DeforaNetworks/deepstate
3
0
Fork 0
Code Releases Activity

some readme cleanup, removing duplicate section

Browse Source
This commit is contained in:
Alex Groce 2019-01-10 14:00:33 -07:00
parent 6faf3dab38
commit 7faf6a2e6b
1 changed files with 17 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
README.md
View File

@ -252,14 +252,6 @@ useful for understanding what a DeepState harness is actually doing;
often, setting `--log_level 1` in either fuzzing or symbolic often, setting `--log_level 1` in either fuzzing or symbolic
execution will give sufficient information to debug your test harness. execution will give sufficient information to debug your test harness.
## A Note on Mac OS and Forking
Normally, when running a test for replay or fuzzing, DeepState forks
in order to cleanly handle crashes of a test. Unfortunately, `fork()`
on mac OS is extremely slow. When using the built-in fuzzer or
replaying tests, it is highly recommended to add the `--no_fork`
option on mac OS, unless you need the added crash handling (that is,
things aren't working without that option).
## Built-In Fuzzer ## Built-In Fuzzer
@ -279,13 +271,13 @@ Note that while symbolic execution only works on Linux, without a
fairly complex cross-compilation process, the brute force fuzzer works fairly complex cross-compilation process, the brute force fuzzer works
on macOS or (as far as we know) any Unix-like system. on macOS or (as far as we know) any Unix-like system.
## A Note on Mac OS and Forking ## A Note on MacOS and Forking
Normally, when running a test for replay or fuzzing, DeepState forks Normally, when running a test for replay or fuzzing, DeepState forks
in order to cleanly handle crashes of a test. Unfortunately, `fork()` in order to cleanly handle crashes of a test. Unfortunately, `fork()`
on mac OS is _extremely_ slow. When using the built-in fuzzer or on macOS is _extremely_ slow. When using the built-in fuzzer or
replaying more than a few tests, it is highly recommended to add the `--no_fork` replaying more than a few tests, it is highly recommended to add the `--no_fork`
option on mac OS, unless you need the added crash handling (that is, option on macOS, unless you need the added crash handling (that is,
only when things aren't working without that option). only when things aren't working without that option).
## Fuzzing with libFuzzer ## Fuzzing with libFuzzer
@ -328,12 +320,17 @@ CC=/usr/local/opt/llvm\@6/bin/clang CXX=/usr/local/opt/llvm\@6/bin/clang++ BUILD
make install make install
``` ```
On mac OS, libFuzzer's normal output is not visible. On any platform,
you can see more about what DeepState under libFuzzer is doing by
setting the `LIBFUZZER_LOUD` environment variable.
Other ways of getting an appropriate LLVM may also work. Other ways of getting an appropriate LLVM may also work.
On macOS, libFuzzer's normal output is not visible. Because libFuzzer
does not fork to execute tests, there is no issue with fork speed on
macOS for this kind of fuzzing.
On any platform,
you can see more about what DeepState under libFuzzer is doing by
setting the `LIBFUZZER_LOUD` environment variable, and tell libFuzzer
to stop upon finding a failing test using `LIBFUZZER_EXIT_ON_FAIL`.
## Test case reduction ## Test case reduction
While tests generated by symbolic execution are likely to be highly While tests generated by symbolic execution are likely to be highly
@ -442,7 +439,11 @@ input file) in your `main`.
Because AFL and other file-based fuzzers only rely on the DeepState Because AFL and other file-based fuzzers only rely on the DeepState
native test executable, they should (like DeepState's built-in simple native test executable, they should (like DeepState's built-in simple
fuzzer) work fine on macOS and other Unix-like OSes. fuzzer) work fine on macOS and other Unix-like OSes. On macOS, you
will want to consider doing the work to use [persistent mode](http://lcamtuf.blogspot.com/2015/06/new-in-afl-persistent-mode.html), or even
running inside a VM, due to AFL (unless in persistent mode) relying
extensively on
forks, which are very slow on macOS.
## Contributing ## Contributing