These are now processed in afl-qemu-trace so that the "copy+paste" code
that is in all of the other AFL tools can be removed.
This also allows the AFL_USE_QASAN flag to work the same when used
with tools like afl-fuzz as it does with afl-qemu-trace. This is
important in situations where loading the QASAN library changes
the address of your desired entrypoint, or for crash validation
using the same environment that afl-fuzz was using.
With this change, the same set of environment variables can be used
in exactly the same way between afl-fuzz, afl-showmap, and
afl-qemu-trace, and you will get exactly the same guest environment.
In QEMU mode (-Q), setting AFL_QEMU_CUSTOM_BIN cause afl-fuzz to skip
prepending afl-qemu-trace to your command line. Use this if you wish to use a
custom afl-qemu-trace or if you need to modify the afl-qemu-trace arguments.
The previous implementation of "@@ handling"
in afl-showmap differed greatly from how it was
handled in afl-fuzz and how the documentation presented it.
It used to require that the @@ be its own argument separated
by whitespace and could not be used in situations like "--file=@@".
This change standardizes it to use detect_file_args() like
everybody else does, so that it will have the expected and
documented behavior.
The previous implementation of "@@ handling"
in afl-cmin differed greatly from how it was
handled in afl-fuzz and how the documentation presented it.
It used to require that the @@ be its own argument separated
by whitespace and could not be used in situations like "--file=@@".
This change standardizes it to just look for @@ to be *in* an argument
in the same manner that afl-cmin.bash does, so that it will have
the expected and documented behavior.
Add checks to free() and malloc_usable_size() to
verify (sort of) that the pointers are actually
pointing at valid allocated memory before dereferencing
them and using the chunk_begin struct info.
This will catch use-after-free and wildly bad pointers
a little bit earlier.
