JIT: Add bounds check for ordinals in ThreeOptLayout::ConsiderEdge
#111719
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When setting up the data structures needed for 3-opt layout, we allocate an array of basic blocks that is large enough to fit everything reachable via a RPO traversal. However, we only initialize as much of the array as needed to fit the blocks hot enough to be considered for reordering, so we frequently have uninitialized space at the end of the array. Thus, our current strategy to check if a block is in the set of candidates takes two steps:
We were neglecting to do the first step in
ThreeOptLayout::ConsiderEdge
, which meant in rare cases, we would compare a block outside the candidate set to some uninitialized part of the array, and the comparison would happen to be equal. Thus, we would later consider creating fallthrough along a flow edge that breaks 3-opt's invariants.To solve this, we can either add the bounds check mentioned above, or we can default-initialize the entire block array to ensure comparisons outside the candidate set are with
nullptr
. I've decided to pursue the former.Fixes #111563.