xref: /aosp_15_r20/external/AFLplusplus/docs/Changelog.md (revision 08b48e0b10e97b33e7b60c5b6e2243bd915777f2)
1# Changelog
2
3  This is the list of all noteworthy changes made in every public
4  release of the tool. See README.md for the general instruction manual.
5
6### Version ++4.10c (release)
7  - afl-fuzz:
8    - default power schedule is now EXPLORE, due a fix in fast schedules
9      explore is slightly better now.
10    - fixed minor issues in the mutation engine, thanks to @futhewo for
11      reporting!
12    - better deterministic fuzzing is now available, benchmarks have shown
13      to improve fuzzing. Enable with -D. Thanks to @kdsjZh for the PR!
14  - afl-cc:
15    - large rewrite by @SonicStark which fixes a few corner cases, thanks!
16    - LTO mode now requires llvm 12+
17    - workaround for ASAN with gcc_plugin mode
18  - instrumentation:
19    - LLVM 18 support, thanks to @devnexen!
20    - Injection (SQL, LDAP, XSS) fuzzing feature now available, see
21      `instrumentation/README.injections.md` how to activate/use/expand.
22    - compcov/LAF-intel:
23      - floating point splitting bug fix by @hexcoder
24      - due a bug in LLVM 17 integer splitting is disabled there!
25      - when splitting floats was selected, integers were always split as well,
26        fixed to require AFL_LLVM_LAF_SPLIT_COMPARES or _ALL as it should
27    - dynamic instrumentation filtering for LLVM NATIVE, thanks @Mozilla!
28      see utils/dynamic_covfilter/README.md
29  - qemu_mode:
30    - plugins are now activated by default and a new module is included that
31      produces drcov compatible traces for lighthouse/lightkeeper/...
32      thanks to @JRomainG to submitting!
33  - updated Nyx checkout (fixes a bug) and some QOL
34  - updated the custom grammar mutator
35  - document afl-cmin does not work on macOS (but afl-cmin.bash does)
36
37### Version ++4.09c (release)
38  - afl-fuzz:
39    - fixed the new mutation implementation for two bugs
40    - added `AFL_FINAL_SYNC` which forces a final fuzzer sync (also for `-F`)
41      before terminating.
42    - added AFL_IGNORE_SEED_PROBLEMS to skip over seeds that time out instead
43      of exiting with an error message
44    - allow -S/-M naming up to 50 characters (from 24)
45    - CMPLOG:
46      - added scale support (-l S)
47      - skip unhelpful insertions (u8)
48    - added --version and --help command line parameters
49    - fixed endless loop when reading malformed dictionaries
50    - new custom mutator function: post_run - thanks to yangzao!
51  - afl-whatsup:
52    - detect instanced that are starting up and show them as such as not dead
53    - now also shows coverage reached
54    - option -m shows only very relevant stats
55    - option -n will not use color in the output
56  - instrumentation:
57    - fix for a few string compare transform functions for LAF
58    - we are instrumenting __cxx internal functions again. this might break
59      a few targets, please report if so.
60  - frida_mode:
61    - fixes support for large map offsets
62  - support for AFL_FUZZER_LOOPCOUNT for afl.rs and LLVMFuzzerTestOneInput
63  - afl-cmin/afl-cmin.bash: prevent unneeded file errors
64  - added new tool afl-addseeds that adds new seeds to a running campaign
65  - added benchmark/benchmark.py if you want to see how good your fuzzing
66    speed is in comparison to other setups.
67
68### Version ++4.08c (release)
69  - afl-fuzz:
70    - new mutation engine: mutations that favor discovery more paths are
71      prefered until no new finds for 10 minutes then switching to mutations
72      that favor triggering crashes. Modes and switch time can be configured
73      with `-P`. Also input mode for the target can be defined with `-a` to
74      be `text` or `binary` (defaults to `generic`)
75    - new custom mutator that has the new afl++ engine (so it can easily
76      incorporated into new custom mutators), and also comes with a standalone
77      command line tool! See custom_mutators/aflpp/standalone/
78    - display the state of the fuzzing run in the UI :-)
79    - fix timeout setting if '+' is used or a session is restarted
80    - -l X option to enable base64 transformation solving
81    - allow to disable CMPLOG with '-c -' (e.g. afl.rs enforces '-c 0' on
82      every instance which is counterproductive).
83  - afl-cmin/afl-cmin.bash:
84    - fixed a bug inherited from vanilla AFL where a coverage of
85      map[123] = 11 would be the same as map[1123] = 1
86    - warn on crashing inputs
87    - adjust threads if less inputs than threads specified
88  - afl-cc:
89    - fixed an off-by-one instrumentation of iselect, hurting coverage a bit.
90      Thanks to @amykweon for spotting and fixing!
91    - @toka fixed a bug in laf-intel signed integer comparison splitting,
92      thanks a lot!!
93    - more LLVM compatability
94  - frida_mode:
95    - support for long form instrumentation on x86_x64 and arm64
96    - renamed utils/get_symbol_addr.sh to utils/frida_get_symbol_addr.sh
97  - qemu_mode:
98    - added qemu_mode/utils/qemu_get_symbol_addr.sh
99
100### Version ++4.07c (release)
101  - afl-fuzz:
102    - reverse reading the seeds only on restarts (increases performance)
103    - new env `AFL_POST_PROCESS_KEEP_ORIGINAL` to keep the orignal
104      data before post process on finds (for atnwalk custom mutator)
105    - new env `AFL_IGNORE_PROBLEMS_COVERAGE` to ignore coverage from
106      loaded libs after forkserver initialization (required by Mozilla)
107  - afl-cc:
108    - added @responsefile support
109    - new env `AFL_LLVM_LTO_SKIPINIT` to support the AFL++ based WASM
110      (https://github.com/fgsect/WAFL) project
111    - error and print help if afl-clan-lto is used with lto=thin
112    - rewrote our PCGUARD pass to be compatible with LLVM 15+ shenanigans,
113      requires LLVM 13+ now instead of 10.0.1+
114    - fallback to native LLVM PCGUARD if our PCGUARD is unavailable
115    - fixed a crash in GCC CMPLOG
116  - afl-showmap:
117    - added custom mutator post_process and send support
118    - add `-I filelist` option, an alternative to `-i in_dir`
119  - afl-cmin + afl-cmin.bash:
120    - `-T threads` parallel task support, can be a huge speedup!
121  - qemu_mode:
122    - Persistent mode + QASAN support for ppc32 targets by @worksbutnottested
123  - a new grammar custom mutator atnwalk was submitted by @voidptr127 !
124  - two new custom mutators are now available:
125    - TritonDSE in custom_mutators/aflpp_tritondse
126    - SymQEMU in custom_mutators/symqemu
127
128### Version ++4.06c (release)
129  - afl-fuzz:
130    - ensure temporary file descriptor is closed when not used
131    - added `AFL_NO_WARN_INSTABILITY`
132    - added time_wo_finds to fuzzer_stats
133    - fixed a crash in pizza (1st april easter egg) mode. Sorry for
134      everyone who was affected!
135    - allow pizza mode to be disabled when AFL_PIZZA_MODE is set to -1
136    - option `-p mmopt` now also selects new queue items more often
137    - fix bug in post_process custom mutator implementation
138    - print name of custom mutator in UI
139    - slight changes that improve fuzzer performance
140  - afl-cc:
141    - add CFI sanitizer variant to gcc targets
142    - llvm 16 + 17 support (thanks to @devnexen!)
143    - support llvm 15 native pcguard changes
144    - support for LLVMFuzzerTestOneInput -1 return
145    - LTO autoken and llvm_mode: added AFL_LLVM_DICT2FILE_NO_MAIN support
146  - qemu_mode:
147    - fix _RANGES envs to allow hyphens in the filenames
148    - basic riscv support
149  - frida_mode:
150    - added `AFL_FRIDA_STATS_INTERVAL`
151    - fix issue on MacOS
152  - unicorn_mode:
153    - updated and minor issues fixed
154  - nyx_mode support for all tools
155  - better sanitizer default options support for all tools
156  - new custom module: autotoken, a grammar free fuzzer for text inputs
157  - fixed custom mutator C examples
158  - more minor fixes and cross-platform support
159
160### Version ++4.05c (release)
161  - MacOS: libdislocator, libtokencap etc. do not work with modern
162    MacOS anymore, but could be patched to work, see this issue if you
163    want to make the effort and send a PR:
164    https://github.com/AFLplusplus/AFLplusplus/issues/1594
165  - afl-fuzz:
166    - added afl_custom_fuzz_send custom mutator feature. Now your can
167      send fuzz data to the target as you need, e.g. via IPC.
168    - cmplog mode now has a -l R option for random colorization, thanks
169      to guyf2010 for the PR!
170    - queue statistics are written every 30 minutes to
171      out/NAME/queue_data if compiled with INTROSPECTION
172    - new env: AFL_FORK_SERVER_KILL_SIGNAL
173  - afl-showmap/afl-cmin
174    - `-t none` now translates to `-t 120000` (120 seconds)
175  - unicorn_mode updated
176  - updated rust custom mutator dependencies and LibAFL custom mutator
177  - overall better sanitizer default setting handling
178  - several minor bugfixes
179
180### Version ++4.04c (release)
181  - fix gramatron and grammar_mutator build scripts
182  - enhancements to the afl-persistent-config and afl-system-config
183    scripts
184  - afl-fuzz:
185    - force writing all stats on exit
186    - ensure targets are killed on exit
187    - `AFL_FORK_SERVER_KILL_SIGNAL` added
188  - afl-cc:
189    - make gcc_mode (afl-gcc-fast) work with gcc down to version 3.6
190  - qemu_mode:
191    - fixed 10x speed degredation in v4.03c, thanks to @ele7enxxh for
192      reporting!
193    - added qemu_mode/fastexit helper library
194  - unicorn_mode:
195    - Enabled tricore arch (by @jma-qb)
196    - Updated Capstone version in Rust bindings
197  - llvm-mode:
198    - AFL runtime will always pass inputs via shared memory, when possible,
199      ignoring the command line.
200
201
202### Version ++4.03c (release)
203  - Building now gives a build summary what succeeded and what not
204  - afl-fuzz:
205    - added AFL_NO_STARTUP_CALIBRATION to start fuzzing at once instead
206      of calibrating all initial seeds first. Good for large queues
207      and long execution times, especially in CIs.
208    - default calibration cycles set to 7 from 8, and only add 5 cycles
209      to variables queue items instead of 12.
210  - afl-cc:
211    - fixed off-by-one bug in our pcguard implemenation, thanks for
212      @tokatoka for reporting
213    - fix for llvm 15 and reenabling LTO, thanks to nikic for the PR!
214    - better handling of -fsanitize=..,...,.. lists
215    - support added for LLVMFuzzerRunDriver()
216    - fix gcc_mode cmplog
217    - obtain the map size of a target with setting AFL_DUMP_MAP_SIZE=1
218      note that this will exit the target before main()
219  - qemu_mode:
220    - added AFL_QEMU_TRACK_UNSTABLE to log the addresses of unstable
221      edges (together with AFL_DEBUG=1 afl-fuzz). thanks to
222      worksbutnottested!
223  - afl-analyze broke at some point, fix by CodeLogicError, thank you!
224  - afl-cmin/afl-cmin.bash now have an -A option to allow also crashing
225    and timeout inputs
226  - unicorn_mode:
227    - updated upstream unicorn version
228    - fixed builds for aarch64
229    - build now uses all available cores
230
231
232### Version ++4.02c (release)
233  - afl-cc:
234    - important fix for the default pcguard mode when LLVM IR vector
235      selects are produced, thanks to @juppytt for reporting!
236  - gcc_plugin:
237    - Adacore submitted CMPLOG support to the gcc_plugin! :-)
238  - llvm_mode:
239    - laf cmp splitting fixed for more comparison types
240  - frida_mode:
241    - now works on Android!
242  - afl-fuzz:
243    - change post_process hook to allow returning NULL and 0 length to
244      tell afl-fuzz to skip this mutated input
245
246### Version ++4.01c (release)
247  - fixed */build_...sh scripts to work outside of git
248  - new custom_mutator: libafl with token fuzzing :)
249  - afl-fuzz:
250    - when you just want to compile once and set CMPLOG, then just
251      set -c 0 to tell afl-fuzz that the fuzzing binary is also for
252      CMPLOG.
253    - new commandline options -g/G to set min/max length of generated
254      fuzz inputs
255    - you can set the time for syncing to other fuzzer now with
256      AFL_SYNC_TIME
257    - reintroduced AFL_PERSISTENT and AFL_DEFER_FORKSRV to allow
258      persistent mode and manual forkserver support if these are not
259      in the target binary (e.g. are in a shared library)
260    - add AFL_EARLY_FORKSERVER to install the forkserver as earliest as
261      possible in the target (for afl-gcc-fast/afl-clang-fast/
262      afl-clang-lto)
263    - "saved timeouts" was wrong information, timeouts are still thrown
264      away by default even if they have new coverage (hangs are always
265      kept), unless AFL_KEEP_TIMEOUTS are set
266    - AFL never implemented auto token inserts (but user token inserts,
267      user token overwrite and auto token overwrite), added now!
268    - fixed a mutation type in havoc mode
269    - Mopt fix to always select the correct algorithm
270    - fix effector map calculation (deterministic mode)
271    - fix custom mutator post_process functionality
272    - document and auto-activate pizza mode on condition
273  - afl-cc:
274    - due a bug in lld of llvm 15 LTO instrumentation wont work atm :-(
275    - converted all passed to use the new llvm pass manager for llvm 11+
276    - AFL++ PCGUARD mode is not available for 10.0.1 anymore (11+ only)
277    - trying to stay on top on all these #$&§!! changes in llvm 15 ...
278  - frida_mode:
279    - update to new frida release, handles now c++ throw/catch
280  - unicorn_mode:
281    - update unicorn engine, fix C example
282  - utils:
283    - removed optimin because it looses coverage due to a bug and is
284      unmaintained :-(
285
286
287### Version ++4.00c (release)
288  - complete documentation restructuring, made possible by Google Season
289    of Docs :) thank you Jana!
290  - we renamed several UI and fuzzer_stat entries to be more precise,
291    e.g. "unique crashes" -> "saved crashes", "total paths" ->
292    "corpus count", "current path" -> "current item".
293    This might need changing custom scripting!
294  - Nyx mode (full system emulation with snapshot capability) has been
295    added - thanks to @schumilo and @eqv!
296  - unicorn_mode:
297    - Moved to unicorn2! by Ziqiao Kong (@lazymio)
298    - Faster, more accurate emulation (newer QEMU base), risc-v support
299    - removed indirections in rust callbacks
300  - new binary-only fuzzing mode: coresight_mode for aarch64 CPUs :)
301    thanks to RICSecLab submitting!
302  - if instrumented libaries are dlopen()'ed after the forkserver you
303    will now see a crash. Before you would have colliding coverage.
304    We changed this to force fixing a broken setup rather then allowing
305    ineffective fuzzing.
306    See docs/best_practices.md how to fix such setups.
307  - afl-fuzz:
308    - cmplog binaries will need to be recompiled for this version
309      (it is better!)
310    - fix a regression introduced in 3.10 that resulted in less
311      coverage being detected. thanks to Collin May for reporting!
312    - ensure all spawned targets are killed on exit
313    - added AFL_IGNORE_PROBLEMS, plus checks to identify and abort on
314      incorrect LTO usage setups and enhanced the READMEs for better
315      information on how to deal with instrumenting libraries
316    - fix -n dumb mode (nobody should use this mode though)
317    - fix stability issue with LTO and cmplog
318    - better banner
319    - more effective cmplog mode
320    - more often update the UI when in input2stage mode
321  - qemu_mode/unicorn_mode: fixed OOB write when using libcompcov,
322      thanks to kotee4ko for reporting!
323  - frida_mode:
324    - better performance, bug fixes
325    - David Carlier added Android support :)
326  - afl-showmap, afl-tmin and afl-analyze:
327    - honor persistent mode for more speed. thanks to dloffre-snl
328      for reporting!
329    - fix bug where targets are not killed on timeouts
330    - moved hidden afl-showmap -A option to -H to be used for
331      coresight_mode
332  - Prevent accidentally killing non-afl/fuzz services when aborting
333    afl-showmap and other tools.
334  - afl-cc:
335    - detect overflow reads on initial input buffer for asan
336    - new cmplog mode (incompatible with older AFL++ versions)
337    - support llvm IR select instrumentation for default PCGUARD and LTO
338    - fix for shared linking on MacOS
339    - better selective instrumentation AFL_LLVM_{ALLOW|DENY}LIST
340      on filename matching (requires llvm 11 or newer)
341    - fixed a potential crash in targets for LAF string handling
342    - fixed a bad assert in LAF split switches
343    - added AFL_USE_TSAN thread sanitizer support
344    - llvm and LTO mode modified to work with new llvm 14-dev (again.)
345    - fix for AFL_REAL_LD
346    - more -z defs filtering
347    - make -v without options work
348  - added the very good grammar mutator "GramaTron" to the
349    custom_mutators
350  - added optimin, a faster and better corpus minimizer by
351    Adrian Herrera. Thank you!
352  - added afl-persistent-config script to set perform permanent system
353    configuration settings for fuzzing, for Linux and Macos.
354    thanks to jhertz!
355  - added xml, curl & exotic string functions to llvm dictionary feature
356  - fix AFL_PRELOAD issues on MacOS
357  - removed utils/afl_frida because frida_mode/ is now so much better
358  - added uninstall target to makefile (todo: update new readme!)
359
360### Version ++3.14c (release)
361  - afl-fuzz:
362    - fix -F when a '/' was part of the parameter
363    - fixed a crash for cmplog for very slow inputs
364    - fix for AFLfast schedule counting
365    - removed implied -D determinstic from -M main
366    - if the target becomes unavailable check out out/default/error.txt
367      for an indicator why
368    - AFL_CAL_FAST was a dead env, now does the same as AFL_FAST_CAL
369    - reverse read the queue on resumes (more effective)
370    - fix custom mutator trimming
371  - afl-cc:
372    - Update to COMPCOV/laf-intel that speeds up the instrumentation
373      process a lot - thanks to Michael Rodler/f0rki for the PR!
374    - Fix for failures for some sized string instrumentations
375    - Fix to instrument global namespace functions in c++
376    - Fix for llvm 13
377    - support partial linking
378    - do honor AFL_LLVM_{ALLOW/DENY}LIST for LTO autodictionary andDICT2FILE
379    - We do support llvm versions from 3.8 to 5.0 again
380  - frida_mode:
381    - several fixes for cmplog
382    - remove need for AFL_FRIDA_PERSISTENT_RETADDR_OFFSET
383    - less coverage collision
384    - feature parity of aarch64 with intel now (persistent, cmplog,
385      in-memory testcases, asan)
386  - afl-cmin and afl-showmap -i do now descend into subdirectories
387    (like afl-fuzz does) - note that afl-cmin.bash does not!
388  - afl_analyze:
389    - fix timeout handling
390    - add forkserver support for better performance
391  - ensure afl-compiler-rt is built for gcc_module
392  - always build aflpp_driver for libfuzzer harnesses
393  - added `AFL_NO_FORKSRV` env variable support to
394    afl-cmin, afl-tmin, and afl-showmap, by @jhertz
395  - removed outdated documents, improved existing documentation
396
397### Version ++3.13c (release)
398  - Note: plot_data switched to relative time from unix time in 3.10
399  - frida_mode - new mode that uses frida to fuzz binary-only targets,
400    it currently supports persistent mode and cmplog.
401    thanks to @WorksButNotTested!
402  - create a fuzzing dictionary with the help of CodeQL thanks to
403    @microsvuln! see utils/autodict_ql
404  - afl-fuzz:
405    - added patch by @realmadsci to support @@ as part of command line
406      options, e.g. `afl-fuzz ... -- ./target --infile=@@`
407    - add recording of previous fuzz attempts for persistent mode
408      to allow replay of non-reproducable crashes, see
409      AFL_PERSISTENT_RECORD in config.h and docs/envs.h
410    - fixed a bug when trimming for stdin targets
411    - cmplog -l: default cmplog level is now 2, better efficiency.
412      level 3 now performs redqueen on everything. use with care.
413    - better fuzzing strategy yield display for enabled options
414    - ensure one fuzzer sync per cycle
415    - fix afl_custom_queue_new_entry original file name when syncing
416      from fuzzers
417    - fixed a crash when more than one custom mutator was used together
418      with afl_custom_post_process
419    - on a crashing seed potentially the wrong input was disabled
420    - added AFL_EXIT_ON_SEED_ISSUES env that will exit if a seed in
421      -i dir crashes the target or results in a timeout. By default
422      AFL++ ignores these and uses them for splicing instead.
423    - added AFL_EXIT_ON_TIME env that will make afl-fuzz exit fuzzing
424      after no new paths have been found for n seconds
425    - when AFL_FAST_CAL is set a variable path will now be calibrated
426      8 times instead of originally 40. Long calibration is now 20.
427    - added AFL_TRY_AFFINITY to try to bind to CPUs but don't error if
428      it fails
429  - afl-cc:
430    - We do not support llvm versions prior 6.0 anymore
431    - added thread safe counters to all modes (`AFL_LLVM_THREADSAFE_INST`),
432      note that this disables NeverZero counters.
433    - Fix for -pie compiled binaries with default afl-clang-fast PCGUARD
434    - Leak Sanitizer (AFL_USE_LSAN) added by Joshua Rogers, thanks!
435    - Removed InsTrim instrumentation as it is not as good as PCGUARD
436    - Removed automatic linking with -lc++ for LTO mode
437    - Fixed a crash in llvm dict2file when a strncmp length was -1
438    - added --afl-noopt support
439  - utils/aflpp_driver:
440    - aflpp_qemu_driver_hook fixed to work with qemu_mode
441    - aflpp_driver now compiled with -fPIC
442  - unicornafl:
443    - fix MIPS delay slot caching, thanks @JackGrence
444    - fixed aarch64 exit address
445    - execution no longer stops at address 0x0
446  - updated afl-system-config to support Arch Linux weirdness and increase
447    MacOS shared memory
448  - updated the grammar custom mutator to the newest version
449  - add -d (add dead fuzzer stats) to afl-whatsup
450  - added AFL_PRINT_FILENAMES to afl-showmap/cmin to print the
451    current filename
452  - afl-showmap/cmin will now process queue items in alphabetical order
453
454### Version ++3.12c (release)
455  - afl-fuzz:
456    - added AFL_TARGET_ENV variable to pass extra env vars to the target
457      (for things like LD_LIBRARY_PATH)
458    - fix map detection, AFL_MAP_SIZE not needed anymore for most cases
459    - fix counting favorites (just a display thing)
460  - afl-cc:
461    - fix cmplog rtn (rare crash and not being able to gather ptr data)
462    - fix our own PCGUARD implementation to compile with llvm 10.0.1
463    - link runtime not to shared libs
464    - ensure shared libraries are properly built and instrumented
465    - AFL_LLVM_INSTRUMENT_ALLOW/DENY were not implemented for LTO, added
466    - show correct LLVM PCGUARD NATIVE mode when auto switching to it
467      and keep fsanitize-coverage-*list=...
468      Short mnemnonic NATIVE is now also accepted.
469  - qemu_mode (thanks @realmadsci):
470    - move AFL_PRELOAD and AFL_USE_QASAN logic inside afl-qemu-trace
471    - add AFL_QEMU_CUSTOM_BIN
472  - unicorn_mode
473    - accidently removed the subfolder from github, re-added
474  - added DEFAULT_PERMISSION to config.h for all files created, default
475    to 0600
476
477### Version ++3.11c (release)
478  - afl-fuzz:
479    - better auto detection of map size
480    - fix sanitizer settings (bug since 3.10c)
481    - fix an off-by-one overwrite in cmplog
482    - add non-unicode variants from unicode-looking dictionary entries
483    - Rust custom mutator API improvements
484    - Imported crash stats painted yellow on resume (only new ones are red)
485  - afl-cc:
486    - added AFL_NOOPT that will just pass everything to the normal
487      gcc/clang compiler without any changes - to pass weird configure
488      scripts
489    - fixed a crash that can occur with ASAN + CMPLOG together plus
490      better support for unicode (thanks to @stbergmann for reporting!)
491    - fixed a crash in LAF transform for empty strings
492    - handle erroneous setups in which multiple afl-compiler-rt are
493      compiled into the target. This now also supports dlopen()
494      instrumented libs loaded before the forkserver and even after the
495      forkserver is started (then with collisions though)
496    - the compiler rt was added also in object building (-c) which
497      should have been fixed years ago but somewhere got lost :(
498    - Renamed CTX to CALLER, added correct/real CTX implementation to
499      CLASSIC
500  - qemu_mode:
501    - added AFL_QEMU_EXCLUDE_RANGES env by @realmadsci, thanks!
502    - if no new/updated checkout is wanted, build with:
503      NO_CHECKOUT=1 ./build_qemu_support.sh
504    - we no longer perform a "git drop"
505  - afl-cmin: support filenames with spaces
506
507### Version ++3.10c (release)
508  - Mac OS ARM64 support
509  - Android support fixed and updated by Joey Jiaojg - thanks!
510  - New selective instrumentation option with __AFL_COVERAGE_* commands
511    to be placed in the source code.
512    Check out instrumentation/README.instrument_list.md
513  - afl-fuzz
514    - Making AFL_MAP_SIZE (mostly) obsolete - afl-fuzz now learns on
515      start the target map size
516    - upgraded cmplog/redqueen: solving for floating point, solving
517      transformations (e.g. toupper, tolower, to/from hex, xor,
518      arithmetics, etc.). This is costly hence new command line option
519      `-l` that sets the intensity (values 1 to 3). Recommended is 2.
520    - added `AFL_CMPLOG_ONLY_NEW` to not use cmplog on initial seeds
521      from `-i` or resumes (these have most likely already been done)
522    - fix crash for very, very fast targets+systems (thanks to mhlakhani
523      for reporting)
524    - on restarts (`-i`)/autoresume (AFL_AUTORESUME) the stats are now
525      reloaded and used, thanks to Vimal Joseph for this patch!
526    - changed the meaning of '+' of the '-t' option, it now means to
527      auto-calculate the timeout with the value given being the max
528      timeout. The original meaning of skipping timeouts instead of
529      abort is now inherent to the -t option.
530    - if deterministic mode is active (`-D`, or `-M` without `-d`) then
531      we sync after every queue entry as this can take very long time
532      otherwise
533    - added minimum SYNC_TIME to include/config.h (30 minutes default)
534    - better detection if a target needs a large shared map
535    - fix for `-Z`
536    - fixed a few crashes
537    - switched to an even faster RNG
538    - added hghwng's patch for faster trace map analysis
539    - printing suggestions for mistyped `AFL_` env variables
540    - added Rust bindings for custom mutators (thanks @julihoh)
541  - afl-cc
542    - allow instrumenting LLVMFuzzerTestOneInput
543    - fixed endless loop for allow/blocklist lines starting with a
544      comment (thanks to Zherya for reporting)
545    - cmplog/redqueen now also tracks floating point, _ExtInt() + 128bit
546    - cmplog/redqueen can now process basic libc++ and libstdc++
547      std::string comparisons (no position or length type variants)
548    - added support for __afl_coverage_interesting() for LTO and our
549      own PCGUARD (llvm 10.0.1+), read more about this function and
550      selective coverage in instrumentation/README.instrument_list.md
551    - added AFL_LLVM_INSTRUMENT option NATIVE for native clang pc-guard
552      support (less performant than our own), GCC for old afl-gcc and
553      CLANG for old afl-clang
554    - fixed a potential crash in the LAF feature
555    - workaround for llvm bitcast lto bug
556    - workaround for llvm 13
557  - qemuafl
558    - QASan (address sanitizer for Qemu) ported to qemuafl!
559      See qemu_mode/libqasan/README.md
560    - solved some persistent mode bugs (thanks Dil4rd)
561    - solved an issue when dumping the memory maps (thanks wizche)
562    - Android support for QASan
563  - unicornafl
564    - Substantial speed gains in python bindings for certain use cases
565    - Improved rust bindings
566    - Added a new example harness to compare python, c and rust bindings
567  - afl-cmin and afl-showmap now support the -f option
568  - afl_plot now also generates a graph on the discovered edges
569  - changed default: no memory limit for afl-cmin and afl-cmin.bash
570  - warn on any _AFL and __AFL env vars.
571  - set AFL_IGNORE_UNKNOWN_ENVS to not warn on unknown AFL_... env vars
572  - added dummy Makefile to instrumentation/
573  - Updated utils/afl_frida to be 5% faster, 7% on x86_x64
574  - Added `AFL_KILL_SIGNAL` env variable (thanks @v-p-b)
575  - @Edznux added a nice documentation on how to use rpc.statsd with
576    AFL++ in docs/rpc_statsd.md, thanks!
577
578### Version ++3.00c (release)
579  - llvm_mode/ and gcc_plugin/ moved to instrumentation/
580  - examples/ renamed to utils/
581  - moved libdislocator, libtokencap and qdbi_mode to utils/
582  - all compilers combined to afl-cc which emulates the previous ones
583  - afl-llvm/gcc-rt.o merged into afl-compiler-rt.o
584  - afl-fuzz
585    - not specifying -M or -S will now auto-set "-S default"
586    - deterministic fuzzing is now disabled by default and can be enabled with
587      -D. It is still enabled by default for -M.
588    - a new seed selection was implemented that uses weighted randoms based on
589      a schedule performance score, which is much better that the previous
590      walk the whole queue approach. Select the old mode with -Z (auto enabled
591      with -M)
592    - Marcel Boehme submitted a patch that improves all AFFast schedules :)
593    - the default schedule is now FAST
594    - memory limits are now disabled by default, set them with -m if required
595    - rpc.statsd support, for stats and charts, by Edznux, thanks a lot!
596    - reading testcases from -i now descends into subdirectories
597    - allow the -x command line option up to 4 times
598    - loaded extras now have a duplication protection
599    - If test cases are too large we do a partial read on the maximum
600      supported size
601    - longer seeds with the same trace information will now be ignored
602      for fuzzing but still be used for splicing
603    - crashing seeds are now not prohibiting a run anymore but are
604      skipped - they are used for splicing, though
605    - update MOpt for expanded havoc modes
606    - setting the env var AFL_NO_AUTODICT will not load an LTO autodictionary
607    - added NO_SPLICING compile option and makefile define
608    - added INTROSPECTION make target that writes all mutations to
609      out/NAME/introspection.txt
610    - print special compile time options used in help output
611    - when using -c cmplog, one of the childs was not killed, fixed
612    - somewhere we broke -n dumb fuzzing, fixed
613    - added afl_custom_describe to the custom mutator API to allow for easy
614      mutation reproduction on crashing inputs
615    - new env. var. AFL_NO_COLOR (or AFL_NO_COLOUR) to suppress colored
616      console output (when configured with USE_COLOR and not ALWAYS_COLORED)
617  - instrumentation
618    - We received an enhanced gcc_plugin module from AdaCore, thank you
619      very much!!
620    - not overriding -Ox or -fno-unroll-loops anymore
621    - we now have our own trace-pc-guard implementation. It is the same as
622      -fsanitize-coverage=trace-pc-guard from llvm 12, but: it is a) inline
623      and b) works from llvm 10.0.1 + onwards :)
624    - new llvm pass: dict2file via AFL_LLVM_DICT2FILE, create afl-fuzz
625      -x dictionary of string comparisons found during compilation
626    - LTO autodict now also collects interesting cmp comparisons,
627      std::string compare + find + ==, bcmp
628    - fix crash in dict2file for integers > 64 bit
629  - custom mutators
630    - added a new custom mutator: symcc -> https://github.com/eurecom-s3/symcc/
631    - added a new custom mutator: libfuzzer that integrates libfuzzer mutations
632    - Our AFL++ Grammar-Mutator is now better integrated into custom_mutators/
633    - added INTROSPECTION support for custom modules
634    - python fuzz function was not optional, fixed
635    - some python mutator speed improvements
636  - afl-cmin/afl-cmin.bash now search first in PATH and last in AFL_PATH
637  - unicornafl synced with upstream version 1.02 (fixes, better rust bindings)
638  - renamed AFL_DEBUG_CHILD_OUTPUT to AFL_DEBUG_CHILD
639  - added AFL_CRASH_EXITCODE env variable to treat a child exitcode as crash
640
641
642### Version ++2.68c (release)
643  - added the GSoC excellent AFL++ grammar mutator by Shengtuo to our
644    custom_mutators/ (see custom_mutators/README.md) - or get it here:
645    https://github.com/AFLplusplus/Grammar-Mutator
646  - a few QOL changes for Apple and its outdated gmake
647  - afl-fuzz:
648    - fix for auto dictionary entries found during fuzzing to not throw out
649      a -x dictionary
650    - added total execs done to plot file
651    - AFL_MAX_DET_EXTRAS env variable added to control the amount of
652      deterministic dict entries without recompiling.
653    - AFL_FORKSRV_INIT_TMOUT env variable added to control the time to wait
654      for the forkserver to come up without the need to increase the overall
655      timeout.
656    - bugfix for cmplog that results in a heap overflow based on target data
657      (thanks to the magma team for reporting!)
658    - write fuzzing setup into out/fuzzer_setup (environment variables and
659      command line)
660  - custom mutators:
661    - added afl_custom_fuzz_count/fuzz_count function to allow specifying
662      the number of fuzz attempts for custom_fuzz
663  - llvm_mode:
664    - ported SanCov to LTO, and made it the default for LTO. better
665      instrumentation locations
666    - Further llvm 12 support (fast moving target like AFL++ :-) )
667    - deprecated LLVM SKIPSINGLEBLOCK env environment
668
669
670### Version ++2.67c (release)
671  - Support for improved AFL++ snapshot module:
672    https://github.com/AFLplusplus/AFL-Snapshot-LKM
673  - Due to the instrumentation needing more memory, the initial memory sizes
674    for -m have been increased
675  - afl-fuzz:
676     - added -F option to allow -M main fuzzers to sync to foreign fuzzers,
677       e.g. honggfuzz or libfuzzer
678     - added -b option to bind to a specific CPU
679     - eliminated CPU affinity race condition for -S/-M runs
680     - expanded havoc mode added, on no cycle finds add extra splicing and
681       MOpt into the mix
682     - fixed a bug in redqueen for strings and made deterministic with -s
683     - Compiletime autodictionary fixes
684  - llvm_mode:
685     - now supports llvm 12
686     - support for AFL_LLVM_ALLOWLIST/AFL_LLVM_DENYLIST (previous
687       AFL_LLVM_WHITELIST and AFL_LLVM_INSTRUMENT_FILE are deprecated and
688       are matched to AFL_LLVM_ALLOWLIST). The format is compatible to llvm
689       sancov, and also supports function matching :)
690     - added neverzero counting to trace-pc/pcgard
691     - fixes for laf-intel float splitting (thanks to mark-griffin for
692       reporting)
693     - fixes for llvm 4.0
694     - skipping ctors and ifuncs for instrumentation
695     - LTO: switch default to the dynamic memory map, set AFL_LLVM_MAP_ADDR
696            for a fixed map address (eg. 0x10000)
697     - LTO: improved stability for persistent mode, no other instrumentation
698            has that advantage
699     - LTO: fixed autodict for long strings
700     - LTO: laf-intel and redqueen/cmplog are now applied at link time
701            to prevent llvm optimizing away the splits
702     - LTO: autodictionary mode is a fixed default now
703     - LTO: instrim instrumentation disabled, only classic support used
704            as it is always better
705     - LTO: env var AFL_LLVM_DOCUMENT_IDS=file will document which edge ID
706            was given to which function during compilation
707     - LTO: single block functions were not implemented by default, fixed
708     - LTO: AFL_LLVM_SKIP_NEVERZERO behaviour was inversed, fixed
709     - setting AFL_LLVM_LAF_SPLIT_FLOATS now activates
710       AFL_LLVM_LAF_SPLIT_COMPARES
711     - support for -E and -shared compilation runs
712  - added honggfuzz mangle as a custom mutator in custom_mutators/honggfuzz
713  - added afl-frida gum solution to examples/afl_frida (mostly imported
714    from https://github.com/meme/hotwax/)
715  - small fixes to afl-plot, afl-whatsup and man page creation
716  - new README, added FAQ
717
718
719### Version ++2.66c (release)
720  - renamed the main branch on Github to "stable"
721  - renamed master/slave to main/secondary
722  - renamed blacklist/whitelist to ignorelist/instrumentlist ->
723    AFL_LLVM_INSTRUMENT_FILE and AFL_GCC_INSTRUMENT_FILE
724  - warn on deprecated environment variables
725  - afl-fuzz:
726     - -S secondary nodes now only sync from the main node to increase
727       performance, the -M main node still syncs from everyone. Added checks
728       that ensure exactly one main node is present and warn otherwise
729     - Add -D after -S to force a secondary to perform deterministic fuzzing
730     - If no main node is present at a sync one secondary node automatically
731       becomes a temporary main node until a real main nodes shows up
732     - Fixed a mayor performance issue we inherited from AFLfast
733     - switched murmur2 hashing and random() for xxh3 and xoshiro256**,
734       resulting in an up to 5.5% speed increase
735     - Resizing the window does not crash afl-fuzz anymore
736     - Ensure that the targets are killed on exit
737     - fix/update to MOpt (thanks to arnow117)
738     - added MOpt dictionary support from repo
739     - added experimental SEEK power schedule. It is EXPLORE with ignoring
740       the runtime and less focus on the length of the test case
741  - llvm_mode:
742    - the default instrumentation is now PCGUARD if the llvm version is >= 7,
743      as it is faster and provides better coverage. The original afl
744      instrumentation can be set via AFL_LLVM_INSTRUMENT=AFL. This is
745      automatically done when the instrument_file list feature is used.
746    - PCGUARD mode is now even better because we made it collision free - plus
747      it has a fixed map size, so it is also faster! :)
748    - some targets want a ld variant for LD that is not gcc/clang but ld,
749      added afl-ld-lto to solve this
750    - lowered minimum required llvm version to 3.4 (except LLVMInsTrim, which
751      needs 3.8.0)
752    - instrument_file list feature now supports wildcards (thanks to sirmc)
753    - small change to cmplog to make it work with current llvm 11-dev
754    - added AFL_LLVM_LAF_ALL, sets all laf-intel settings
755    - LTO instrument_files functionality rewritten, now main, _init etc functions
756      need not to be listed anymore
757    - fixed crash in compare-transform-pass when strcasecmp/strncasecmp was
758      tried to be instrumented with LTO
759    - fixed crash in cmplog with LTO
760    - enable snapshot lkm also for persistent mode
761  - Unicornafl
762    - Added powerPC support from unicorn/next
763    - rust bindings!
764  - CMPLOG/Redqueen now also works for MMAP sharedmem
765  - ensure shmem is released on errors
766  - we moved radamsa to be a custom mutator in ./custom_mutators/. It is not
767    compiled by default anymore.
768  - allow running in /tmp (only unsafe with umask 0)
769  - persistent mode shared memory testcase handover (instead of via
770    files/stdin) - 10-100% performance increase
771  - General support for 64 bit PowerPC, RiscV, Sparc etc.
772  - fix afl-cmin.bash
773  - slightly better performance compilation options for AFL++ and targets
774  - fixed afl-gcc/afl-as that could break on fast systems reusing pids in
775    the same second
776  - added lots of dictionaries from oss-fuzz, go-fuzz and Jakub Wilk
777  - added former post_library examples to examples/custom_mutators/
778  - Dockerfile upgraded to Ubuntu 20.04 Focal and installing llvm 11 and
779    gcc 10 so afl-clang-lto can be build
780
781
782### Version ++2.65c (release):
783  - afl-fuzz:
784     - AFL_MAP_SIZE was not working correctly
785     - better python detection
786     - an old, old bug in AFL that would show negative stability in rare
787       circumstances is now hopefully fixed
788     - AFL_POST_LIBRARY was deprecated, use AFL_CUSTOM_MUTATOR_LIBRARY
789       instead (see docs/custom_mutators.md)
790  - llvm_mode:
791     - afl-clang-fast/lto now do not skip single block functions. This
792       behaviour can be reactivated with AFL_LLVM_SKIPSINGLEBLOCK
793     - if LLVM 11 is installed the posix shm_open+mmap is used and a fixed
794       address for the shared memory map is used as this increases the
795       fuzzing speed
796     - InsTrim now has an LTO version! :-) That is the best and fastest mode!
797     - fixes to LTO mode if instrumented edges > MAP_SIZE
798     - CTX and NGRAM can now be used together
799     - CTX and NGRAM are now also supported in CFG/INSTRIM mode
800     - AFL_LLVM_LAF_TRANSFORM_COMPARES could crash, fixed
801     - added AFL_LLVM_SKIP_NEVERZERO to skip the never zero coverage counter
802       implementation. For targets with few or no loops or heavily called
803       functions. Gives a small performance boost.
804  - qemu_mode:
805    - add information on PIE/PIC load addresses for 32 bit
806    - better dependency checks
807  - gcc_plugin:
808    - better dependency checks
809  - unicorn_mode:
810    - validate_crash_callback can now count non-crashing inputs as crash as well
811    - better submodule handling
812  - afl-showmap: fix for -Q mode
813  - added examples/afl_network_proxy which allows to fuzz a target over the
814    network (not fuzzing tcp/ip services but running afl-fuzz on one system
815    and the target being on an embedded device)
816  - added examples/afl_untracer which does a binary-only fuzzing with the
817    modifications done in memory (intel32/64 and aarch64 support)
818  - added examples/afl_proxy which can be easily used to fuzz and instrument
819    non-standard things
820  - all:
821    - forkserver communication now also used for error reporting
822    - fix 32 bit build options
823    - make clean now leaves qemu-3.1.1.tar.xz and the unicornafl directory
824      intact if in a git/svn checkout - unless "deepclean" is used
825
826
827### Version ++2.64c (release):
828  - llvm_mode LTO mode:
829    - now requires llvm11 - but compiles all targets! :)
830    - autodictionary feature added, enable with `AFL_LLVM_LTO_AUTODICTIONARY`
831    - variable map size usage
832  - afl-fuzz:
833    - variable map size support added (only LTO mode can use this)
834    - snapshot feature usage now visible in UI
835    - Now setting `-L -1` will enable MOpt in parallel to normal mutation.
836      Additionally, this allows to run dictionaries, radamsa and cmplog.
837    - fix for cmplog/redqueen mode if stdin was used
838    - fix for writing a better plot_data file
839  - qemu_mode: fix for persistent mode (which would not terminate or get stuck)
840  - compare-transform/AFL_LLVM_LAF_TRANSFORM_COMPARES now transforms also
841    static global and local variable comparisons (cannot find all though)
842  - extended forkserver: map_size and more information is communicated to
843    afl-fuzz (and afl-fuzz acts accordingly)
844  - new environment variable: AFL_MAP_SIZE to specify the size of the shared map
845  - if AFL_CC/AFL_CXX is set but empty AFL compilers did fail, fixed
846    (this bug is in vanilla AFL too)
847  - added NO_PYTHON flag to disable python support when building afl-fuzz
848  - more refactoring
849
850
851### Version ++2.63c (release):
852
853  ! the repository was moved from vanhauser-thc to AFLplusplus. It is now
854    an own organisation :)
855  ! development and acceptance of PRs now happen only in the dev branch
856    and only occasionally when everything is fine we PR to master
857  - all:
858    - big code changes to make afl-fuzz thread-safe so afl-fuzz can spawn
859      multiple fuzzing threads in the future or even become a library
860    - AFL basic tools now report on the environment variables picked up
861    - more tools get environment variable usage info in the help output
862    - force all output to stdout (some OK/SAY/WARN messages were sent to
863      stdout, some to stderr)
864    - uninstrumented mode uses an internal forkserver ("fauxserver")
865    - now builds with `-D_FORTIFY_SOURCE=2`
866    - drastically reduced number of (de)allocations during fuzzing
867  - afl-fuzz:
868    - python mutator modules and custom mutator modules now use the same
869      interface and hence the API changed
870    - AFL_AUTORESUME will resume execution without the need to specify `-i -`
871    - added experimental power schedules (-p):
872      - mmopt: ignores runtime of queue entries, gives higher weighting to
873               the last 5 queue entries
874      - rare: puts focus on queue entries that hits rare branches, also ignores
875              runtime
876  - llvm_mode:
877    - added SNAPSHOT feature (using https://github.com/AFLplusplus/AFL-Snapshot-LKM)
878    - added Control Flow Integrity sanitizer (AFL_USE_CFISAN)
879    - added AFL_LLVM_INSTRUMENT option to control the instrumentation type
880      easier: DEFAULT, CFG (INSTRIM), LTO, CTX, NGRAM-x (x=2-16)
881    - made USE_TRACE_PC compile obsolete
882  - LTO collision free instrumented added in llvm_mode with afl-clang-lto -
883    this mode is amazing but requires you to build llvm 11 yourself
884  - Added llvm_mode NGRAM prev_loc coverage by Adrean Herrera
885    (https://github.com/adrianherrera/afl-ngram-pass/), activate by setting
886    AFL_LLVM_INSTRUMENT=NGRAM-<value> or AFL_LLVM_NGRAM_SIZE=<value>
887  - Added llvm_mode context sensitive branch coverage, activated by setting
888    AFL_LLVM_INSTRUMENT=CTX or AFL_LLVM_CTX=1
889  - llvm_mode InsTrim mode:
890    - removed workaround for bug where paths were not instrumented and
891      imported fix by author
892    - made skipping 1 block functions an option and is disabled by default,
893      set AFL_LLVM_INSTRIM_SKIPSINGLEBLOCK=1 to re-enable this
894  - qemu_mode:
895    - qemu_mode now uses solely the internal capstone version to fix builds
896      on modern Linux distributions
897    - QEMU now logs routine arguments for CmpLog when the target is x86
898  - afl-tmin:
899    - now supports hang mode `-H` to minimize hangs
900    - fixed potential afl-tmin missbehavior for targets with multiple hangs
901  - Pressing Control-c in afl-cmin did not terminate it for some OS
902  - the custom API was rewritten and is now the same for Python and shared
903    libraries.
904
905
906### Version ++2.62c (release):
907
908  - Important fix for memory allocation functions that result in afl-fuzz
909    not identifying crashes - UPDATE!
910  - Small fix for -E/-V to release the CPU
911  - CmpLog does not need sancov anymore
912
913
914### Version ++2.61c (release):
915
916  - use -march=native if available
917  - most tools now check for mistyped environment variables
918  - gcc 10 is now supported
919  - the memory safety checks are now disabled for a little more speed during
920    fuzzing (only affects creating queue entries), can be toggled in config.h
921  - afl-fuzz:
922     - MOpt out of bounds writing crash fixed
923     - now prints the real python version support compiled in
924     - set stronger performance compile options and little tweaks
925     - Android: prefer bigcores when selecting a CPU
926     - CmpLog forkserver
927     - Redqueen input-2-state mutator (cmp instructions only ATM)
928     - all Python 2+3 versions supported now
929     - changed execs_per_sec in fuzzer_stats from "current" execs per second
930       (which is pointless) to total execs per second
931     - bugfix for dictionary insert stage count (fix via Google repo PR)
932     - added warning if -M is used together with custom mutators with _ONLY option
933     - AFL_TMPDIR checks are now later and better explained if they fail
934  - llvm_mode
935     - InsTrim: three bug fixes:
936        1. (minor) no pointless instrumentation of 1 block functions
937        2. (medium) path bug that leads a few blocks not instrumented that
938           should be
939        3. (major) incorrect prev_loc was written, fixed!
940  - afl-clang-fast:
941     - show in the help output for which llvm version it was compiled for
942     - now does not need to be recompiled between trace-pc and pass
943       instrumentation. compile normally and set AFL_LLVM_USE_TRACE_PC :)
944     - LLVM 11 is supported
945     - CmpLog instrumentation using SanCov (see llvm_mode/README.cmplog.md)
946  - afl-gcc, afl-clang-fast, afl-gcc-fast:
947     - experimental support for undefined behaviour sanitizer UBSAN
948       (set AFL_USE_UBSAN=1)
949     - the instrumentation summary output now also lists activated sanitizers
950     - afl-as: added isatty(2) check back in
951     - added AFL_DEBUG (for upcoming merge)
952  - qemu_mode:
953     - persistent mode is now also available for arm and aarch64
954     - CmpLog instrumentation for QEMU (-c afl-fuzz command line option)
955       for x86, x86_64, arm and aarch64
956     - AFL_PERSISTENT_HOOK callback module for persistent QEMU
957       (see examples/qemu_persistent_hook)
958     - added qemu_mode/README.persistent.md documentation
959     - AFL_ENTRYPOINT now has instruction granularity
960  - afl-cmin is now a sh script (invoking awk) instead of bash for portability
961    the original script is still present as afl-cmin.bash
962  - afl-showmap: -i dir option now allows processing multiple inputs using the
963     forkserver. This is for enhanced speed in afl-cmin.
964  - added blacklist and instrument_filesing function check in all modules of llvm_mode
965  - added fix from Debian project to compile libdislocator and libtokencap
966  - libdislocator: AFL_ALIGNED_ALLOC to force size alignment to max_align_t
967
968
969### Version ++2.60c (release):
970
971  - fixed a critical bug in afl-tmin that was introduced during ++2.53d
972  - added test cases for afl-cmin and afl-tmin to test/test.sh
973  - added ./examples/argv_fuzzing ld_preload library by Kjell Braden
974  - added preeny's desock_dup ld_preload library as
975    ./examples/socket_fuzzing for network fuzzing
976  - added AFL_AS_FORCE_INSTRUMENT environment variable for afl-as - this is
977    for the retrorewrite project
978  - we now set QEMU_SET_ENV from AFL_PRELOAD when qemu_mode is used
979
980
981### Version ++2.59c (release):
982
983  - qbdi_mode: fuzz android native libraries via QBDI framework
984  - unicorn_mode: switched to the new unicornafl, thanks domenukk
985                  (see https://github.com/vanhauser-thc/unicorn)
986  - afl-fuzz:
987     - added radamsa as (an optional) mutator stage (-R[R])
988     - added -u command line option to not unlink the fuzz input file
989     - Python3 support (autodetect)
990     - AFL_DISABLE_TRIM env var to disable the trim stage
991     - CPU affinity support for DragonFly
992  - llvm_mode:
993     - float splitting is now configured via AFL_LLVM_LAF_SPLIT_FLOATS
994     - support for llvm 10 included now (thanks to devnexen)
995  - libtokencap:
996     - support for *BSD/OSX/Dragonfly added
997     - hook common *cmp functions from widely used libraries
998  - compcov:
999     - hook common *cmp functions from widely used libraries
1000     - floating point splitting support for QEMU on x86 targets
1001  - qemu_mode: AFL_QEMU_DISABLE_CACHE env to disable QEMU TranslationBlocks caching
1002  - afl-analyze: added AFL_SKIP_BIN_CHECK support
1003  - better random numbers for gcc_plugin and llvm_mode (thanks to devnexen)
1004  - Dockerfile by courtesy of devnexen
1005  - added regex.dictionary
1006  - qemu and unicorn download scripts now try to download until the full
1007    download succeeded. f*ckin travis fails downloading 40% of the time!
1008  - more support for Android (please test!)
1009  - added the few Android stuff we didnt have already from Google AFL repository
1010  - removed unnecessary warnings
1011
1012
1013### Version ++2.58c (release):
1014
1015  - reverted patch to not unlink and recreate the input file, it resulted in
1016    performance loss of ~10%
1017  - added test/test-performance.sh script
1018  - (re)added gcc_plugin, fast inline instrumentation is not yet finished,
1019    however it includes the instrument_filesing and persistance feature! by hexcoder-
1020  - gcc_plugin tests added to testing framework
1021
1022
1023### Version ++2.54d-2.57c (release):
1024
1025  - we jump to 2.57 instead of 2.55 to catch up with Google's versioning
1026  - persistent mode for QEMU (see qemu_mode/README.md)
1027  - custom mutator library is now an additional mutator, to exclusivly use it
1028    add AFL_CUSTOM_MUTATOR_ONLY (that will trigger the previous behaviour)
1029  - new library qemu_mode/unsigaction which filters sigaction events
1030  - afl-fuzz: new command line option -I to execute a command on a new crash
1031  - no more unlinking the input file, this way the input file can also be a
1032    FIFO or disk partition
1033  - setting LLVM_CONFIG for llvm_mode will now again switch to the selected
1034    llvm version. If your setup is correct.
1035  - fuzzing strategy yields for custom mutator were missing from the UI, added them :)
1036  - added "make tests" which will perform checks to see that all functionality
1037    is working as expected. this is currently the starting point, its not complete :)
1038  - added mutation documentation feature ("make document"), creates afl-fuzz-document
1039    and saves all mutations of the first run on the first file into out/queue/mutations
1040  - libtokencap and libdislocator now compile to the afl_root directory and are
1041    installed to the .../lib/afl directory when present during make install
1042  - more BSD support, e.g. free CPU binding code for FreeBSD (thanks to devnexen)
1043  - reducing duplicate code in afl-fuzz
1044  - added "make help"
1045  - removed compile warnings from python internal stuff
1046  - added man page for afl-clang-fast[++]
1047  - updated documentation
1048  - Wine mode to run Win32 binaries with the QEMU instrumentation (-W)
1049  - CompareCoverage for ARM target in QEMU/Unicorn
1050  - laf-intel in llvm_mode now also handles floating point comparisons
1051
1052
1053### Version ++2.54c (release):
1054
1055  - big code refactoring:
1056    * all includes are now in include/
1057    * all AFL sources are now in src/ - see src/README.md
1058    * afl-fuzz was split up in various individual files for including
1059      functionality in other programs (e.g. forkserver, memory map, etc.)
1060      for better readability.
1061    * new code indention everywhere
1062  - auto-generating man pages for all (main) tools
1063  - added AFL_FORCE_UI to show the UI even if the terminal is not detected
1064  - llvm 9 is now supported (still needs testing)
1065  - Android is now supported (thank to JoeyJiao!) - still need to modify the Makefile though
1066  - fix building qemu on some Ubuntus (thanks to floyd!)
1067  - custom mutator by a loaded library is now supported (thanks to kyakdan!)
1068  - added PR that includes peak_rss_mb and slowest_exec_ms in the fuzzer_stats report
1069  - more support for *BSD (thanks to devnexen!)
1070  - fix building on *BSD (thanks to tobias.kortkamp for the patch)
1071  - fix for a few features to support different map sized than 2^16
1072  - afl-showmap: new option -r now shows the real values in the buckets (stock
1073    AFL never did), plus shows tuple content summary information now
1074  - small docu updates
1075  - NeverZero counters for QEMU
1076  - NeverZero counters for Unicorn
1077  - CompareCoverage Unicorn
1078  - immediates-only instrumentation for CompareCoverage
1079
1080
1081### Version ++2.53c (release):
1082
1083  - README is now README.md
1084  - imported the few minor changes from the 2.53b release
1085  - unicorn_mode got added - thanks to domenukk for the patch!
1086  - fix llvm_mode AFL_TRACE_PC with modern llvm
1087  - fix a crash in qemu_mode which also exists in stock afl
1088  - added libcompcov, a laf-intel implementation for qemu! :)
1089    see qemu_mode/libcompcov/README.libcompcov.md
1090  - afl-fuzz now displays the selected core in the status screen (blue {#})
1091  - updated afl-fuzz and afl-system-config for new scaling governor location
1092    in modern kernels
1093  - using the old ineffective afl-gcc will now show a deprecation warning
1094  - all queue, hang and crash files now have their discovery time in their name
1095  - if llvm_mode was compiled, afl-clang/afl-clang++ will point to these
1096    instead of afl-gcc
1097  - added instrim, a much faster llvm_mode instrumentation at the cost of
1098    path discovery. See llvm_mode/README.instrim.md (https://github.com/csienslab/instrim)
1099  - added MOpt (github.com/puppet-meteor/MOpt-AFL) mode, see docs/README.MOpt.md
1100  - added code to make it more portable to other platforms than Intel Linux
1101  - added never zero counters for afl-gcc and optionally (because of an
1102    optimization issue in llvm < 9) for llvm_mode (AFL_LLVM_NEVER_ZERO=1)
1103  - added a new doc about binary only fuzzing: docs/binaryonly_fuzzing.txt
1104  - more cpu power for afl-system-config
1105  - added forkserver patch to afl-tmin, makes it much faster (originally from
1106    github.com/nccgroup/TriforceAFL)
1107  - added instrument_files support for llvm_mode via AFL_LLVM_WHITELIST to allow
1108    only to instrument what is actually interesting. Gives more speed and less
1109    map pollution (originally by choller@mozilla)
1110  - added Python Module mutator support, python2.7-dev is autodetected.
1111    see docs/python_mutators.txt (originally by choller@mozilla)
1112  - added AFL_CAL_FAST for slow applications and AFL_DEBUG_CHILD_OUTPUT for
1113    debugging
1114  - added -V time and -E execs option to better comparison runs, runs afl-fuzz
1115    for a specific time/executions.
1116  - added a -s seed switch to allow AFL run with a fixed initial
1117    seed that is not updated. This is good for performance and path discovery
1118    tests as the random numbers are deterministic then
1119  - llvm_mode LAF_... env variables can now be specified as AFL_LLVM_LAF_...
1120    that is longer but in line with other llvm specific env vars
1121
1122
1123### Version ++2.52c (2019-06-05):
1124
1125  - Applied community patches. See docs/PATCHES for the full list.
1126    LLVM and Qemu modes are now faster.
1127    Important changes:
1128      afl-fuzz: -e EXTENSION commandline option
1129      llvm_mode: LAF-intel performance (needs activation, see llvm/README.laf-intel.md)
1130      a few new environment variables for afl-fuzz, llvm and qemu, see docs/env_variables.md
1131  - Added the power schedules of AFLfast by Marcel Boehme, but set the default
1132    to the AFL schedule, not to the FAST schedule. So nothing changes unless
1133    you use the new -p option :-) - see docs/power_schedules.md
1134  - added afl-system-config script to set all system performance options for fuzzing
1135  - llvm_mode works with llvm 3.9 up to including 8 !
1136  - qemu_mode got upgraded from 2.1 to 3.1 - incorporated from
1137    https://github.com/andreafioraldi/afl and with community patches added
1138
1139
1140### Version 2.52b (2017-11-04):
1141
1142  - Upgraded QEMU patches from 2.3.0 to 2.10.0. Required troubleshooting
1143    several weird issues. All the legwork done by Andrew Griffiths.
1144
1145  - Added setsid to afl-showmap. See the notes for 2.51b.
1146
1147  - Added target mode (deferred, persistent, qemu, etc) to fuzzer_stats.
1148    Requested by Jakub Wilk.
1149
1150  - afl-tmin should now save a partially minimized file when Ctrl-C
1151    is pressed. Suggested by Jakub Wilk.
1152
1153  - Added an option for afl-analyze to dump offsets in hex. Suggested by
1154    Jakub Wilk.
1155
1156  - Added support for parameters in triage_crashes.sh. Patch by Adam of
1157    DC949.
1158
1159### Version 2.51b (2017-08-30):
1160
1161  - Made afl-tmin call setsid to prevent glibc traceback junk from showing
1162    up on the terminal in some distros. Suggested by Jakub Wilk.
1163
1164### Version 2.50b (2017-08-19):
1165
1166  - Fixed an interesting timing corner case spotted by Jakub Wilk.
1167
1168  - Addressed a libtokencap / pthreads incompatibility issue. Likewise, spotted
1169    by Jakub Wilk.
1170
1171  - Added a mention of afl-kit and Pythia.
1172
1173  - Added AFL_FAST_CAL.
1174
1175  - In-place resume now preserves .synced. Suggested by Jakub Wilk.
1176
1177### Version 2.49b (2017-07-18):
1178
1179  - Added AFL_TMIN_EXACT to allow path constraint for crash minimization.
1180
1181  - Added dates for releases (retroactively for all of 2017).
1182
1183### Version 2.48b (2017-07-17):
1184
1185  - Added AFL_ALLOW_TMP to permit some scripts to run in /tmp.
1186
1187  - Fixed cwd handling in afl-analyze (similar to the quirk in afl-tmin).
1188
1189  - Made it possible to point -o and -f to the same file in afl-tmin.
1190
1191### Version 2.47b (2017-07-14):
1192
1193  - Fixed cwd handling in afl-tmin. Spotted by Jakub Wilk.
1194
1195### Version 2.46b (2017-07-10):
1196
1197  - libdislocator now supports AFL_LD_NO_CALLOC_OVER for folks who do not
1198    want to abort on calloc() overflows.
1199
1200  - Made a minor fix to libtokencap. Reported by Daniel Stender.
1201
1202  - Added a small JSON dictionary, inspired on a dictionary done by Jakub Wilk.
1203
1204### Version 2.45b (2017-07-04):
1205
1206  - Added strstr, strcasestr support to libtokencap. Contributed by
1207    Daniel Hodson.
1208
1209  - Fixed a resumption offset glitch spotted by Jakub Wilk.
1210
1211  - There are definitely no bugs in afl-showmap -c now.
1212
1213### Version 2.44b (2017-06-28):
1214
1215  - Added a visual indicator of ASAN / MSAN mode when compiling. Requested
1216    by Jakub Wilk.
1217
1218  - Added support for afl-showmap coredumps (-c). Suggested by Jakub Wilk.
1219
1220  - Added LD_BIND_NOW=1 for afl-showmap by default. Although not really useful,
1221    it reportedly helps reproduce some crashes. Suggested by Jakub Wilk.
1222
1223  - Added a note about allocator_may_return_null=1 not always working with
1224    ASAN. Spotted by Jakub Wilk.
1225
1226### Version 2.43b (2017-06-16):
1227
1228  - Added AFL_NO_ARITH to aid in the fuzzing of text-based formats.
1229    Requested by Jakub Wilk.
1230
1231### Version 2.42b (2017-06-02):
1232
1233  - Renamed the R() macro to avoid a problem with llvm_mode in the latest
1234    versions of LLVM. Fix suggested by Christian Holler.
1235
1236### Version 2.41b (2017-04-12):
1237
1238  - Addressed a major user complaint related to timeout detection. Timing out
1239    inputs are now binned as "hangs" only if they exceed a far more generous
1240    time limit than the one used to reject slow paths.
1241
1242### Version 2.40b (2017-04-02):
1243
1244  - Fixed a minor oversight in the insertion strategy for dictionary words.
1245    Spotted by Andrzej Jackowski.
1246
1247  - Made a small improvement to the havoc block insertion strategy.
1248
1249  - Adjusted color rules for "is it done yet?" indicators.
1250
1251### Version 2.39b (2017-02-02):
1252
1253  - Improved error reporting in afl-cmin. Suggested by floyd.
1254
1255  - Made a minor tweak to trace-pc-guard support. Suggested by kcc.
1256
1257  - Added a mention of afl-monitor.
1258
1259### Version 2.38b (2017-01-22):
1260
1261  - Added -mllvm -sanitizer-coverage-block-threshold=0 to trace-pc-guard
1262    mode, as suggested by Kostya Serebryany.
1263
1264### Version 2.37b (2017-01-22):
1265
1266  - Fixed a typo. Spotted by Jakub Wilk.
1267
1268  - Fixed support for make install when using trace-pc. Spotted by
1269    Kurt Roeckx.
1270
1271  - Switched trace-pc to trace-pc-guard, which should be considerably
1272    faster and is less quirky. Kudos to Konstantin Serebryany (and sorry
1273    for dragging my feet).
1274
1275    Note that for some reason, this mode doesn't perform as well as
1276    "vanilla" afl-clang-fast / afl-clang.
1277
1278### Version 2.36b (2017-01-14):
1279
1280  - Fixed a cosmetic bad free() bug when aborting -S sessions. Spotted
1281    by Johannes S.
1282
1283  - Made a small change to afl-whatsup to sort fuzzers by name.
1284
1285  - Fixed a minor issue with malloc(0) in libdislocator. Spotted by
1286    Rene Freingruber.
1287
1288  - Changed the clobber pattern in libdislocator to a slightly more
1289    reliable one. Suggested by Rene Freingruber.
1290
1291  - Added a note about THP performance. Suggested by Sergey Davidoff.
1292
1293  - Added a somewhat unofficial support for running afl-tmin with a
1294    baseline "mask" that causes it to minimize only for edges that
1295    are unique to the input file, but not to the "boring" baseline.
1296    Suggested by Sami Liedes.
1297
1298  - "Fixed" a getPassName() problem with newer versions of clang.
1299    Reported by Craig Young and several other folks.
1300
1301  Yep, I know I have a backlog on several other feature requests.
1302  Stay tuned!
1303
1304### Version 2.35b:
1305
1306  - Fixed a minor cmdline reporting glitch, spotted by Leo Barnes.
1307
1308  - Fixed a silly bug in libdislocator. Spotted by Johannes Schultz.
1309
1310### Version 2.34b:
1311
1312  - Added a note about afl-tmin to technical_details.txt.
1313
1314  - Added support for AFL_NO_UI, as suggested by Leo Barnes.
1315
1316### Version 2.33b:
1317
1318  - Added code to strip -Wl,-z,defs and -Wl,--no-undefined for afl-clang-fast,
1319    since they interfere with -shared. Spotted and diagnosed by Toby Hutton.
1320
1321  - Added some fuzzing tips for Android.
1322
1323### Version 2.32b:
1324
1325  - Added a check for AFL_HARDEN combined with AFL_USE_*SAN. Suggested by
1326    Hanno Boeck.
1327
1328  - Made several other cosmetic adjustments to cycle timing in the wake of the
1329    big tweak made in 2.31b.
1330
1331### Version 2.31b:
1332
1333  - Changed havoc cycle counts for a marked performance boost, especially
1334    with -S / -d. See the discussion of FidgetyAFL in:
1335
1336    https://groups.google.com/forum/#!topic/afl-users/fOPeb62FZUg
1337
1338    While this does not implement the approach proposed by the authors of
1339    the CCS paper, the solution is a result of digging into that research;
1340    more improvements may follow as I do more experiments and get more
1341    definitive data.
1342
1343### Version 2.30b:
1344
1345  - Made minor improvements to persistent mode to avoid the remote
1346    possibility of "no instrumentation detected" issues with very low
1347    instrumentation densities.
1348
1349  - Fixed a minor glitch with a leftover process in persistent mode.
1350    Reported by Jakub Wilk and Daniel Stender.
1351
1352  - Made persistent mode bitmaps a bit more consistent and adjusted the way
1353    this is shown in the UI, especially in persistent mode.
1354
1355### Version 2.29b:
1356
1357  - Made a minor #include fix to llvm_mode. Suggested by Jonathan Metzman.
1358
1359  - Made cosmetic updates to the docs.
1360
1361### Version 2.28b:
1362
1363  - Added "life pro tips" to docs/.
1364
1365  - Moved testcases/_extras/ to dictionaries/ for visibility.
1366
1367  - Made minor improvements to install scripts.
1368
1369  - Added an important safety tip.
1370
1371### Version 2.27b:
1372
1373  - Added libtokencap, a simple feature to intercept strcmp / memcmp and
1374    generate dictionary entries that can help extend coverage.
1375
1376  - Moved libdislocator to its own dir, added README.md.
1377
1378  - The demo in examples/instrumented_cmp is no more.
1379
1380### Version 2.26b:
1381
1382  - Made a fix for libdislocator.so to compile on MacOS X.
1383
1384  - Added support for DYLD_INSERT_LIBRARIES.
1385
1386  - Renamed AFL_LD_PRELOAD to AFL_PRELOAD.
1387
1388### Version 2.25b:
1389
1390  - Made some cosmetic updates to libdislocator.so, renamed one env
1391    variable.
1392
1393### Version 2.24b:
1394
1395  - Added libdislocator.so, an experimental, abusive allocator. Try
1396    it out with AFL_LD_PRELOAD=/path/to/libdislocator.so when running
1397    afl-fuzz.
1398
1399### Version 2.23b:
1400
1401  - Improved the stability metric for persistent mode binaries. Problem
1402    spotted by Kurt Roeckx.
1403
1404  - Made a related improvement that may bring the metric to 100% for those
1405    targets.
1406
1407### Version 2.22b:
1408
1409  - Mentioned the potential conflicts between MSAN / ASAN and FORTIFY_SOURCE.
1410    There is no automated check for this, since some distros may implicitly
1411    set FORTIFY_SOURCE outside of the compiler's argv[].
1412
1413  - Populated the support for AFL_LD_PRELOAD to all companion tools.
1414
1415  - Made a change to the handling of ./afl-clang-fast -v. Spotted by
1416    Jan Kneschke.
1417
1418### Version 2.21b:
1419
1420  - Added some crash reporting notes for Solaris in docs/INSTALL, as
1421    investigated by Martin Carpenter.
1422
1423  - Fixed a minor UI mix-up with havoc strategy stats.
1424
1425### Version 2.20b:
1426
1427  - Revamped the handling of variable paths, replacing path count with a
1428    "stability" score to give users a much better signal. Based on the
1429    feedback from Vegard Nossum.
1430
1431  - Made a stability improvement to the syncing behavior with resuming
1432    fuzzers. Based on the feedback from Vegard.
1433
1434  - Changed the UI to include current input bitmap density along with
1435    total density. Ditto.
1436
1437  - Added experimental support for parallelizing -M.
1438
1439### Version 2.19b:
1440
1441  - Made a fix to make sure that auto CPU binding happens at non-overlapping
1442    times.
1443
1444### Version 2.18b:
1445
1446  - Made several performance improvements to has_new_bits() and
1447    classify_counts(). This should offer a robust performance bump with
1448    fast targets.
1449
1450### Version 2.17b:
1451
1452  - Killed the error-prone and manual -Z option. On Linux, AFL will now
1453    automatically bind to the first free core (or complain if there are no
1454    free cores left).
1455
1456  - Made some doc updates along these lines.
1457
1458### Version 2.16b:
1459
1460  - Improved support for older versions of clang (hopefully without
1461    breaking anything).
1462
1463  - Moved version data from Makefile to config.h. Suggested by
1464    Jonathan Metzman.
1465
1466### Version 2.15b:
1467
1468  - Added a README section on looking for non-crashing bugs.
1469
1470  - Added license data to several boring files. Contributed by
1471    Jonathan Metzman.
1472
1473### Version 2.14b:
1474
1475  - Added FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION as a macro defined when
1476    compiling with afl-gcc and friends. Suggested by Kostya Serebryany.
1477
1478  - Refreshed some of the non-x86 docs.
1479
1480### Version 2.13b:
1481
1482  - Fixed a spurious build test error with trace-pc and llvm_mode/Makefile.
1483    Spotted by Markus Teufelberger.
1484
1485  - Fixed a cosmetic issue with afl-whatsup. Spotted by Brandon Perry.
1486
1487### Version 2.12b:
1488
1489  - Fixed a minor issue in afl-tmin that can make alphabet minimization less
1490    efficient during passes > 1. Spotted by Daniel Binderman.
1491
1492### Version 2.11b:
1493
1494  - Fixed a minor typo in instrumented_cmp, spotted by Hanno Eissfeldt.
1495
1496  - Added a missing size check for deterministic insertion steps.
1497
1498  - Made an improvement to afl-gotcpu when -Z not used.
1499
1500  - Fixed a typo in post_library_png.so.c in examples/. Spotted by Kostya
1501    Serebryany.
1502
1503### Version 2.10b:
1504
1505  - Fixed a minor core counting glitch, reported by Tyler Nighswander.
1506
1507### Version 2.09b:
1508
1509  - Made several documentation updates.
1510
1511  - Added some visual indicators to promote and simplify the use of -Z.
1512
1513### Version 2.08b:
1514
1515  - Added explicit support for -m32 and -m64 for llvm_mode. Inspired by
1516    a request from Christian Holler.
1517
1518  - Added a new benchmarking option, as requested by Kostya Serebryany.
1519
1520### Version 2.07b:
1521
1522  - Added CPU affinity option (-Z) on Linux. With some caution, this can
1523    offer a significant (10%+) performance bump and reduce jitter.
1524    Proposed by Austin Seipp.
1525
1526  - Updated afl-gotcpu to use CPU affinity where supported.
1527
1528  - Fixed confusing CPU_TARGET error messages with QEMU build. Spotted by
1529    Daniel Komaromy and others.
1530
1531### Version 2.06b:
1532
1533  - Worked around LLVM persistent mode hiccups with -shared code.
1534    Contributed by Christian Holler.
1535
1536  - Added __AFL_COMPILER as a convenient way to detect that something is
1537    built under afl-gcc / afl-clang / afl-clang-fast and enable custom
1538    optimizations in your code. Suggested by Pedro Corte-Real.
1539
1540  - Upstreamed several minor changes developed by Franjo Ivancic to
1541    allow AFL to be built as a library. This is fairly use-specific and
1542    may have relatively little appeal to general audiences.
1543
1544### Version 2.05b:
1545
1546  - Put __sanitizer_cov_module_init & co behind #ifdef to avoid problems
1547    with ASAN. Spotted by Christian Holler.
1548
1549### Version 2.04b:
1550
1551  - Removed indirect-calls coverage from -fsanitize-coverage (since it's
1552    redundant). Spotted by Kostya Serebryany.
1553
1554### Version 2.03b:
1555
1556  - Added experimental -fsanitize-coverage=trace-pc support that goes with
1557    some recent additions to LLVM, as implemented by Kostya Serebryany.
1558    Right now, this is cumbersome to use with common build systems, so
1559    the mode remains undocumented.
1560
1561  - Made several substantial improvements to better support non-standard
1562    map sizes in LLVM mode.
1563
1564  - Switched LLVM mode to thread-local execution tracing, which may offer
1565    better results in some multithreaded apps.
1566
1567  - Fixed a minor typo, reported by Heiko Eissfeldt.
1568
1569  - Force-disabled symbolization for ASAN, as suggested by Christian Holler.
1570
1571  - AFL_NOX86 renamed to AFL_NO_X86 for consistency.
1572
1573  - Added AFL_LD_PRELOAD to allow LD_PRELOAD to be set for targets without
1574    affecting AFL itself. Suggested by Daniel Godas-Lopez.
1575
1576### Version 2.02b:
1577
1578  - Fixed a "lcamtuf can't count to 16" bug in the havoc stage. Reported
1579    by Guillaume Endignoux.
1580
1581### Version 2.01b:
1582
1583  - Made an improvement to cycle counter color coding, based on feedback
1584    from Shai Sarfaty.
1585
1586  - Added a mention of aflize to sister_projects.txt.
1587
1588  - Fixed an installation issue with afl-as, as spotted by ilovezfs.
1589
1590### Version 2.00b:
1591
1592  - Cleaned up color handling after a minor snafu in 1.99b (affecting some
1593    terminals).
1594
1595  - Made minor updates to the documentation.
1596
1597### Version 1.99b:
1598
1599  - Substantially revamped the output and the internal logic of afl-analyze.
1600
1601  - Cleaned up some of the color handling code and added support for
1602    background colors.
1603
1604  - Removed some stray files (oops).
1605
1606  - Updated docs to better explain afl-analyze.
1607
1608### Version 1.98b:
1609
1610  - Improved to "boring string" detection in afl-analyze.
1611
1612  - Added technical_details.txt for afl-analyze.
1613
1614### Version 1.97b:
1615
1616  - Added afl-analyze, a nifty tool to analyze the structure of a file
1617    based on the feedback from AFL instrumentation. This is kinda experimental,
1618    so field reports welcome.
1619
1620  - Added a mention of afl-cygwin.
1621
1622  - Fixed a couple of typos, as reported by Jakub Wilk and others.
1623
1624### Version 1.96b:
1625
1626  - Added -fpic to CFLAGS for the clang plugin, as suggested by Hanno Boeck.
1627
1628  - Made another clang change (IRBuilder) suggested by Jeff Trull.
1629
1630  - Fixed several typos, spotted by Jakub Wilk.
1631
1632  - Added support for AFL_SHUFFLE_QUEUE, based on discussions with
1633    Christian Holler.
1634
1635### Version 1.95b:
1636
1637  - Fixed a harmless bug when handling -B. Spotted by Jacek Wielemborek.
1638
1639  - Made the exit message a bit more accurate when AFL_EXIT_WHEN_DONE is set.
1640
1641  - Added some error-checking for old-style forkserver syntax. Suggested by
1642    Ben Nagy.
1643
1644  - Switched from exit() to _exit() in injected code to avoid snafus with
1645    destructors in C++ code. Spotted by sunblate.
1646
1647  - Made a change to avoid spuriously setting __AFL_SHM_ID when
1648    AFL_DUMB_FORKSRV is set in conjunction with -n. Spotted by Jakub Wilk.
1649
1650### Version 1.94b:
1651
1652  - Changed allocator alignment to improve support for non-x86 systems (now
1653    that llvm_mode makes this more feasible).
1654
1655  - Fixed a minor typo in afl-cmin. Spotted by Jonathan Neuschafer.
1656
1657  - Fixed an obscure bug that would affect people trying to use afl-gcc
1658    with $TMP set but $TMPDIR absent. Spotted by Jeremy Barnes.
1659
1660### Version 1.93b:
1661
1662  - Hopefully fixed a problem with MacOS X and persistent mode, spotted by
1663    Leo Barnes.
1664
1665### Version 1.92b:
1666
1667  - Made yet another C++ fix (namespaces). Reported by Daniel Lockyer.
1668
1669### Version 1.91b:
1670
1671  - Made another fix to make 1.90b actually work properly with C++ (d'oh).
1672    Problem spotted by Daniel Lockyer.
1673
1674### Version 1.90b:
1675
1676  - Fixed a minor typo spotted by Kai Zhao; and made several other minor updates
1677    to docs.
1678
1679  - Updated the project URL for python-afl. Requested by Jakub Wilk.
1680
1681  - Fixed a potential problem with deferred mode signatures getting optimized
1682    out by the linker (with --gc-sections).
1683
1684### Version 1.89b:
1685
1686  - Revamped the support for persistent and deferred forkserver modes.
1687    Both now feature simpler syntax and do not require companion env
1688    variables. Suggested by Jakub Wilk.
1689
1690  - Added a bit more info about afl-showmap. Suggested by Jacek Wielemborek.
1691
1692### Version 1.88b:
1693
1694  - Made AFL_EXIT_WHEN_DONE work in non-tty mode. Issue spotted by
1695    Jacek Wielemborek.
1696
1697### Version 1.87b:
1698
1699  - Added QuickStartGuide.txt, a one-page quick start doc.
1700
1701  - Fixed several typos spotted by Dominique Pelle.
1702
1703  - Revamped several parts of README.
1704
1705### Version 1.86b:
1706
1707  - Added support for AFL_SKIP_CRASHES, which is a very hackish solution to
1708    the problem of resuming sessions with intermittently crashing inputs.
1709
1710  - Removed the hard-fail terminal size check, replaced with a dynamic
1711    warning shown in place of the UI. Based on feedback from Christian Holler.
1712
1713  - Fixed a minor typo in show_stats. Spotted by Dingbao Xie.
1714
1715### Version 1.85b:
1716
1717  - Fixed a garbled sentence in notes on parallel fuzzing. Thanks to Jakub Wilk.
1718
1719  - Fixed a minor glitch in afl-cmin. Spotted by Jonathan Foote.
1720
1721### Version 1.84b:
1722
1723  - Made SIMPLE_FILES behave as expected when naming backup directories for
1724    crashes and hangs.
1725
1726  - Added the total number of favored paths to fuzzer_stats. Requested by
1727    Ben Nagy.
1728
1729  - Made afl-tmin, afl-fuzz, and afl-cmin reject negative values passed to
1730    -t and -m, since they generally won't work as expected.
1731
1732  - Made a fix for no lahf / sahf support on older versions of FreeBSD.
1733    Patch contributed by Alex Moneger.
1734
1735### Version 1.83b:
1736
1737  - Fixed a problem with xargs -d on non-Linux systems in afl-cmin. Spotted by
1738    teor2345 and Ben Nagy.
1739
1740  - Fixed an implicit declaration in LLVM mode on MacOS X. Reported by
1741    Kai Zhao.
1742
1743### Version 1.82b:
1744
1745  - Fixed a harmless but annoying race condition in persistent mode - signal
1746    delivery is a bit more finicky than I thought.
1747
1748  - Updated the documentation to explain persistent mode a bit better.
1749
1750  - Tweaked AFL_PERSISTENT to force AFL_NO_VAR_CHECK.
1751
1752### Version 1.81b:
1753
1754  - Added persistent mode for in-process fuzzing. See llvm_mode/README.llvm.
1755    Inspired by Kostya Serebryany and Christian Holler.
1756
1757  - Changed the in-place resume code to preserve crashes/README.txt. Suggested
1758    by Ben Nagy.
1759
1760  - Included a potential fix for LLVM mode issues on MacOS X, based on the
1761    investigation done by teor2345.
1762
1763### Version 1.80b:
1764
1765  - Made afl-cmin tolerant of whitespaces in filenames. Suggested by
1766    Jonathan Neuschafer and Ketil Froyn.
1767
1768  - Added support for AFL_EXIT_WHEN_DONE, as suggested by Michael Rash.
1769
1770### Version 1.79b:
1771
1772  - Added support for dictionary levels, see testcases/README.testcases.
1773
1774  - Reworked the SQL dictionary to use levels.
1775
1776  - Added a note about Preeny.
1777
1778### Version 1.78b:
1779
1780  - Added a dictionary for PDF, contributed by Ben Nagy.
1781
1782  - Added several references to afl-cov, a new tool by Michael Rash.
1783
1784  - Fixed a problem with crash reporter detection on MacOS X, as reported by
1785    Louis Dassy.
1786
1787### Version 1.77b:
1788
1789  - Extended the -x option to support single-file dictionaries.
1790
1791  - Replaced factory-packaged dictionaries with file-based variants.
1792
1793  - Removed newlines from HTML keywords in testcases/_extras/html/.
1794
1795### Version 1.76b:
1796
1797  - Very significantly reduced the number of duplicate execs during
1798    deterministic checks, chiefly in int16 and int32 stages. Confirmed
1799    identical path yields. This should improve early-stage efficiency by
1800    around 5-10%.
1801
1802  - Reduced the likelihood of duplicate non-deterministic execs by
1803    bumping up lowest stacking factor from 1 to 2. Quickly confirmed
1804    that this doesn't seem to have significant impact on coverage with
1805    libpng.
1806
1807  - Added a note about integrating afl-fuzz with third-party tools.
1808
1809### Version 1.75b:
1810
1811  - Improved argv_fuzzing to allow it to emit empty args. Spotted by Jakub
1812    Wilk.
1813
1814  - afl-clang-fast now defines __AFL_HAVE_MANUAL_INIT. Suggested by Jakub Wilk.
1815
1816  - Fixed a libtool-related bug with afl-clang-fast that would make some
1817    ./configure invocations generate incorrect output. Spotted by Jakub Wilk.
1818
1819  - Removed flock() on Solaris. This means no locking on this platform,
1820    but so be it. Problem reported by Martin Carpenter.
1821
1822  - Fixed a typo. Reported by Jakub Wilk.
1823
1824### Version 1.74b:
1825
1826  - Added an example argv[] fuzzing wrapper in examples/argv_fuzzing.
1827    Reworked the bash example to be faster, too.
1828
1829  - Clarified llvm_mode prerequisites for FreeBSD.
1830
1831  - Improved afl-tmin to use /tmp if cwd is not writeable.
1832
1833  - Removed redundant includes for sys/fcntl.h, which caused warnings with
1834    some nitpicky versions of libc.
1835
1836  - Added a corpus of basic HTML tags that parsers are likely to pay attention
1837    to (no attributes).
1838
1839  - Added EP_EnabledOnOptLevel0 to llvm_mode, so that the instrumentation is
1840    inserted even when AFL_DONT_OPTIMIZE=1 is set.
1841
1842  - Switched qemu_mode to use the newly-released QEMU 2.3.0, which contains
1843    a couple of minor bugfixes.
1844
1845### Version 1.73b:
1846
1847  - Fixed a pretty stupid bug in effector maps that could sometimes cause
1848    AFL to fuzz slightly more than necessary; and in very rare circumstances,
1849    could lead to SEGV if eff_map is aligned with page boundary and followed
1850    by an unmapped page. Spotted by Jonathan Gray.
1851
1852### Version 1.72b:
1853
1854  - Fixed a glitch in non-x86 install, spotted by Tobias Ospelt.
1855
1856  - Added a minor safeguard to llvm_mode Makefile following a report from
1857    Kai Zhao.
1858
1859### Version 1.71b:
1860
1861  - Fixed a bug with installed copies of AFL trying to use QEMU mode. Spotted
1862    by G.M. Lime.
1863
1864  - Added last find / crash / hang times to fuzzer_stats, suggested by
1865    Richard Hipp.
1866
1867  - Fixed a typo, thanks to Jakub Wilk.
1868
1869### Version 1.70b:
1870
1871  - Modified resumption code to reuse the original timeout value when resuming
1872    a session if -t is not given. This prevents timeout creep in continuous
1873    fuzzing.
1874
1875  - Added improved error messages for failed handshake when AFL_DEFER_FORKSRV
1876    is set.
1877
1878  - Made a slight improvement to llvm_mode/Makefile based on feedback from
1879    Jakub Wilk.
1880
1881  - Refreshed several bits of documentation.
1882
1883  - Added a more prominent note about the MacOS X trade-offs to Makefile.
1884
1885### Version 1.69b:
1886
1887  - Added support for deferred initialization in LLVM mode. Suggested by
1888    Richard Godbee.
1889
1890### Version 1.68b:
1891
1892  - Fixed a minor PRNG glitch that would make the first seconds of a fuzzing
1893    job deterministic. Thanks to Andreas Stieger.
1894
1895  - Made tmp[] static in the LLVM runtime to keep Valgrind happy (this had
1896    no impact on anything else). Spotted by Richard Godbee.
1897
1898  - Clarified the footnote in README.
1899
1900### Version 1.67b:
1901
1902  - Made one more correction to llvm_mode Makefile, spotted by Jakub Wilk.
1903
1904### Version 1.66b:
1905
1906  - Added CC / CXX support to llvm_mode Makefile. Requested by Charlie Eriksen.
1907
1908  - Fixed 'make clean' with gmake. Suggested by Oliver Schneider.
1909
1910  - Fixed 'make -j n clean all'. Suggested by Oliver Schneider.
1911
1912  - Removed build date and time from banners to give people deterministic
1913    builds. Requested by Jakub Wilk.
1914
1915### Version 1.65b:
1916
1917  - Fixed a snafu with some leftover code in afl-clang-fast.
1918
1919  - Corrected even moar typos.
1920
1921### Version 1.64b:
1922
1923  - Further simplified afl-clang-fast runtime by reverting .init_array to
1924    __attribute__((constructor(0)). This should improve compatibility with
1925    non-ELF platforms.
1926
1927  - Fixed a problem with afl-clang-fast and -shared libraries. Simplified
1928    the code by getting rid of .preinit_array and replacing it with a .comm
1929    object. Problem reported by Charlie Eriksen.
1930
1931  - Removed unnecessary instrumentation density adjustment for the LLVM mode.
1932    Reported by Jonathan Neuschafer.
1933
1934### Version 1.63b:
1935
1936  - Updated cgroups_asan/ with a new version from Sam, made a couple changes
1937    to streamline it and keep parallel AFL instances in separate groups.
1938
1939  - Fixed typos, thanks to Jakub Wilk.
1940
1941### Version 1.62b:
1942
1943  - Improved the handling of -x in afl-clang-fast,
1944
1945  - Improved the handling of low AFL_INST_RATIO settings for QEMU and
1946    LLVM modes.
1947
1948  - Fixed the llvm-config bug for good (thanks to Tobias Ospelt).
1949
1950### Version 1.61b:
1951
1952  - Fixed an obscure bug compiling OpenSSL with afl-clang-fast. Patch by
1953    Laszlo Szekeres.
1954
1955  - Fixed a 'make install' bug on non-x86 systems, thanks to Tobias Ospelt.
1956
1957  - Fixed a problem with half-broken llvm-config on Odroid, thanks to
1958    Tobias Ospelt. (There is another odd bug there that hasn't been fully
1959    fixed - TBD).
1960
1961### Version 1.60b:
1962
1963  - Allowed examples/llvm_instrumentation/ to graduate to llvm_mode/.
1964
1965  - Removed examples/arm_support/, since it's completely broken and likely
1966    unnecessary with LLVM support in place.
1967
1968  - Added ASAN cgroups script to examples/asan_cgroups/, updated existing
1969    docs. Courtesy Sam Hakim and David A. Wheeler.
1970
1971  - Refactored afl-tmin to reduce the number of execs in common use cases.
1972    Ideas from Jonathan Neuschafer and Turo Lamminen.
1973
1974  - Added a note about CLAs at the bottom of README.
1975
1976  - Renamed testcases_readme.txt to README.testcases for some semblance of
1977    consistency.
1978
1979  - Made assorted updates to docs.
1980
1981  - Added MEM_BARRIER() to afl-showmap and afl-tmin, just to be safe.
1982
1983### Version 1.59b:
1984
1985  - Imported Laszlo Szekeres' experimental LLVM instrumentation into
1986    examples/llvm_instrumentation. I'll work on including it in the
1987    "mainstream" version soon.
1988
1989  - Fixed another typo, thanks to Jakub Wilk.
1990
1991### Version 1.58b:
1992
1993  - Added a workaround for abort() behavior in -lpthread programs in QEMU mode.
1994    Spotted by Aidan Thornton.
1995
1996  - Made several documentation updates, including links to the static
1997    instrumentation tool (sister_projects.txt).
1998
1999### Version 1.57b:
2000
2001  - Fixed a problem with exception handling on some versions of MacOS X.
2002    Spotted by Samir Aguiar and Anders Wang Kristensen.
2003
2004  - Tweaked afl-gcc to use BIN_PATH instead of a fixed string in help
2005    messages.
2006
2007### Version 1.56b:
2008
2009  - Renamed related_work.txt to historical_notes.txt.
2010
2011  - Made minor edits to the ASAN doc.
2012
2013  - Added docs/sister_projects.txt with a list of inspired or closely
2014    related utilities.
2015
2016### Version 1.55b:
2017
2018  - Fixed a glitch with afl-showmap opening /dev/null with O_RDONLY when
2019    running in quiet mode. Spotted by Tyler Nighswander.
2020
2021### Version 1.54b:
2022
2023  - Added another postprocessor example for PNG.
2024
2025  - Made a cosmetic fix to realloc() handling in examples/post_library/,
2026    suggested by Jakub Wilk.
2027
2028  - Improved -ldl handling. Suggested by Jakub Wilk.
2029
2030### Version 1.53b:
2031
2032  - Fixed an -l ordering issue that is apparently still a problem on Ubuntu.
2033    Spotted by William Robinet.
2034
2035### Version 1.52b:
2036
2037  - Added support for file format postprocessors. Requested by Ben Nagy. This
2038    feature is intentionally buried, since it's fairly easy to misuse and
2039    useful only in some scenarios. See examples/post_library/.
2040
2041### Version 1.51b:
2042
2043  - Made it possible to properly override LD_BIND_NOW after one very unusual
2044    report of trouble.
2045
2046  - Cleaned up typos, thanks to Jakub Wilk.
2047
2048  - Fixed a bug in AFL_DUMB_FORKSRV.
2049
2050### Version 1.50b:
2051
2052  - Fixed a flock() bug that would prevent dir reuse errors from kicking
2053    in every now and then.
2054
2055  - Renamed references to ppvm (the project is now called recidivm).
2056
2057  - Made improvements to file descriptor handling to avoid leaving some fds
2058    unnecessarily open in the child process.
2059
2060  - Fixed a typo or two.
2061
2062### Version 1.49b:
2063
2064  - Added code to save original command line in fuzzer_stats and
2065    crashes/README.txt. Also saves fuzzer version in fuzzer_stats.
2066    Requested by Ben Nagy.
2067
2068### Version 1.48b:
2069
2070  - Fixed a bug with QEMU fork server crashes when translation is attempted
2071    after a jump to an invalid pointer in the child process (i.e., after
2072    bumping into a particularly nasty security bug in the tested binary).
2073    Reported by Tyler Nighswander.
2074
2075### Version 1.47b:
2076
2077  - Fixed a bug with afl-cmin in -Q mode complaining about binary being not
2078    instrumented. Thanks to Jonathan Neuschafer for the bug report.
2079
2080  - Fixed another bug with argv handling for afl-fuzz in -Q mode. Reported
2081    by Jonathan Neuschafer.
2082
2083  - Improved the use of colors when showing crash counts in -C mode.
2084
2085### Version 1.46b:
2086
2087  - Improved instrumentation performance on 32-bit systems by getting rid of
2088    xor-swap (oddly enough, xor-swap is still faster on 64-bit) and tweaking
2089    alignment.
2090
2091  - Made path depth numbers more accurate with imported test cases.
2092
2093### Version 1.45b:
2094
2095  - Added support for SIMPLE_FILES in config.h for folks who don't like
2096    descriptive file names. Generates very simple names without colons,
2097    commas, plus signs, dashes, etc.
2098
2099  - Replaced zero-sized files with symlinks in the variable behavior state
2100    dir to simplify examining the relevant test cases.
2101
2102  - Changed the period of limited-range block ops from 5 to 10 minutes based
2103    on a couple of experiments. The basic goal of this delay timer behavior
2104    is to better support jobs that are seeded with completely invalid files,
2105    in which case, the first few queue cycles may be completed very quickly
2106    without discovering new paths. Should have no effect on well-seeded jobs.
2107
2108  - Made several minor updates to docs.
2109
2110### Version 1.44b:
2111
2112  - Corrected two bungled attempts to get the -C mode work properly
2113    with afl-cmin (accounting for the short-lived releases tagged 1.42 and
2114    1.43b) - sorry.
2115
2116  - Removed AFL_ALLOW_CRASHES in favor of the -C mode in said tool.
2117
2118  - Said goodbye to Hello Kitty, as requested by Padraig Brady.
2119
2120### Version 1.41b:
2121
2122  - Added AFL_ALLOW_CRASHES=1 to afl-cmin. Allows crashing inputs in the
2123    output corpus. Changed the default behavior to disallow it.
2124
2125  - Made the afl-cmin output dir default to 0700, not 0755, to be consistent
2126    with afl-fuzz; documented the rationale for 0755 in afl-plot.
2127
2128  - Lowered the output dir reuse time limit to 25 minutes as a dice-roll
2129    compromise after a discussion on afl-users@.
2130
2131  - Made afl-showmap accept -o /dev/null without borking out.
2132
2133  - Added support for crash / hang info in exit codes of afl-showmap.
2134
2135  - Tweaked block operation scaling to also factor in ballpark run time
2136    in cases where queue passes take very little time.
2137
2138  - Fixed typos and made improvements to several docs.
2139
2140### Version 1.40b:
2141
2142  - Switched to smaller block op sizes during the first passes over the
2143    queue. Helps keep test cases small.
2144
2145  - Added memory barrier for run_target(), just in case compilers get
2146    smarter than they are today.
2147
2148  - Updated a bunch of docs.
2149
2150### Version 1.39b:
2151
2152  - Added the ability to skip inputs by sending SIGUSR1 to the fuzzer.
2153
2154  - Reworked several portions of the documentation.
2155
2156  - Changed the code to reset splicing perf scores between runs to keep
2157    them closer to intended length.
2158
2159  - Reduced the minimum value of -t to 5 for afl-fuzz (~200 exec/sec)
2160    and to 10 for auxiliary tools (due to the absence of a fork server).
2161
2162  - Switched to more aggressive default timeouts (rounded up to 25 ms
2163    versus 50 ms - ~40 execs/sec) and made several other cosmetic changes
2164    to the timeout code.
2165
2166### Version 1.38b:
2167
2168  - Fixed a bug in the QEMU build script, spotted by William Robinet.
2169
2170  - Improved the reporting of skipped bitflips to keep the UI counters a bit
2171    more accurate.
2172
2173  - Cleaned up related_work.txt and added some non-goals.
2174
2175  - Fixed typos, thanks to Jakub Wilk.
2176
2177### Version 1.37b:
2178
2179  - Added effector maps, which detect regions that do not seem to respond
2180    to bitflips and subsequently exclude them from more expensive steps
2181    (arithmetics, known ints, etc). This should offer significant performance
2182    improvements with quite a few types of text-based formats, reducing the
2183    number of deterministic execs by a factor of 2 or so.
2184
2185  - Cleaned up mem limit handling in afl-cmin.
2186
2187  - Switched from uname -i to uname -m to work around Gentoo-specific
2188    issues with coreutils when building QEMU. Reported by William Robinet.
2189
2190  - Switched from PID checking to flock() to detect running sessions.
2191    Problem, against all odds, bumped into by Jakub Wilk.
2192
2193  - Added SKIP_COUNTS and changed the behavior of COVERAGE_ONLY in config.h.
2194    Useful only for internal benchmarking.
2195
2196  - Made improvements to UI refresh rates and exec/sec stats to make them
2197    more stable.
2198
2199  - Made assorted improvements to the documentation and to the QEMU build
2200    script.
2201
2202  - Switched from perror() to strerror() in error macros, thanks to Jakub
2203    Wilk for the nag.
2204
2205  - Moved afl-cmin back to bash, wasn't thinking straight. It has to stay
2206    on bash because other shells may have restrictive limits on array sizes.
2207
2208### Version 1.36b:
2209
2210  - Switched afl-cmin over to /bin/sh. Thanks to Jonathan Gray.
2211
2212  - Fixed an off-by-one bug in queue limit check when resuming sessions
2213    (could cause NULL ptr deref if you are *really* unlucky).
2214
2215  - Fixed the QEMU script to tolerate i686 if returned by uname -i. Based on
2216    a problem report from Sebastien Duquette.
2217
2218  - Added multiple references to Jakub's ppvm tool.
2219
2220  - Made several minor improvements to the Makefile.
2221
2222  - Believe it or not, fixed some typos. Thanks to Jakub Wilk.
2223
2224### Version 1.35b:
2225
2226  - Cleaned up regular expressions in some of the scripts to avoid errors
2227    on *BSD systems. Spotted by Jonathan Gray.
2228
2229### Version 1.34b:
2230
2231  - Performed a substantial documentation and program output cleanup to
2232    better explain the QEMU feature.
2233
2234### Version 1.33b:
2235
2236  - Added support for AFL_INST_RATIO and AFL_INST_LIBS in the QEMU mode.
2237
2238  - Fixed a stack allocation crash in QEMU mode (bug in QEMU, fixed with
2239    an extra patch applied to the downloaded release).
2240
2241  - Added code to test the QEMU instrumentation once the afl-qemu-trace
2242    binary is built.
2243
2244  - Modified afl-tmin and afl-showmap to search $PATH for binaries and to
2245    better handle QEMU support.
2246
2247  - Added a check for instrumented binaries when passing -Q to afl-fuzz.
2248
2249### Version 1.32b:
2250
2251  - Fixed 'make install' following the QEMU changes. Spotted by Hanno Boeck.
2252
2253  - Fixed EXTRA_PAR handling in afl-cmin.
2254
2255### Version 1.31b:
2256
2257  - Hallelujah! Thanks to Andrew Griffiths, we now support very fast, black-box
2258    instrumentation of binary-only code. See qemu_mode/README.qemu.
2259
2260    To use this feature, you need to follow the instructions in that
2261    directory and then run afl-fuzz with -Q.
2262
2263### Version 1.30b:
2264
2265  - Added -s (summary) option to afl-whatsup. Suggested by Jodie Cunningham.
2266
2267  - Added a sanity check in afl-tmin to detect minimization to zero len or
2268    excess hangs.
2269
2270  - Fixed alphabet size counter in afl-tmin.
2271
2272  - Slightly improved the handling of -B in afl-fuzz.
2273
2274  - Fixed process crash messages with -m none.
2275
2276### Version 1.29b:
2277
2278  - Improved the naming of test cases when orig: is already present in the file
2279    name.
2280
2281  - Made substantial improvements to technical_details.txt.
2282
2283### Version 1.28b:
2284
2285  - Made a minor tweak to the instrumentation to preserve the directionality
2286    of tuples (i.e., A -> B != B -> A) and to maintain the identity of tight
2287    loops (A -> A). You need to recompile targeted binaries to leverage this.
2288
2289  - Cleaned up some of the afl-whatsup stats.
2290
2291  - Added several sanity checks to afl-cmin.
2292
2293### Version 1.27b:
2294
2295  - Made afl-tmin recursive. Thanks to Hanno Boeck for the tip.
2296
2297  - Added docs/technical_details.txt.
2298
2299  - Changed afl-showmap search strategy in afl-cmap to just look into the
2300    same place that afl-cmin is executed from. Thanks to Jakub Wilk.
2301
2302  - Removed current_todo.txt and cleaned up the remaining docs.
2303
2304### Version 1.26b:
2305
2306  - Added total execs/sec stat for afl-whatsup.
2307
2308  - afl-cmin now auto-selects between cp or ln. Based on feedback from
2309    Even Huus.
2310
2311  - Fixed a typo. Thanks to Jakub Wilk.
2312
2313  - Made afl-gotcpu a bit more accurate by using getrusage instead of
2314    times. Thanks to Jakub Wilk.
2315
2316  - Fixed a memory limit issue during the build process on NetBSD-current.
2317    Reported by Thomas Klausner.
2318
2319### Version 1.25b:
2320
2321  - Introduced afl-whatsup, a simple tool for querying the status of
2322    local synced instances of afl-fuzz.
2323
2324  - Added -x compiler to clang options on Darwin. Suggested by Filipe
2325    Cabecinhas.
2326
2327  - Improved exit codes for afl-gotcpu.
2328
2329  - Improved the checks for -m and -t values in afl-cmin. Bug report
2330    from Evan Huus.
2331
2332### Version 1.24b:
2333
2334  - Introduced afl-getcpu, an experimental tool to empirically measure
2335    CPU preemption rates. Thanks to Jakub Wilk for the idea.
2336
2337### Version 1.23b:
2338
2339  - Reverted one change to afl-cmin that actually made it slower.
2340
2341### Version 1.22b:
2342
2343  - Reworked afl-showmap.c to support normal options, including -o, -q,
2344    -e. Also added support for timeouts and memory limits.
2345
2346  - Made changes to afl-cmin and other scripts to accommodate the new
2347    semantics.
2348
2349  - Officially retired AFL_EDGES_ONLY.
2350
2351  - Fixed another typo in afl-tmin, courtesy of Jakub Wilk.
2352
2353### Version 1.21b:
2354
2355  - Graduated minimize_corpus.sh to afl-cmin. It is now a first-class
2356    utility bundled with the fuzzer.
2357
2358  - Made significant improvements to afl-cmin to make it faster, more
2359    robust, and more versatile.
2360
2361  - Refactored some of afl-tmin code to make it a bit more readable.
2362
2363  - Made assorted changes to the doc to document afl-cmin and other stuff.
2364
2365### Version 1.20b:
2366
2367  - Added AFL_DUMB_FORKSRV, as requested by Jakub Wilk. This works only
2368    in -n mode and allows afl-fuzz to run with "dummy" fork servers that
2369    don't output any instrumentation, but follow the same protocol.
2370
2371  - Renamed AFL_SKIP_CHECKS to AFL_SKIP_BIN_CHECK to make it at least
2372    somewhat descriptive.
2373
2374  - Switched to using clang as the default assembler on MacOS X to work
2375    around Xcode issues with newer builds of clang. Testing and patch by
2376    Nico Weber.
2377
2378  - Fixed a typo (via Jakub Wilk).
2379
2380### Version 1.19b:
2381
2382  - Improved exec failure detection in afl-fuzz and afl-showmap.
2383
2384  - Improved Ctrl-C handling in afl-showmap.
2385
2386  - Added afl-tmin, a handy instrumentation-enabled minimizer.
2387
2388### Version 1.18b:
2389
2390  - Fixed a serious but short-lived bug in the resumption behavior introduced
2391    in version 1.16b.
2392
2393  - Added -t nn+ mode for soft-skipping timing-out paths.
2394
2395### Version 1.17b:
2396
2397  - Fixed a compiler warning introduced in 1.16b for newer versions of GCC.
2398    Thanks to Jakub Wilk and Ilfak Guilfanov.
2399
2400  - Improved the consistency of saving fuzzer_stats, bitmap info, and
2401    auto-dictionaries when aborting fuzzing sessions.
2402
2403  - Made several noticeable performance improvements to deterministic arith
2404    and known int steps.
2405
2406### Version 1.16b:
2407
2408  - Added a bit of code to make resumption pick up from the last known
2409    offset in the queue, rather than always rewinding to the start. Suggested
2410    by Jakub Wilk.
2411
2412  - Switched to tighter timeout control for slow programs (3x rather than
2413    5x average exec speed at init).
2414
2415### Version 1.15b:
2416
2417  - Added support for AFL_NO_VAR_CHECK to speed up resumption and inhibit
2418    variable path warnings for some programs.
2419
2420  - Made the trimmer run even for variable paths, since there is no special
2421    harm in doing so and it can be very beneficial if the trimming still
2422    pans out.
2423
2424  - Made the UI a bit more descriptive by adding "n/a" instead of "0" in a
2425    couple of corner cases.
2426
2427### Version 1.14b:
2428
2429  - Added a (partial) dictionary for JavaScript.
2430
2431  - Added AFL_NO_CPU_RED, as suggested by Jakub Wilk.
2432
2433  - Tweaked the havoc scaling logic added in 1.12b.
2434
2435### Version 1.13b:
2436
2437  - Improved the performance of minimize_corpus.sh by switching to a
2438    sort-based approach.
2439
2440  - Made several minor revisions to the docs.
2441
2442### Version 1.12b:
2443
2444  - Made an improvement to dictionary generation to avoid runs of identical
2445    bytes.
2446
2447  - Added havoc cycle scaling to help with slow binaries in -d mode. Based on
2448    a thread with Sami Liedes.
2449
2450  - Added AFL_SYNC_FIRST for afl-fuzz. This is useful for those who obsess
2451    over stats, no special purpose otherwise.
2452
2453  - Switched to more robust box drawing codes, suggested by Jakub Wilk.
2454
2455  - Created faster 64-bit variants of several critical-path bitmap functions
2456    (sorry, no difference on 32 bits).
2457
2458  - Fixed moar typos, as reported by Jakub Wilk.
2459
2460### Version 1.11b:
2461
2462  - Added a bit more info about dictionary strategies to the status screen.
2463
2464### Version 1.10b:
2465
2466  - Revised the dictionary behavior to use insertion and overwrite in
2467    deterministic steps, rather than just the latter. This improves coverage
2468    with SQL and the like.
2469
2470  - Added a mention of "*" in status_screen.txt, as suggested by Jakub Wilk.
2471
2472### Version 1.09b:
2473
2474  - Corrected a cosmetic problem with 'extras' stage count not always being
2475    accurate in the stage yields view.
2476
2477  - Fixed a typo reported by Jakub Wilk and made some minor documentation
2478    improvements.
2479
2480### Version 1.08b:
2481
2482  - Fixed a div-by-zero bug in the newly-added code when using a dictionary.
2483
2484### Version 1.07b:
2485
2486  - Added code that automatically finds and extracts syntax tokens from the
2487    input corpus.
2488
2489  - Fixed a problem with ld dead-code removal option on MacOS X, reported
2490    by Filipe Cabecinhas.
2491
2492  - Corrected minor typos spotted by Jakub Wilk.
2493
2494  - Added a couple of more exotic archive format samples.
2495
2496### Version 1.06b:
2497
2498  - Switched to slightly more accurate (if still not very helpful) reporting
2499    of short read and short write errors. These theoretically shouldn't happen
2500    unless you kill the forkserver or run out of disk space. Suggested by
2501    Jakub Wilk.
2502
2503  - Revamped some of the allocator and debug code, adding comments and
2504    cleaning up other mess.
2505
2506  - Tweaked the odds of fuzzing non-favored test cases to make sure that
2507    baseline coverage of all inputs is reached sooner.
2508
2509### Version 1.05b:
2510
2511  - Added a dictionary for WebP.
2512
2513  - Made some additional performance improvements to minimize_corpus.sh,
2514    getting deeper into the bash woods.
2515
2516### Version 1.04b:
2517
2518  - Made substantial performance improvements to minimize_corpus.sh with
2519    large datasets, albeit at the expense of having to switch back to bash
2520    (other shells may have limits on array sizes, etc).
2521
2522  - Tweaked afl-showmap to support the format used by the new script.
2523
2524### Version 1.03b:
2525
2526  - Added code to skip README.txt in the input directory to make the crash
2527    exploration mode work better. Suggested by Jakub Wilk.
2528
2529  - Added a dictionary for SQLite.
2530
2531### Version 1.02b:
2532
2533  - Reverted the ./ search path in minimize_corpus.sh because people did
2534    not like it.
2535
2536  - Added very explicit warnings not to run various shell scripts that
2537    read or write to /tmp/ (since this is generally a pretty bad idea on
2538    multi-user systems).
2539
2540  - Added a check for /tmp binaries and -f locations in afl-fuzz.
2541
2542### Version 1.01b:
2543
2544  - Added dictionaries for XML and GIF.
2545
2546### Version 1.00b:
2547
2548  - Slightly improved the performance of minimize_corpus.sh, especially on
2549    Linux.
2550
2551  - Made a couple of improvements to calibration timeouts for resumed scans.
2552
2553### Version 0.99b:
2554
2555  - Fixed minimize_corpus.sh to work with dash, as suggested by Jakub Wilk.
2556
2557  - Modified minimize_corpus.sh to try locate afl-showmap in $PATH and ./.
2558    The first part requested by Jakub Wilk.
2559
2560  - Added support for afl-as --version, as required by one funky build
2561    script. Reported by William Robinet.
2562
2563### Version 0.98b:
2564
2565  - Added a dictionary for TIFF.
2566
2567  - Fixed another cosmetic snafu with stage exec counts for -x.
2568
2569  - Switched afl-plot to /bin/sh, since it seems bashism-free. Also tried
2570    to remove any obvious bashisms from other examples/ scripts,
2571    most notably including minimize_corpus.sh and triage_crashes.sh.
2572    Requested by Jonathan Gray.
2573
2574### Version 0.97b:
2575
2576  - Fixed cosmetic issues around the naming of -x strategy files.
2577
2578  - Added a dictionary for JPEG.
2579
2580  - Fixed a very rare glitch when running instrumenting 64-bit code that makes
2581    heavy use of xmm registers that are also touched by glibc.
2582
2583### Version 0.96b:
2584
2585  - Added support for extra dictionaries, provided testcases/_extras/png/
2586    as a demo.
2587
2588  - Fixed a minor bug in number formatting routines used by the UI.
2589
2590  - Added several additional PNG test cases that are relatively unlikely
2591    to be hit by chance.
2592
2593  - Fixed afl-plot syntax for gnuplot 5.x. Reported by David Necas.
2594
2595### Version 0.95b:
2596
2597  - Cleaned up the OSX ReportCrash code. Thanks to Tobias Ospelt for help.
2598
2599  - Added some extra tips for AFL_NO_FORKSERVER on OSX.
2600
2601  - Refreshed the INSTALL file.
2602
2603### Version 0.94b:
2604
2605  - Added in-place resume (-i-) to address a common user complaint.
2606
2607  - Added an awful workaround for ReportCrash on MacOS X. Problem
2608    spotted by Joseph Gentle.
2609
2610### Version 0.93b:
2611
2612  - Fixed the link() workaround, as reported by Jakub Wilk.
2613
2614### Version 0.92b:
2615
2616  - Added support for reading test cases from another filesystem.
2617    Requested by Jakub Wilk.
2618
2619  - Added pointers to the mailing list.
2620
2621  - Added a sample PDF document.
2622
2623### Version 0.91b:
2624
2625  - Refactored minimize_corpus.sh to make it a bit more user-friendly and to
2626    select for smallest files, not largest bitmaps. Offers a modest corpus
2627    size improvement in most cases.
2628
2629  - Slightly improved the performance of splicing code.
2630
2631### Version 0.90b:
2632
2633  - Moved to an algorithm where paths are marked as preferred primarily based
2634    on size and speed, rather than bitmap coverage. This should offer
2635    noticeable performance gains in many use cases.
2636
2637  - Refactored path calibration code; calibration now takes place as soon as a
2638    test case is discovered, to facilitate better prioritization decisions later
2639    on.
2640
2641  - Changed the way of marking variable paths to avoid .state metadata
2642    inconsistencies.
2643
2644  - Made sure that calibration routines always create a new test case to avoid
2645    hypothetical problems with utilities that modify the input file.
2646
2647  - Added bitmap saturation to fuzzer stats and plot data.
2648
2649  - Added a testcase for JPEG XR.
2650
2651  - Added a tty check for the colors warning in Makefile, to keep distro build
2652    logs tidy. Suggested by Jakub Wilk.
2653
2654### Version 0.89b:
2655
2656  - Renamed afl-plot.sh to afl-plot, as requested by Padraig Brady.
2657
2658  - Improved the compatibility of afl-plot with older versions of gnuplot.
2659
2660  - Added banner information to fuzzer_stats, populated it to afl-plot.
2661
2662### Version 0.88b:
2663
2664  - Added support for plotting, with design and implementation based on a
2665    prototype design proposed by Michael Rash. Huge thanks!
2666
2667  - Added afl-plot.sh, which allows you to, well, generate a nice plot using
2668    this data.
2669
2670  - Refactored the code slightly to make more frequent updates to fuzzer_stats
2671    and to provide more detail about synchronization.
2672
2673  - Added an fflush(stdout) call for non-tty operation, as requested by
2674    Joonas Kuorilehto.
2675
2676  - Added some detail to fuzzer_stats for parity with plot_file.
2677
2678### Version 0.87b:
2679
2680  - Added support for MSAN, via AFL_USE_MSAN, same gotchas as for ASAN.
2681
2682### Version 0.86b:
2683
2684  - Added AFL_NO_FORKSRV, allowing the forkserver to be bypassed. Suggested
2685    by Ryan Govostes.
2686
2687  - Simplified afl-showmap.c to make use of the no-forkserver mode.
2688
2689  - Made minor improvements to crash_triage.sh, as suggested by Jakub Wilk.
2690
2691### Version 0.85b:
2692
2693  - Fixed the CPU counting code - no sysctlbyname() on OpenBSD, d'oh. Bug
2694    reported by Daniel Dickman.
2695
2696  - Made a slight correction to error messages - the advice on testing
2697    with ulimit was a tiny bit off by a factor of 1024.
2698
2699### Version 0.84b:
2700
2701  - Added support for the CPU widget on some non-Linux platforms (I hope).
2702    Based on feedback from Ryan Govostes.
2703
2704  - Cleaned up the changelog (very meta).
2705
2706### Version 0.83b:
2707
2708  - Added examples/clang_asm_normalize/ and related notes in
2709    env_variables.txt and afl-as.c. Thanks to Ryan Govostes for the idea.
2710
2711  - Added advice on hardware utilization in README.
2712
2713### Version 0.82b:
2714
2715  - Made additional fixes for Xcode support, juggling -Q and -q flags. Thanks to
2716    Ryan Govostes.
2717
2718  - Added a check for __asm__ blocks and switches to .intel_syntax in assembly.
2719    Based on feedback from Ryan Govostes.
2720
2721### Version 0.81b:
2722
2723  - A workaround for Xcode 6 as -Q flag glitch. Spotted by Ryan Govostes.
2724
2725  - Improved Solaris build instructions, as suggested by Martin Carpenter.
2726
2727  - Fix for a slightly busted path scoring conditional. Minor practical impact.
2728
2729### Version 0.80b:
2730
2731  - Added a check for $PATH-induced loops. Problem noticed by Kartik Agaram.
2732
2733  - Added AFL_KEEP_ASSEMBLY for easier troubleshooting.
2734
2735  - Added an override for AFL_USE_ASAN if set at AFL compile time. Requested by
2736    Hanno Boeck.
2737
2738### Version 0.79b:
2739
2740  - Made minor adjustments to path skipping logic.
2741
2742  - Made several documentation updates to reflect the path selection changes
2743    made in 0.78b.
2744
2745### Version 0.78b:
2746
2747  - Added a CPU governor check. Bug report from Joe Zbiciak.
2748
2749  - Favored paths are now selected strictly based on new edges, not hit
2750    counts. This speeds up the first pass by a factor of 3-6x without
2751    significantly impacting ultimate coverage (tested with libgif, libpng,
2752    libjpeg).
2753
2754    It also allows some performance & memory usage improvements by making
2755    some of the in-memory bitmaps much smaller.
2756
2757  - Made multiple significant performance improvements to bitmap checking
2758    functions, plus switched to a faster hash.
2759
2760  - Owing largely to these optimizations, bumped the size of the bitmap to
2761    64k and added a warning to detect older binaries that rely on smaller
2762    bitmaps.
2763
2764### Version 0.77b:
2765
2766  - Added AFL_SKIP_CHECKS to bypass binary checks when really warranted.
2767    Feature requested by Jakub Wilk.
2768
2769  - Fixed a couple of typos.
2770
2771  - Added a warning for runs that are aborted early on.
2772
2773### Version 0.76b:
2774
2775  - Incorporated another signal handling fix for Solaris. Suggestion
2776    submitted by Martin Carpenter.
2777
2778### Version 0.75b:
2779
2780  - Implemented a slightly more "elegant" kludge for the %llu glitch (see
2781    types.h).
2782
2783  - Relaxed CPU load warnings to stay in sync with reality.
2784
2785### Version 0.74b:
2786
2787  - Switched to more responsive exec speed averages and better UI speed
2788    scaling.
2789
2790  - Fixed a bug with interrupted reads on Solaris. Issue spotted by Martin
2791    Carpenter.
2792
2793### Version 0.73b:
2794
2795  - Fixed a stray memcpy() instead of memmove() on overlapping buffers.
2796    Mostly harmless but still dumb. Mistake spotted thanks to David Higgs.
2797
2798### Version 0.72b:
2799
2800  - Bumped map size up to 32k. You may want to recompile instrumented
2801    binaries (but nothing horrible will happen if you don't).
2802
2803  - Made huge performance improvements for bit-counting functions.
2804
2805  - Default optimizations now include -funroll-loops. This should have
2806    interesting effects on the instrumentation. Frankly, I'm just going to
2807    ship it and see what happens next. I have a good feeling about this.
2808
2809  - Made a fix for stack alignment crash on MacOS X 10.10; looks like the
2810    rhetorical question in the comments in afl-as.h has been answered.
2811    Tracked down by Mudge Zatko.
2812
2813### Version 0.71b:
2814
2815  - Added a fix for the nonsensical MacOS ELF check. Spotted by Mudge Zatko.
2816
2817  - Made some improvements to ASAN checks.
2818
2819### Version 0.70b:
2820
2821  - Added explicit detection of ASANified binaries.
2822
2823  - Fixed compilation issues on Solaris. Reported by Martin Carpenter.
2824
2825### Version 0.69b:
2826
2827  - Improved the detection of non-instrumented binaries.
2828
2829  - Made the crash counter in -C mode accurate.
2830
2831  - Fixed an obscure install bug that made afl-as non-functional with the tool
2832    installed to /usr/bin instead of /usr/local/bin. Found by Florian Kiersch.
2833
2834  - Fixed for a cosmetic SIGFPE when Ctrl-C is pressed while the fork server
2835    is spinning up.
2836
2837### Version 0.68b:
2838
2839  - Added crash exploration mode! Woot!
2840
2841### Version 0.67b:
2842
2843  - Fixed several more typos, the project is now cartified 100% typo-free.
2844    Thanks to Thomas Jarosch and Jakub Wilk.
2845
2846  - Made a change to write fuzzer_stats early on.
2847
2848  - Fixed a glitch when (not!) running on MacOS X as root. Spotted by Tobias
2849    Ospelt.
2850
2851  - Made it possible to override -O3 in Makefile. Suggested by Jakub Wilk.
2852
2853### Version 0.66b:
2854
2855  - Fixed a very obscure issue with build systems that use gcc as an assembler
2856    for hand-written .s files; this would confuse afl-as. Affected nss, reported
2857    by Hanno Boeck.
2858
2859  - Fixed a bug when cleaning up synchronized fuzzer output dirs. Issue reported
2860    by Thomas Jarosch.
2861
2862### Version 0.65b:
2863
2864  - Cleaned up shell printf escape codes in Makefile. Reported by Jakub Wilk.
2865
2866  - Added more color to fuzzer_stats, provided short documentation of the file
2867    format, and made several other stats-related improvements.
2868
2869### Version 0.64b:
2870
2871  - Enabled GCC support on MacOS X.
2872
2873### Version 0.63b:
2874
2875  - Provided a new, simplified way to pass data in files (@@). See README.
2876
2877  - Made additional fixes for 64-bit MacOS X, working around a crashing bug in
2878    their linker (umpf) and several other things. It's alive!
2879
2880  - Added a minor workaround for a bug in 64-bit FreeBSD (clang -m32 -g doesn't
2881    work on that platform, but clang -m32 does, so we no longer insert -g).
2882
2883  - Added a build-time warning for inverse video terminals and better
2884    instructions in status_screen.txt.
2885
2886### Version 0.62b:
2887
2888  - Made minor improvements to the allocator, as suggested by Tobias Ospelt.
2889
2890  - Added example instrumented memcmp() in examples/instrumented_cmp.
2891
2892  - Added a speculative fix for MacOS X (clang detection, again).
2893
2894  - Fixed typos in parallel_fuzzing.txt. Problems spotted by Thomas Jarosch.
2895
2896### Version 0.61b:
2897
2898  - Fixed a minor issue with clang detection on systems with a clang cc
2899    wrapper, so that afl-gcc doesn't confuse it with GCC.
2900
2901  - Made cosmetic improvements to docs and to the CPU load indicator.
2902
2903  - Fixed a glitch with crash removal (README.txt left behind, d'oh).
2904
2905### Version 0.60b:
2906
2907  - Fixed problems with jump tables generated by exotic versions of GCC. This
2908    solves an outstanding problem on OpenBSD when using afl-gcc + PIE (not
2909    present with afl-clang).
2910
2911  - Fixed permissions on one of the sample archives.
2912
2913  - Added a lahf / sahf workaround for OpenBSD (their assembler doesn't know
2914    about these opcodes).
2915
2916  - Added docs/INSTALL.
2917
2918### Version 0.59b:
2919
2920  - Modified 'make install' to also install test cases.
2921
2922  - Provided better pointers to installed README in afl-fuzz.
2923
2924  - More work on RLIMIT_AS for OpenBSD.
2925
2926### Version 0.58b:
2927
2928  - Added a core count check on Linux.
2929
2930  - Refined the code for the lack-of-RLIMIT_AS case on OpenBSD.
2931
2932  - Added a rudimentary CPU utilization meter to help with optimal loading.
2933
2934### Version 0.57b:
2935
2936  - Made fixes to support FreeBSD and OpenBSD: use_64bit is now inferred if not
2937    explicitly specified when calling afl-as, and RLIMIT_AS is behind an #ifdef.
2938    Thanks to Fabian Keil and Jonathan Gray for helping troubleshoot this.
2939
2940  - Modified 'make install' to also install docs (in /usr/local/share/doc/afl).
2941
2942  - Fixed a typo in status_screen.txt.
2943
2944  - Made a couple of Makefile improvements as proposed by Jakub Wilk.
2945
2946### Version 0.56b:
2947
2948  - Added probabilistic instrumentation density reduction in ASAN mode. This
2949    compensates for ASAN-specific branches in a crude but workable way.
2950
2951  - Updated notes_for_asan.txt.
2952
2953### Version 0.55b:
2954
2955  - Implemented smarter out_dir behavior, automatically deleting directories
2956    that don't contain anything of special value. Requested by several folks,
2957    including Hanno Boeck.
2958
2959  - Added more detail in fuzzer_stats (start time, run time, fuzzer PID).
2960
2961  - Implemented support for configurable install prefixes in Makefile
2962    ($PREFIX), as requested by Luca Barbato.
2963
2964  - Made it possible to resume by doing -i <out_dir>, without having to specify
2965    -i <out_dir>/queue/.
2966
2967### Version 0.54b:
2968
2969  - Added a fix for -Wformat warning messages (oops, I thought this had been in
2970    place for a while).
2971
2972### Version 0.53b:
2973
2974  - Redesigned the crash & hang duplicate detection code to better deal with
2975    fault conditions that can be reached in a multitude of ways.
2976
2977    The old approach could be compared to hashing stack traces to de-dupe
2978    crashes, a method prone to crash count inflation. The alternative I
2979    wanted to avoid would be equivalent to just looking at crash %eip,
2980    which can have false negatives in common functions such as memcpy().
2981
2982    The middle ground currently used in afl-fuzz can be compared to looking
2983    at every line item in the stack trace and tagging crashes as unique if
2984    we see any function name that we haven't seen before (or if something that
2985    we have *always* seen there suddenly disappears). We do the comparison
2986    without paying any attention to ordering or hit counts. This can still
2987    cause some crash inflation early on, but the problem will quickly taper
2988    off. So, you may get 20 dupes instead of 5,000.
2989
2990  - Added a fix for harmless but absurd trim ratios shown if the first exec in
2991    the trimmer timed out. Spotted by @EspenGx.
2992
2993### Version 0.52b:
2994
2995  - Added a quick summary of the contents in examples/.
2996
2997  - Made a fix to the process of writing fuzzer_stats.
2998
2999  - Slightly reorganized the .state/ directory, now recording redundant paths,
3000    too. Note that this breaks the ability to properly resume older sessions
3001    - sorry about that.
3002
3003    (To fix this, simply move <out_dir>/.state/* from an older run
3004    to <out_dir>/.state/deterministic_done/*.)
3005
3006### Version 0.51b:
3007
3008  - Changed the search order for afl-as to avoid the problem with older copies
3009    installed system-wide; this also means that I can remove the Makefile check
3010    for that.
3011
3012  - Made it possible to set instrumentation ratio of 0%.
3013
3014  - Introduced some typos, fixed others.
3015
3016  - Fixed the test_prev target in Makefile, as reported by Ozzy Johnson.
3017
3018### Version 0.50b:
3019
3020  - Improved the 'make install' logic, as suggested by Padraig Brady.
3021
3022  - Revamped various bits of the documentation, especially around perf_tips.txt;
3023    based on the feedback from Alexander Cherepanov.
3024
3025  - Added AFL_INST_RATIO to afl-as. The only case where this comes handy is
3026    ffmpeg, at least as far as I can tell. (Trivia: the current version of
3027    ffmpeg ./configure also ignores CC and --cc, probably unintentionally).
3028
3029  - Added documentation for all environmental variables (env_variables.txt).
3030
3031  - Implemented a visual warning for excessive or insufficient bitmap density.
3032
3033  - Changed afl-gcc to add -O3 by default; use AFL_DONT_OPTIMIZE if you don't
3034    like that. Big speed gain for ffmpeg, so seems like a good idea.
3035
3036  - Made a regression fix to afl-as to ignore .LBB labels in gcc mode.
3037
3038### Version 0.49b:
3039
3040  - Fixed more typos, as found by Jakub Wilk.
3041
3042  - Added support for clang!
3043
3044  - Changed AFL_HARDEN to *not* include ASAN by default. Use AFL_USE_ASAN if
3045    needed. The reasons for this are in notes_for_asan.txt.
3046
3047  - Switched from configure auto-detection to isatty() to keep afl-as and
3048    afl-gcc quiet.
3049
3050  - Improved installation process to properly create symlinks, rather than
3051    copies of binaries.
3052
3053### Version 0.48b:
3054
3055  - Improved afl-fuzz to force-set ASAN_OPTIONS=abort_on_error=1. Otherwise,
3056    ASAN crashes wouldn't be caught at all. Reported by Hanno Boeck.
3057
3058  - Improved Makefile mkdir logic, as suggested by Hanno Boeck.
3059
3060  - Improved the 64-bit instrumentation to properly save r8-r11 registers in
3061    the x86 setup code. The old behavior could cause rare problems running
3062    *without* instrumentation when the first function called in a particular
3063    .o file has 5+ parameters. No impact on code running under afl-fuzz or
3064    afl-showmap. Issue spotted by Padraig Brady.
3065
3066### Version 0.47b:
3067
3068  - Fixed another Makefile bug for parallel builds of afl. Problem identified
3069    by Richard W. M. Jones.
3070
3071  - Added support for suffixes for -m.
3072
3073  - Updated the documentation and added notes_for_asan.txt. Based on feedback
3074    from Hanno Boeck, Ben Laurie, and others.
3075
3076  - Moved the project to https://lcamtuf.coredump.cx/afl/.
3077
3078### Version 0.46b:
3079
3080  - Cleaned up Makefile dependencies for parallel builds. Requested by
3081    Richard W. M. Jones.
3082
3083  - Added support for DESTDIR in Makefile. Once again suggested by
3084    Richard W. M. Jones :-)
3085
3086  - Removed all the USE_64BIT stuff; we now just auto-detect compilation mode.
3087    As requested by many callers to the show.
3088
3089  - Fixed rare problems with programs that use snippets of assembly and
3090    switch between .code32 and .code64. Addresses a glitch spotted by
3091    Hanno Boeck with compiling ToT gdb.
3092
3093### Version 0.45b:
3094
3095  - Implemented a test case trimmer. Results in 20-30% size reduction for many
3096    types of work loads, with very pronounced improvements in path discovery
3097    speeds.
3098
3099  - Added better warnings for various problems with input directories.
3100
3101  - Added a Makefile warning for older copies, based on counterintuitive
3102    behavior observed by Hovik Manucharyan.
3103
3104  - Added fuzzer_stats file for status monitoring. Suggested by @dronesec.
3105
3106  - Fixed moar typos, thanks to Alexander Cherepanov.
3107
3108  - Implemented better warnings for ASAN memory requirements, based on calls
3109    from several angry listeners.
3110
3111  - Switched to saner behavior with non-tty stdout (less output generated,
3112    no ANSI art).
3113
3114### Version 0.44b:
3115
3116  - Added support for AFL_CC and AFL_CXX, based on a patch from Ben Laurie.
3117
3118  - Replaced afl-fuzz -S -D with -M for simplicity.
3119
3120  - Added a check for .section .text; lack of this prevented main() from
3121    getting instrumented for some users. Reported by Tom Ritter.
3122
3123  - Reorganized the testcases/ directory.
3124
3125  - Added an extra check to confirm that the build is operational.
3126
3127  - Made more consistent use of color reset codes, as suggested by Oliver
3128    Kunz.
3129
3130### Version 0.43b:
3131
3132  - Fixed a bug with 64-bit gcc -shared relocs.
3133
3134  - Removed echo -e from Makefile for compatibility with dash. Suggested
3135    by Jakub Wilk.
3136
3137  - Added status_screen.txt.
3138
3139  - Added examples/canvas_harness.
3140
3141  - Made a minor change to the Makefile GCC check. Suggested by Hanno Boeck.
3142
3143### Version 0.42b:
3144
3145  - Fixed a bug with red zone handling for 64-bit (oops!). Problem reported by
3146    Felix Groebert.
3147
3148  - Implemented horribly experimental ARM support in examples/arm_support.
3149
3150  - Made several improvements to error messages.
3151
3152  - Added AFL_QUIET to silence afl-gcc and afl-as when using wonky build
3153    systems. Reported by Hanno Boeck.
3154
3155  - Improved check for 64-bit compilation, plus several sanity checks
3156    in Makefile.
3157
3158### Version 0.41b:
3159
3160  - Fixed a fork served bug for processes that call execve().
3161
3162  - Made minor compatibility fixes to Makefile, afl-gcc; suggested by Jakub
3163    Wilk.
3164
3165  - Fixed triage_crashes.sh to work with the new layout of output directories.
3166    Suggested by Jakub Wilk.
3167
3168  - Made multiple performance-related improvements to the injected
3169    instrumentation.
3170
3171  - Added visual indication of the number of imported paths.
3172
3173  - Fixed afl-showmap to make it work well with new instrumentation.
3174
3175  - Added much better error messages for crashes when importing test cases
3176    or otherwise calibrating the binary.
3177
3178### Version 0.40b:
3179
3180  - Added support for parallelized fuzzing. Inspired by earlier patch
3181    from Sebastian Roschke.
3182
3183  - Added an example in examples/distributed_fuzzing/.
3184
3185### Version 0.39b:
3186
3187  - Redesigned status screen, now 90% more spiffy.
3188
3189  - Added more verbose and user-friendly messages for some common problems.
3190
3191  - Modified the resumption code to reconstruct path depth.
3192
3193  - Changed the code to inhibit core dumps and improve the ability to detect
3194    SEGVs.
3195
3196  - Added a check for redirection of core dumps to programs.
3197
3198  - Made a minor improvement to the handling of variable paths.
3199
3200  - Made additional performance tweaks to afl-fuzz, chiefly around mem limits.
3201
3202  - Added performance_tips.txt.
3203
3204### Version 0.38b:
3205
3206  - Fixed an fd leak and +cov tracking bug resulting from changes in 0.37b.
3207
3208  - Implemented auto-scaling for screen update speed.
3209
3210  - Added a visual indication when running in non-instrumented mode.
3211
3212### Version 0.37b:
3213
3214  - Added fuzz state tracking for more seamless resumption of aborted
3215    fuzzing sessions.
3216
3217  - Removed the -D option, as it's no longer necessary.
3218
3219  - Refactored calibration code and improved startup reporting.
3220
3221  - Implemented dynamically scaled timeouts, so that you don't need to
3222    play with -t except in some very rare cases.
3223
3224  - Added visual notification for slow binaries.
3225
3226  - Improved instrumentation to explicitly cover the other leg of every
3227    branch.
3228
3229### Version 0.36b:
3230
3231  - Implemented fork server support to avoid the overhead of execve(). A
3232    nearly-verbatim design from Jann Horn; still pending part 2 that would
3233    also skip initial setup steps (thinking about reliable heuristics now).
3234
3235  - Added a check for shell scripts used as fuzz targets.
3236
3237  - Added a check for fuzz jobs that don't seem to be finding anything.
3238
3239  - Fixed the way IGNORE_FINDS works (was a bit broken after adding splicing
3240    and path skip heuristics).
3241
3242### Version 0.35b:
3243
3244  - Properly integrated 64-bit instrumentation into afl-as.
3245
3246### Version 0.34b:
3247
3248  - Added a new exec count classifier (the working theory is that it gets
3249    meaningful coverage with fewer test cases spewed out).
3250
3251### Version 0.33b:
3252
3253  - Switched to new, somewhat experimental instrumentation that tries to
3254    target only arcs, rather than every line. May be fragile, but is a lot
3255    faster (2x+).
3256
3257  - Made several other cosmetic fixes and typo corrections, thanks to
3258    Jakub Wilk.
3259
3260### Version 0.32b:
3261
3262  - Another take at fixing the C++ exception thing. Reported by Jakub Wilk.
3263
3264### Version 0.31b:
3265
3266  - Made another fix to afl-as to address a potential problem with newer
3267    versions of GCC (introduced in 0.28b). Thanks to Jann Horn.
3268
3269### Version 0.30b:
3270
3271  - Added more detail about the underlying operations in file names.
3272
3273### Version 0.29b:
3274
3275  - Made some general improvements to chunk operations.
3276
3277### Version 0.28b:
3278
3279  - Fixed C++ exception handling in newer versions of GCC. Problem diagnosed
3280    by Eberhard Mattes.
3281
3282  - Fixed the handling of the overflow flag. Once again, thanks to
3283    Eberhard Mattes.
3284
3285### Version 0.27b:
3286
3287  - Added prioritization of new paths over the already-fuzzed ones.
3288
3289  - Included spliced test case ID in the output file name.
3290
3291  - Fixed a rare, cosmetic null ptr deref after Ctrl-C.
3292
3293  - Refactored the code to make copies of test cases in the output directory.
3294
3295  - Switched to better output file names, keeping track of stage and splicing
3296    sources.
3297
3298### Version 0.26b:
3299
3300  - Revamped storage of testcases, -u option removed,
3301
3302  - Added a built-in effort minimizer to get rid of potentially redundant
3303    inputs,
3304
3305  - Provided a testcase count minimization script in examples/,
3306
3307  - Made miscellaneous improvements to directory and file handling.
3308
3309  - Fixed a bug in timeout detection.
3310
3311### Version 0.25b:
3312
3313  - Improved count-based instrumentation.
3314
3315  - Improved the hang deduplication logic.
3316
3317  - Added -cov prefixes for test cases.
3318
3319  - Switched from readdir() to scandir() + alphasort() to preserve ordering of
3320    test cases.
3321
3322  - Added a splicing strategy.
3323
3324  - Made various minor UI improvements and several other bugfixes.
3325
3326### Version 0.24b:
3327
3328  - Added program name to the status screen, plus the -T parameter to go with
3329    it.
3330
3331### Version 0.23b:
3332
3333  - Improved the detection of variable behaviors.
3334
3335  - Added path depth tracking,
3336
3337  - Improved the UI a bit,
3338
3339  - Switched to simplified (XOR-based) tuple instrumentation.
3340
3341### Version 0.22b:
3342
3343  - Refactored the handling of long bitflips and some swaps.
3344
3345  - Fixed the handling of gcc -pipe, thanks to anonymous reporter.
3346
3347### Version 0.21b (2013-11-12):
3348
3349  - Initial public release.
3350
3351  - Added support for use of multiple custom mutators which can be specified using
3352    the environment variable AFL_CUSTOM_MUTATOR_LIBRARY.
3353