vgvassilev wrote: > I also considered the process-wide nature of this design. It would **not > allow concurrent compiler sessions** within the same process or WebAssembly > instance to select different informational streams. However, our > xeus-cpp/JupyterLite and WasmBolt execution models serialize compiler > invocations, and native kernels and Web Workers have isolated process or > WebAssembly state. CppInterOp also currently uses a process-wide > active-interpreter model rather than concurrent interpreters with independent > output routing.
Not sure I follow. Any stream will have concurrency issues and in that sense `llvm::errs` and `llvm::outs` are no different. The xeus-cpp and CppInterOp example can use the usual solution we have for `llvm::errs` -- how is that different? https://github.com/llvm/llvm-project/pull/226022 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
