mirror of
https://github.com/GaloisInc/macaw.git
synced 2024-11-29 21:44:11 +03:00
6a4f406c68
It turns out that we have to be more conservative with tail call identification, as incorrectly identifying a block as the target of a tail call (instead of a branch) can cause other branch classifiers to fail if that block is the target of another jump. Ultimately, we will need to give up some tail call recognition (since they are in general indistinguishable from jumps), and instead only identify known call targets as tail call candidates. With additional global analysis we could do better. Fixes #294 |
||
|---|---|---|
| .. | ||
| x64 | ||
| ElfX64Linux.hs | ||
| Main.hs | ||
| README.rst | ||
This directory contains some automated tests for macaw-x86. The tests currently cover the code discovery implementation in macaw. Specifically, they check: 1) That the correct number of functions are found, 2) Each function starts at the expected address, 3) Each function contains the correct number of basic blocks, 4) Each basic block starts at the expected address The ``Makefile`` in the tests directory rebuilds the test files from source. The resulting binaries are checked in to the repository so that Mac OS users can run the tests without having to have a Linux build toolchain available. Additionally, having the binaries in the repository keeps the code layout stable. Expected addresses are hard-coded into the expected results of the test suite; regenerating the binaries on a different system or with a different compiler could change the offsets in the test binaries, which would require the expected results to be tweaked. This is unfortunate, but difficult to avoid in a robust way. Note that the tests are bare-bones and do not link against libc. They issue the exit system call directly to terminate. This lets us get small and comprehensible test cases.