
@@ -1,7 +1,7 @@
 ///////////////////////////////////////////////////////////////////////////////
 //
-/// \file       check_byteswap.h
-/// \brief      Byteswapping needed by the checks
+/// \file       bswap.h
+/// \brief      Byte swapping
 //
 //  This code has been put into the public domain.
 //
@@ -11,18 +11,19 @@
 //
 ///////////////////////////////////////////////////////////////////////////////
 
-#ifndef LZMA_CHECK_BYTESWAP_H
-#define LZMA_CHECK_BYTESWAP_H
+#ifndef LZMA_BSWAP_H
+#define LZMA_BSWAP_H
 
-#ifdef HAVE_CONFIG_H
-#	include <config.h>
-#endif
+// NOTE: We assume that config.h is already #included.
 
 // byteswap.h is a GNU extension. It contains inline assembly versions
 // for byteswapping. When byteswap.h is not available, we use generic code.
 #ifdef HAVE_BYTESWAP_H
 #	include <byteswap.h>
 #else
+#	define bswap_16(num) \
+		(((num) << 8) | ((num) >> 8))
+
 #	define bswap_32(num) \
 		( (((num) << 24)                       ) \
 		| (((num) <<  8) & UINT32_C(0x00FF0000)) \
diff --git a/src/liblzma/check/Makefile.am b/src/liblzma/check/Makefile.am
index e436cb59..182e0868 100644
--- a/src/liblzma/check/Makefile.am
+++ b/src/liblzma/check/Makefile.am
@@ -14,7 +14,6 @@ libcheck_la_SOURCES = \
 	check.c \
 	check.h \
 	check_init.c \
-	check_byteswap.h \
 	crc_macros.h
 libcheck_la_CPPFLAGS = \
 	-I@top_srcdir@/src/liblzma/api \
diff --git a/src/liblzma/check/check.c b/src/liblzma/check/check.c
index ba59af2e..388b57e8 100644
--- a/src/liblzma/check/check.c
+++ b/src/liblzma/check/check.c
@@ -13,8 +13,15 @@
 
 #include "check.h"
 
-// See the .lzma header format specification section 2.2.2.
-LZMA_API const uint32_t lzma_check_sizes[8] = { 0, 4, 4, 8, 16, 32, 32, 64 };
+// See the .lzma header format specification section 2.1.1.2.
+LZMA_API const uint32_t lzma_check_sizes[LZMA_CHECK_ID_MAX + 1] = {
+	0,
+	4, 4, 4,
+	8, 8, 8,
+	16, 16, 16,
+	32, 32, 32,
+	64, 64, 64
+};
 
 
 LZMA_API const lzma_bool lzma_available_checks[LZMA_CHECK_ID_MAX + 1] = {
@@ -27,6 +34,7 @@ LZMA_API const lzma_bool lzma_available_checks[LZMA_CHECK_ID_MAX + 1] = {
 #endif
 
 	false,  // Reserved
+	false,  // Reserved
 
 #ifdef HAVE_CHECK_CRC64
 	true,
@@ -35,6 +43,10 @@ LZMA_API const lzma_bool lzma_available_checks[LZMA_CHECK_ID_MAX + 1] = {
 #endif
 
 	false,  // Reserved
+	false,  // Reserved
+	false,  // Reserved
+	false,  // Reserved
+	false,  // Reserved
 
 #ifdef HAVE_CHECK_SHA256
 	true,
@@ -44,6 +56,9 @@ LZMA_API const lzma_bool lzma_available_checks[LZMA_CHECK_ID_MAX + 1] = {
 
 	false,  // Reserved
 	false,  // Reserved
+	false,  // Reserved
+	false,  // Reserved
+	false,  // Reserved
 };
 
 
@@ -58,24 +73,24 @@ lzma_check_init(lzma_check *check, lzma_check_type type)
 
 #ifdef HAVE_CHECK_CRC32
 	case LZMA_CHECK_CRC32:
-		check->crc32 = 0;
+		check->state.crc32 = 0;
 		break;
 #endif
 
 #ifdef HAVE_CHECK_CRC64
 	case LZMA_CHECK_CRC64:
-		check->crc64 = 0;
+		check->state.crc64 = 0;
 		break;
 #endif
 
 #ifdef HAVE_CHECK_SHA256
 	case LZMA_CHECK_SHA256:
-		lzma_sha256_init(&check->sha256);
+		lzma_sha256_init(check);
 		break;
 #endif
 
 	default:
-		if (type <= LZMA_CHECK_ID_MAX)
+		if ((unsigned)(type) <= LZMA_CHECK_ID_MAX)
 			ret = LZMA_UNSUPPORTED_CHECK;
 		else
 			ret = LZMA_PROG_ERROR;
@@ -93,19 +108,19 @@ lzma_check_update(lzma_check *check, lzma_check_type type,
 	switch (type) {
 #ifdef HAVE_CHECK_CRC32
 	case LZMA_CHECK_CRC32:
-		check->crc32 = lzma_crc32(buf, size, check->crc32);
+		check->state.crc32 = lzma_crc32(buf, size, check->state.crc32);
 		break;
 #endif
 
 #ifdef HAVE_CHECK_CRC64
 	case LZMA_CHECK_CRC64:
-		check->crc64 = lzma_crc64(buf, size, check->crc64);
+		check->state.crc64 = lzma_crc64(buf, size, check->state.crc64);
 		break;
 #endif
 
 #ifdef HAVE_CHECK_SHA256
 	case LZMA_CHECK_SHA256:
-		lzma_sha256_update(buf, size, &check->sha256);
+		lzma_sha256_update(buf, size, check);
 		break;
 #endif
 
@@ -120,11 +135,29 @@ lzma_check_update(lzma_check *check, lzma_check_type type,
 extern void
 lzma_check_finish(lzma_check *check, lzma_check_type type)
 {
+	switch (type) {
+#ifdef HAVE_CHECK_CRC32
+	case LZMA_CHECK_CRC32:
+		*(uint32_t *)(check->buffer) = check->state.crc32;
+		break;
+#endif
+
+#ifdef HAVE_CHECK_CRC64
+	case LZMA_CHECK_CRC64:
+		*(uint64_t *)(check->buffer) = check->state.crc64;
+		break;
+#endif
+
 #ifdef HAVE_CHECK_SHA256
-	if (type == LZMA_CHECK_SHA256)
-		lzma_sha256_finish(&check->sha256);
+	case LZMA_CHECK_SHA256:
+		lzma_sha256_finish(check);
+		break;
 #endif
 
+	default:
+		break;
+	}
+
 	return;
 }
 
diff --git a/src/liblzma/check/check.h b/src/liblzma/check/check.h
index 74279ceb..45ca25e9 100644
--- a/src/liblzma/check/check.h
+++ b/src/liblzma/check/check.h
@@ -17,15 +17,21 @@
 #include "common.h"
 
 
+// Index hashing used to verify the Index with O(1) memory usage needs
+// a good hash function.
+#if defined(HAVE_CHECK_SHA256)
+#	define LZMA_CHECK_BEST LZMA_CHECK_SHA256
+#elif defined(HAVE_CHECK_CRC64)
+#	define LZMA_CHECK_BEST LZMA_CHECK_CRC64
+#else
+#	define LZMA_CHECK_BEST LZMA_CHECK_CRC32
+#endif
+
+
 typedef struct {
 	/// Internal state
 	uint32_t state[8];
 
-	/// Temporary 8-byte aligned buffer to hold incomplete chunk.
-	/// After lzma_check_finish(), the first 32 bytes will contain
-	/// the final digest in big endian byte order.
-	uint8_t buffer[64];
-
 	/// Size of the message excluding padding
 	uint64_t size;
 
@@ -34,10 +40,27 @@ typedef struct {
 
 /// \note       This is not in the public API because this structure will
 ///             change in future.
-typedef union {
-	uint32_t crc32;
-	uint64_t crc64;
-	lzma_sha256 sha256;
+typedef struct {
+	// FIXME Guarantee 8-byte alignment
+
+	/// Buffer to hold the final result; this is also used as a temporary
+	/// buffer in SHA256. Note that this buffer must be 8-byte aligned.
+	uint8_t buffer[64];
+
+	/// Check-specific data
+	union {
+		uint32_t crc32;
+		uint64_t crc64;
+
+		struct {
+			/// Internal state
+			uint32_t state[8];
+
+			/// Size of the message excluding padding
+			uint64_t size;
+		} sha256;
+	} state;
+
 } lzma_check;
 
 
@@ -91,12 +114,12 @@ extern void lzma_crc64_init(void);
 
 // SHA256
 
-extern void lzma_sha256_init(lzma_sha256 *sha256);
+extern void lzma_sha256_init(lzma_check *check);
 
 extern void lzma_sha256_update(
-		const uint8_t *buf, size_t size, lzma_sha256 *sha256);
+		const uint8_t *buf, size_t size, lzma_check *check);
 
-extern void lzma_sha256_finish(lzma_sha256 *sha256);
+extern void lzma_sha256_finish(lzma_check *check);
 
 
 #endif
diff --git a/src/liblzma/check/crc32_init.c b/src/liblzma/check/crc32_init.c
index 0dd402a4..8b596091 100644
--- a/src/liblzma/check/crc32_init.c
+++ b/src/liblzma/check/crc32_init.c
@@ -17,7 +17,7 @@
 #endif
 
 #ifdef WORDS_BIGENDIAN
-#	include "check_byteswap.h"
+#	include "../../common/bswap.h"
 #endif
 
 
diff --git a/src/liblzma/check/crc64_init.c b/src/liblzma/check/crc64_init.c
index 4c91a771..0029987a 100644
--- a/src/liblzma/check/crc64_init.c
+++ b/src/liblzma/check/crc64_init.c
@@ -17,7 +17,7 @@
 #endif
 
 #ifdef WORDS_BIGENDIAN
-#	include "check_byteswap.h"
+#	include "../../common/bswap.h"
 #endif
 
 
diff --git a/src/liblzma/check/crc_macros.h b/src/liblzma/check/crc_macros.h
index 5fbecf07..e827d07d 100644
--- a/src/liblzma/check/crc_macros.h
+++ b/src/liblzma/check/crc_macros.h
@@ -12,7 +12,7 @@
 ///////////////////////////////////////////////////////////////////////////////
 
 #ifdef WORDS_BIGENDIAN
-#	include "check_byteswap.h"
+#	include "../../common/bswap.h"
 
 #	define A(x) ((x) >> 24)
 #	define B(x) (((x) >> 16) & 0xFF)
diff --git a/src/liblzma/check/sha256.c b/src/liblzma/check/sha256.c
index 8e3d375a..ea51896e 100644
--- a/src/liblzma/check/sha256.c
+++ b/src/liblzma/check/sha256.c
@@ -20,7 +20,7 @@
 #include "check.h"
 
 #ifndef WORDS_BIGENDIAN
-#	include "check_byteswap.h"
+#	include "../../common/bswap.h"
 #endif
 
 // At least on x86, GCC is able to optimize this to a rotate instruction.
@@ -104,18 +104,18 @@ transform(uint32_t state[static 8], const uint32_t data[static 16])
 
 
 static void
-process(lzma_sha256 *sha256)
+process(lzma_check *check)
 {
 #ifdef WORDS_BIGENDIAN
-	transform(sha256->state, (uint32_t *)(sha256->buffer));
+	transform(check->state.sha256.state, (uint32_t *)(check->buffer));
 
 #else
 	uint32_t data[16];
 
 	for (size_t i = 0; i < 16; ++i)
-		data[i] = bswap_32(*((uint32_t*)(sha256->buffer) + i));
+		data[i] = bswap_32(*((uint32_t*)(check->buffer) + i));
 
-	transform(sha256->state, data);
+	transform(check->state.sha256.state, data);
 #endif
 
 	return;
@@ -123,41 +123,41 @@ process(lzma_sha256 *sha256)
 
 
 extern void
-lzma_sha256_init(lzma_sha256 *sha256)
+lzma_sha256_init(lzma_check *check)
 {
 	static const uint32_t s[8] = {
 		0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A,
 		0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19,
 	};
 
-	memcpy(sha256->state, s, sizeof(s));
-	sha256->size = 0;
+	memcpy(check->state.sha256.state, s, sizeof(s));
+	check->state.sha256.size = 0;
 
 	return;
 }
 
 
 extern void
-lzma_sha256_update(const uint8_t *buf, size_t size, lzma_sha256 *sha256)
+lzma_sha256_update(const uint8_t *buf, size_t size, lzma_check *check)
 {
 	// Copy the input data into a properly aligned temporary buffer.
 	// This way we can be called with arbitrarily sized buffers
 	// (no need to be multiple of 64 bytes), and the code works also
 	// on architectures that don't allow unaligned memory access.
 	while (size > 0) {
-		const size_t copy_start = sha256->size & 0x3F;
+		const size_t copy_start = check->state.sha256.size & 0x3F;
 		size_t copy_size = 64 - copy_start;
 		if (copy_size > size)
 			copy_size = size;
 
-		memcpy(sha256->buffer + copy_start, buf, copy_size);
+		memcpy(check->buffer + copy_start, buf, copy_size);
 
 		buf += copy_size;
 		size -= copy_size;
-		sha256->size += copy_size;
+		check->state.sha256.size += copy_size;
 
-		if ((sha256->size & 0x3F) == 0)
-			process(sha256);
+		if ((check->state.sha256.size & 0x3F) == 0)
+			process(check);
 	}
 
 	return;
@@ -165,38 +165,41 @@ lzma_sha256_update(const uint8_t *buf, size_t size, lzma_sha256 *sha256)
 
 
 extern void
-lzma_sha256_finish(lzma_sha256 *sha256)
+lzma_sha256_finish(lzma_check *check)
 {
 	// Add padding as described in RFC 3174 (it describes SHA-1 but
 	// the same padding style is used for SHA-256 too).
-	size_t pos = sha256->size & 0x3F;
-	sha256->buffer[pos++] = 0x80;
+	size_t pos = check->state.sha256.size & 0x3F;
+	check->buffer[pos++] = 0x80;
 
 	while (pos != 64 - 8) {
 		if (pos == 64) {
-			process(sha256);
+			process(check);
 			pos = 0;
 		}
 
-		sha256->buffer[pos++] = 0x00;
+		check->buffer[pos++] = 0x00;
 	}
 
 	// Convert the message size from bytes to bits.
-	sha256->size *= 8;
+	check->state.sha256.size *= 8;
 
 #ifdef WORDS_BIGENDIAN
-	*(uint64_t *)(sha256->buffer + 64 - 8) = sha256->size;
+	*(uint64_t *)(check->buffer + 64 - 8) = check->state.sha256.size;
 #else
-	*(uint64_t *)(sha256->buffer + 64 - 8) = bswap_64(sha256->size);
+	*(uint64_t *)(check->buffer + 64 - 8)
+			= bswap_64(check->state.sha256.size);
 #endif
 
-	process(sha256);
+	process(check);
 
 	for (size_t i = 0; i < 8; ++i)
 #ifdef WORDS_BIGENDIAN
-		((uint32_t *)(sha256->buffer))[i] = sha256->state[i];
+		((uint32_t *)(check->buffer))[i]
+				= check->state.sha256.state[i];
 #else
-		((uint32_t *)(sha256->buffer))[i] = bswap_32(sha256->state[i]);
+		((uint32_t *)(check->buffer))[i]
+				= bswap_32(check->state.sha256.state[i]);
 #endif
 
 	return;