* 1.5.12, October 27, 2006 - Cleanups in the initialization routines - Add some x86-64 assembly for multiply-add - Fix problems generating very small (below 384 bit) RSA keys - Support out of tree builds - Bring some of the documentation up to date - More improvements to the Python bindings * 1.5.11, September 10, 2006 - Removed the Algorithm base class - Various cleanups in the public key inheritance hierarchy - Major overhaul of the configure/build setup - Added x86 assembler implementations of Serpent and low-level MPI code - Optimizations for the SHA-1 x86 assembler - Various improvements to the Python wrappers - Work around a Visual Studio compiler bug * 1.5.10, August 13, 2006 - Add x86 assembler versions of MD4, MD5, and SHA-1 - Expand InitializerOptions' language to support on/off switches - Fix definition of OID 2.5.4.8; was accidentally changed in 1.5.9 - Fix possible resource leaks in the mmap allocator - Slightly optimized buffering in MDx_HashFunction - Initialization failures are dealt with somewhat better - Add an example implementing Pollard's Rho algorithm - Better option handling in the test/benchmark tool - Expand the xor_ciph example to support longer keys - Some updates to the documentation * 1.5.9, July 12, 2006 - Fixed bitrot in the AEP engine - Fix support for marking certificate/CRL extensions as critical - Significant cleanups in the library state / initialization code - LibraryInitializer takes an explicit InitializerOptions object - Make Mutex_Factory an abstract class, add Default_Mutex_Factory - Change configuration access to using global_state() - Add support for global named mutexes throughout the library - Add some STL wrappers for the delete operator - Change how certificates are created to be more flexible and general * 1.5.8, June 23, 2006 - Many internal cleanups to the X.509 cert/CRL code - Allow for application code to support new X.509 extensions - Change the return type of X509_Certificate::{subject,issuer}_info - Allow for alternate character set handling mechanisms - Fix a bug that was slowing squaring performance somewhat - Fix a very hard to hit overflow bug in the C version of word3_muladd - Minor cleanups to the assembler modules - Disable es_unix module on FreeBSD due to build problem on FreeBSD 6.1 - Support for GCC 2.95.x has been dropped in this release * 1.5.7, May 28, 2006 - Further, major changes to the BER/DER coding system - Updated the Qt mutex module to use Mutex_Factory - Moved the library global state object into an anonymous namespace - Drop the Visual C++ x86 assembly module due to bugs * 1.5.6, March 1, 2006 - The low-level DER/BER coding system was redesigned and rewritten - Portions of the certificate code were cleaned up internally - Use macros to substantially clean up the GCC assembly code - Added 32-bit x86 assembly for Visual C++ (by Luca Piccarreta) - Avoid a couple of spurious warnings under Visual C++ - Some slight cleanups in X509_PublicKey::key_id * 1.5.5, February 4, 2006 - Fixed a potential infinite loop in the memory pool code (Matt Johnston) - Made Pooling_Allocator::Memory_Block an actual class of sorts - Some small optimizations to the division and modulo computations - Cleaned up the implementation of some of the BigInt operators - Reduced use of dynamic memory allocation in low-level BigInt functions - A few simplifications in the Randpool mixing function - Removed power(), as it was not particularly useful (or fast) - Fixed some annoying bugs in the benchmark code - Added a real credits file * 1.5.4, January 29, 2006 - Integrated x86 and amd64 assembly code, contributed by Luca Piccarreta - Fixed a memory access off-by-one in the Karatsuba code - Changed Pooling_Allocator's free list search to a log(N) algorithm - Merged ModularReducer with its only subclass, Barrett_Reducer - Fixed sign-handling bugs in some of the division and modulo code - Renamed the module description files to modinfo.txt - Further cleanups in the initialization code - Removed BigInt::add and BigInt::sub - Merged all the division-related functions into just divide() - Modified the functions to allow for better optimizations - Made the number of bits polled from an EntropySource user configurable - Avoid including in - Fixed some build problems with Sun Forte - Removed some dead code from bigint_modop - Fix the definition of same_mem * 1.5.3, January 24, 2006 - Many optimizations in the low-level multiple precision integer code - Added hooks for assembly implementations of the MPI code - Support for the X.509 issuer alternative name extension in new certs - Fixed a bug in the decompression modules; found and patched by Matt Johnston - New Windows mutex module (mux_win32), by Luca Piccarreta - Changed the Windows timer module to use QueryPerformanceCounter - mem_pool.cpp was using std::set iterators instead of std::multiset ones - Fixed a bug in X509_CA preventing users from disabling particular extensions - Fixed the mp_asm64 module, which was entirely broken in 1.5.2 - Fixed some module build problems on FreeBSD and Tru64 * 1.5.2, January 15, 2006 - Fixed an off-by-one memory read in MISTY1::key() - Fixed a nasty memory leak in Output_Buffers::retire() - Reimplemented the memory allocator for scratch - Improved memory caching in Montgomery exponentiation - Optimizations for multiple precision addition and subtraction - Fixed a build problem in the hardware timer module on 64-bit PowerPC - Changed default Karatsuba cutoff to 12 words (was 14) - Removed MemoryRegion::bits(), which was unused and incorrect - Changed maximum HMAC keylength to 1024 bits - Various minor Makefile and build system changes - Avoid using std::min in to bypass Windows libc macro pollution - Switched checks/clock.cpp back to using clock() by default - Enabled the symmetric algorithm tests, which were accidentally off in 1.5.1 - Removed the Default_Mutex's unused clone() member function * 1.5.1, January 8, 2006 - Implemented Montgomery exponentiation - Implemented generalized Karatsuba multiplication and squaring - Implemented Comba squaring for 4, 6, and 8 word inputs - Added new Modular_Exponentiator and Power_Mod classes - Removed FixedBase_Exp and FixedExponent_Exp - Fixed a performance regression in get_allocator introduced in 1.5.0 - Engines can now offer S2K algorithms and block cipher padding methods - Merged the remaining global 'algolist' code into Default_Engine - The low-level MPI code is linked as C again - Replaced BigInt's get_nibble with the more general get_substring - Some documentation updates * 1.5.0, January 1, 2006 - Moved all global/shared library state into a single object - Mutex objects are created through mutex factories instead of a global - Removed ::get_mutex(), ::initialize_mutex(), and Mutex::clone() - Removed the RNG_Quality enum entirely - There is now only a single global-use PRNG - Removed the no_aliases and no_oids options for LibraryInitializer - Removed the deprecated algorithms SEAL, ISAAC, and HAVAL - Change es_ftw to use unbuffered I/O