Such dbg.assigns will occur if you write zero-sized memcpys (see https://reviews.llvm.org/D146987#4240016). I figure this is safer to handle this in AssignmentTrackingAnalysis (back end) rather than AssignmentTrackingPass (declare-to-assign) in case it is possible to reproduce this as a result of optimisations.
Details
Details
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Comment Actions
LGTM, although I feel like this is the kind of thing that we should design-out of the compiler, by adding an assertion in fragment creation that you're not creating a zero sized fragment. Seems like this is a reasonable patch for now though.