Both routines (on Linux, at least) utilize a cache; protect the cache with a mutex to allow concurrent callers.
Details
Diff Detail
- Repository
- rL LLVM
Event Timeline
This is not necessary. NativeProcess classes are only used in lldb-server, which is completely single threaded. If you want to change that, then we should start with a discussion of what you intend to achieve.
Let me post the other two changes to start a broader discussion. We can center the conversation around whether/how to prime the caches; the other two changes naturally follow from that.
All your other changes are client-side, so still think this will not matter, but I'll take a look. :)
I meant to respond to the discussion today, but I got sidetracked by the ipv6 thingy. I'll try to look at that tomorrow.
Interesting. Maybe this only matters in the context of unit tests? I was still getting crashes without this change, but now I'm not (I've since git pull'd and switched to another machine). I'll keep investigating.
Cool, glad that's sorted. I've had some ideas about introducing limited amount of paralellism to the server side, but that's probably is not interesting to you if you're not doing remote debugging.