[Lldb-commits] [PATCH] D68434: SBFile support in SBCommandReturnObject

Pavel Labath via Phabricator via lldb-commits lldb-commits at lists.llvm.org
Fri Oct 4 02:06:50 PDT 2019

labath added inline comments.

Comment at: lldb/include/lldb/Utility/ReproducerInstrumentation.h:19
+#include "lldb/API/SBFile.h"
This is definitely not right. `Utility` should never include other lldb libraries. I think this stuff should be defined in the API folder. Either in the SBFile.cpp directly, or some additional cpp file.

Comment at: lldb/source/Utility/ReproducerInstrumentation.cpp:38
+template <> lldb::SBFile Deserializer::Deserialize<lldb::SBFile>() {
+    //@JDevlieghere I'm pretty sure this is not the right thing to
lawrence_danna wrote:
> @JDevlieghere advice please.
This is not totally surprising as SBDebugger::SetInputFile ignores the input SBFile when it's in replay mode. However, it still doesn't seem like the right fix. I am guessing that something special needs to happen in the record/replay logic of the SBFile constructor, but off-hand, it's not fully clear to me what would that be.

I think ideally, we'd have the reproducer shadowing that's currently done in SBDebugger::SetInputFileHandle kick in earlier (in the SBFile constructor) so that it covers all SBFiles, and not just those that later become the debugger input handles, but I don't think it should be on you to make all of that work. Maybe the appropriate fix would be to just make sure the SBFile constructor creates empty objects and acknowledge that reproducers won't work for all SBFiles until someone actually implements the appropriate support for them.

  rG LLVM Github Monorepo



More information about the lldb-commits mailing list