* Revert "Insert into corpus if feedback is_interesting on crash/timeout (#1327)"
This reverts commit 871dfa0a013f31f84e43e125105febca2f137049.
* unused
* Fix issue with libafl_qemu being repeatedly rebuilt
* Changes to make qemu_launcher a production ready fuzzer
* Remove _get prefix
* Don't collect DrCov data during the campaign
* Fix poor performance
* Better validation for core selection
* Changes to print debug when running in verbose mode
* Autofix
* Remove afl++-clang
* Fix build error on 32-bit
* Fix some clippy
* Fix OSX
* Set default version of clang/clang++
* Review changes
* Fix issue with fd sharing between processes
---------
Co-authored-by: Your Name <you@example.com>
Co-authored-by: Andrea Fioraldi <andreafioraldi@gmail.com>
* endif
* just fix every profile
* WHY YOU DONT JUST USE THE SAME NAME FOR DIR
* fix
* Some fuzzer want release
* WHY I ALWAYS FORGET COMMA
* NO MORE SPACE
* rename
* stb doesn't like debug build
* just use release
* another just use release
* impr(frida): Don't keep FuzzerOptions in Helper
Instead, keep the actual values that are needed. This allows us to make
a builder for FridaInstrumentationBuilder in a subsequent commit.
* refactor(frida): Move workaround to separate method
This is just code movement.
* refactor(frida): move transformer initialization
Mostly code movement here, sets up replacing `new` with a builder. The
one exception is the introduction of a lifetime bound on RT, which needs
to outlive the transformer. This could be generic, but there's probably
no reason to introduce an additional lifetime.
However, because of this lifetime introduction, this is _technically_ a
breaking change.
* impr(frida): Pass module map to runtimes
Instead of passing a slice of modules to instrument, and re-building the
modulemap, pass a Ref-counted module map directly to the initialization.
* feat(frida): Builder for InstrumentationHelper
Co-authored-by: Dominik Maier <domenukk@gmail.com>
* impr(frida/alloc): optional options in allocator
Move all the initialization into Default::default with sensible defaults
and override parameters set from options in new.
* impr(frida): remove options from AsanError
The only option AsanError uses is whether to continue on error. Instead
of keeping a whole clone of the options around, just store that single
boolean value.
* impr(frida/asan): Use less FuzzerOptions
* Implement Default::default to get a good default AsanRuntime
---------
Co-authored-by: Dominik Maier <domenukk@gmail.com>
* update tuneable: consistently access metadata + force 'choice'
* oops, loop in the wrong place
* clarify API some; allow for least of set configuration
