[Lldb-commits] [lldb] [LLDB][Telemetry]Define DebuggerTelemetryInfo and related methods (PR #127696)

[Pavel Labath via lldb-commits]

================
@@ -56,13 +60,83 @@ struct LLDBBaseTelemetryInfo : public llvm::telemetry::TelemetryInfo {
   void serialize(llvm::telemetry::Serializer &serializer) const override;
 };
 
+/// Describes the exit status of a debugger.
+struct ExitDescription {
+  int exit_code;
+  std::string description;
+};
+
+struct DebuggerTelemetryInfo : public LLDBBaseTelemetryInfo {
----------------
labath wrote:

I have many thoughts here. In no particular order, they are:
- I don't find the I/O argument particularly convincing, as I expect that any reasonable backend will want to cache/batch sending of these entries anyway
- I can believe that for analysis purposes, it's nicer to have a single "this is everything I know about the environment of the session" entry. I can also understand the desire for easy filtering of unwanted data. I don't know if there's a reasonable compromise here. Group this data into two sub-structs?
- the collect-then-filter approach makes sense for things which cannot be collected in any other way. However, pretty much everything being collected here is static data. There's no reason why it has to be collected at a particular location. (Like, even if we just stopped collecting the username here, a backend could just refetch it if it wanted to). Maybe we could not include things that recomputed in the backend? Or at least things that are likely to be problematic?

https://github.com/llvm/llvm-project/pull/127696