[clang] [llvm] [clang][MBD] set up module build daemon infrastructure (PR #67562)

[Connor Sughrue via llvm-commits]

================
@@ -0,0 +1,25 @@
+// Check that a clang invocation can spawn and handshake with a module build daemon
+
+// RUN: %kill-process "-cc1modbuildd mbd-handshake"
+// RUN: rm -rf mbd-handshake %t
+// RUN: split-file %s %t
+
+//--- main.c
+int main() {return 0;}
+
+// RUN: %clang -fmodule-build-daemon=mbd-handshake -Rmodule-build-daemon %t/main.c &> %t/output-new || true
----------------
cpsughrue wrote:

That's understandable. To ensure I understand, do you want to see an example where the client dies after receiving the handshake response and fails to deliver a payload or do you want to see a payload and response sent back and forth after the handshake? I think it would be sufficient to carry out a second handshake if it's the latter. The only difference between a second handshake and a new payload message is the contents of the buffer, which I don't think is super important since we mainly care about the mechanisms behind preventing zombie processes.

https://github.com/llvm/llvm-project/pull/67562