OSS-Fuzz rolling release

This release should only used for integration with OSS-Fuzz.
It's guaranteed to work with Linux/Ubuntu/x86-64/ptrace only.

PS: It's a rolling release.

Version 2.6

• Pastis patch
• env NO_COLOR support
• fix problems with linux/bfd init_disassemble_info
• Support --exit_on_time
• strlcat supported in libhfuzz/memorycmp
• Some patches for MacOSX, sadly no support for ARM yet
• added missing add missing disassemble_free_target()
• using rfork() under FreeBSD
• support for enabling/disabling ASLR under FreeBSD

Version 2.5

• fixed build for Android NDK >= 23
• fixed build for CygWin
• improved hfuzz-cc, so it supports -x correctly
• error returned if unknown cmd-line parameters are provided
• support for thread CPU pinning
• various fixes for *BSD
• increased number of dictionary entries (to 8192)

Version 2.4

• better NetBSD support (compatiblity)
• fixed mangle_getLen() #360
• disabled --linux_net_ns by default, as the Linux kernel dies under heavy load of new net namespaces
• added support for address compression in IntelPT mode
• simplified input_skipFactor() which now yields better results (faster coverage acquisition)
• added --save_smaller for the use with Rust fuzzing
• fixed Android build under MacOSX
• simplified some mangle() functions

Version 2.3.1

• Fixed 'make install' with #356

Version 2.3

• honggfuzz.h - split run_t into substructs
• clang-format options in .clang-format
• added missing mutex initializers
• removed unncessary comparisons to 'true' and 'false'
• improved NetBSD compatibility
• removed unnecessary memory fences (speed ups)
• faster searching through the binary for const 4/8-byte values
• removed unnecessary includes with iwyu
• libhfnetdriver - general improvements around local socket fuzzing and timeouts

Version 2.2

• Added 8bitcnt instrumentation - use hfuzz-cc/hfuzz-8bitcnt-(gcc|clang) for that
• PC-guard instrumentation now uses edge counting
• --experimental_const_feedback is now set to true by default
• additional string instrumentation wrappers: glib, lcms
• additional mutators: splicing, changing ascii numbers
• additional integer comparison instrumentation (adding integers to the dynamic dictionary)
• fixed linking with ld.lld
• removed sanitizer-coverage-prune-blocks from hfuzz-cc.c
• most mutators have now either overwrite or insert versions
• fixed memory barriers in libhfuzz/
• implemented skip_factor which dictates how often a given input is fuzzed
• lowered the default timeout to 1 second
• honggfuzz now uses microseconds, instead of milliseconds across the code
• added some new functions to libhfcommon/files
• enabled more aggressive inlining in hfuzz-cc/
• fixed compilation dependency under MacOS X

Version 2.1

• string/int comparison enabled for targets built with *SAN, but w/o hfuzz-cc
• Parallel work made faster by using faster ATOMIC constructs (check first, then update)
• Implement --experimental_const_feedback - const string/integer feedback (used as an additional dictionary)
• Sanitizer report files are "better"-deleted (i.e. based on PID and not TID)
• New patches for fuzzing added (e.g. for bind-9.16.0/9.15.7)
• Buffered output enabled in display.c
• Some functions moved from per-arch arch.c to common subproc.c
• Compilation under MacOS X 10.15 (Catalina) is now supported
• Added suport for bfd/binutils-2.33

Version 2.0

• Coverage-based corpus minimizer with '-M'
• QEmu mode: coverage feedback for Linux binaries
• *SAN sanitizer stack-parsing improved for Linux and for POSIX
• Move signal functionality to libhfcommon/
• Fixed Android builds with newer unwind and capstone
• NetDriver: more functionality - e.g. specifying custom addresses and custom tmpfs mount points
• Examples: for /usr/bin/file, newer ISC Bind patch, improved OpenSSL code

Version 1.9

• Don't include netdriver if not needed
• Updated examples (bind/openssl)
• Add missing TEMP_FAILURE_RETRY() wrappers
• Add additional _HF_STATE_DYNAMIC_SWITCH_TO_MAIN state