Hi all, We have just pushed Hyperscan 4.2.0 to Github. The changelog on Github is a good summary of what has changed since 4.1, but I will also include it here: [4.2.0] 2016-05-31 • Introduce an interpreter for many complex actions to replace the use of internal reports within the core of Hyperscan (the "Rose" engine). This improves scanning performance and reduces database size for many pattern sets. • Many enhancements to the acceleration framework used by NFA and DFA engines, including more flexible multibyte implementations and more AVX2 support. This improves scanning performance for many pattern sets. • Improved prefiltering support for complex patterns containing very large bounded repeats (R{M,N} with large N). • Improve scanning performance of pattern sets with a very large number of EOD-anchored patterns. • Improve scanning performance of large pattern sets that use the HS_FLAG_SINGLEMATCH flag. • Improve scanning performance of pattern sets that contain a single literal by improving the "Noodle" literal matcher. • Small reductions in total stream state for many pattern sets. • Improve runtime detection of AVX2 support. • Disable -Werror for release builds, in order to behave better for packagers and users with different compiler combinations than those that we test. • Improve support for building on Windows with MSVC 2015 (github issue #14). Support for Hyperscan on Windows is still experimental. • Small updates to fix warnings identified by Coverity. • Remove Python codegen for the "FDR" and "Teddy" literal matchers. These are now implemented directly in C code. • Remove the specialist "Sidecar" engine in favour of using our more general repeat engines. • New API function: add the hs_expression_ext_info() function. This is a variant of hs_expression_info() that can accept patterns with extended parameters. • New API error value: add the HS_SCRATCH_IN_USE error, which is returned when Hyperscan detects that a scratch region is already in use on entry to an API function. regards, Matt.