Michael137 added a comment. In D134493#3811097 <https://reviews.llvm.org/D134493#3811097>, @labath wrote:
> In D134493#3810290 <https://reviews.llvm.org/D134493#3810290>, @Michael137 > wrote: > >> Wasn't sure how to properly test this since the original reproducer >> technically relies on implementation-defined behaviour (i.e., initialising a >> bitfield with an out-of-range value). Suggestions are welcome > > I'm probably missing something, but what exactly is undefined about that test > program? The number eight fits comfortably in four bits, and afaik it is a > valid value for the `EnumVals` type because it has a fixed underlying type. > > Other than that, this seems fine to me. Ah yes, absolutely true! Misinterpreted it from some of the other reproducers. It's enough for the value we assign (here `8`) to be out of range of the bit-field (if it were signed). Since we previously treated all bit-fields of enum type as signed, that would incorrectly left-fill with 1s. I.e., if the bit-field were of width 16, then 32768 would sign-extend incorrectly. I'll update the test description accordingly Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D134493/new/ https://reviews.llvm.org/D134493 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
