[libcxx-commits] [PATCH] D96842: [dfsan] Do not specialize vector<bool> for DFSan

[stephan.yichao.zhao via Phabricator via libcxx-commits]

stephan.yichao.zhao added a comment.

DFSan's memory and CPU overhead is 2-3x because each user byte has 1-2 shadow byte, and user instructions have additional inserted code to propagate shadow data. So the overhead added to std::vector<bool> is acceptable.

The ideal fix is like this change. This way DFSan can accurately track flows.

Zeroing out after reserve can ensure DFSan does not read undefined shadow data, but because vector<bool> updates data at bit-level while dfsan tracks at byte-level, dfsan's tracking would not be accurate.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D96842/new/

https://reviews.llvm.org/D96842