[Lldb-commits] [compiler-rt] [libcxx] [libcxxabi] [libunwind] [lldb] [llvm] [compiler-rt] Disable LLVM_ENABLE_PER_TARGET_RUNTIME_DIR=ON on AIX. (PR #131200)

[David Tenty via lldb-commits]

================
@@ -223,6 +223,13 @@ endif()
 # This can be used to detect whether we're in the runtimes build.
 set(LLVM_RUNTIMES_BUILD ON)
 
+if (LLVM_ENABLE_PER_TARGET_RUNTIME_DIR AND UNIX AND ${CMAKE_SYSTEM_NAME} MATCHES "AIX")
+  # Set LLVM_ENABLE_PER_TARGET_RUNTIME_DIR=OFF as AIX doesn't support it
+  message(WARNING
+  "LLVM_ENABLE_PER_TARGET_RUNTIME_DIR=ON is not supported on AIX. LLVM_ENABLE_PER_TARGET_RUNTIME_DIR is set to OFF.")
+  set(LLVM_ENABLE_PER_TARGET_RUNTIME_DIR OFF CACHE BOOL "" FORCE)
+endif()
----------------
daltenty wrote:

Yeah, this argument makes a lot of sense to me as well. 

> we don't want to have configuration logic, especially when it's imperative, in the CMake files

IIUC it sounds like what we are say is that the CMakeLists shouldn’t be making decisions about how the target is configured. Specifying platform configuration defaults is best left to other mechanisms such as caches file (some of which already handle this option for example)

That being the case, if the user specifies `LLVM_ENABLE_PER_TARGET_RUNTIME_DIR=ON` we know walking into an config we know happens to be un-implemented for the target. But they asked for what they asked for, so issuing a fatal error (rather than giving them something else) makes sense to me.


https://github.com/llvm/llvm-project/pull/131200