rocallahan wrote: > (The precommit test failures look real though)
Indeed. The problem was in `ThreadList::WillResume()`, which is a tricky function that my changes made even trickier. The basic problem is that we make execution direction depend on the state of the thread plans and which thread we're going to run, but `SetupToStepOverBreakpointIfNeeded()` depends on the direction and and can change the thread plan and which thread we're going to run --- so conceptually we have cyclic dependencies. It all works out in the end because `SetupToStepOverBreakpointIfNeeded()` only does something if `direction` is forward, and it can't change the direction to backward, but we have to be very careful. To make this more correct and easier to review, I've rewritten my changes to `ThreadList::WillResume()` and factored them out into multiple commits in this PR. First there's a series of commits that refactor `WillResume()` on the main branch, step by step, getting it into a state where the changes for reverse execution are simple. Hopefully it's clear how each of those refactoring commits preserves the current behavior. Then there's the reverse-execution commit itself. Then there's one more commit that does a little bit more cleanup of `WillResume()`. Up to you/Jim whether you want me to squash all these together after review, or rebase and merge all of them. Let me know. If it's the latter, then I should probably clean up the commit messages. https://github.com/llvm/llvm-project/pull/112079 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
