Details
Details
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Unit Tests
Unit Tests
Time | Test | |
---|---|---|
2,350 ms | x64 debian > libarcher.races::task-two.c |
Event Timeline
Comment Actions
Updated patch
- changed switch statement so that adding new item to VariantKind would cause compilation warning if the respective case is not added.
- updated RISCVMCExpr::getVariantKindForName as well.
llvm/lib/Target/RISCV/MCTargetDesc/RISCVMCExpr.cpp | ||
---|---|---|
170 | Can this just be llvm_unreachable and then you wouldn't need the first assert? |
Comment Actions
I do not know how to test this. But I changed the patch so that if a new item would be added to VariantKind and that item would not be handled in the switch statement, compilation would give a warning.
I got crash when executed llc -debug -filetype=obj. I am not sure if it can be found in MIR, but this looks unlikely, as it is an MC layer feature.
llvm/lib/Target/RISCV/MCTargetDesc/RISCVMCExpr.cpp | ||
---|---|---|
136 | Looking again. The only caller of this is RISCVAsmParser::parseOperandWithModifier. Are we allowing something additional in assembly parsing with this change? Is that something we should test? |
Looking again. The only caller of this is RISCVAsmParser::parseOperandWithModifier. Are we allowing something additional in assembly parsing with this change? Is that something we should test?