Previously it was optional to cater for potential future cases where we'd not
want errors reported while still verifying types. That lead to some
inconvenient usage patterns inside inference function. One can get the same
behavior via the diagnostic handler and given the vast majority of uses has a
location, flip it so that the common case is the easy case.
Details
Details
- Reviewers
mehdi_amini rriddle herhut
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Unit Tests
Unit Tests
Time | Test | |
---|---|---|
40 ms | x64 debian > Flang.Semantics::resolve102.f90 |
Event Timeline
Comment Actions
LGTM in general, though one comment about the diagnostic handler.
mlir/test/lib/Dialect/Test/TestPatterns.cpp | ||
---|---|---|
103 | I don't think we have a general utility for this. We would need to add something like a NullDiagnosticHandler or DiscardingDiagnosticHandler, that only ignores diagnostics from the current thread(or child threads perhaps?). |
I don't think we have a general utility for this. We would need to add something like a NullDiagnosticHandler or DiscardingDiagnosticHandler, that only ignores diagnostics from the current thread(or child threads perhaps?).