It looks like we're skipping these for the Makefile builds, but the problem shows up in the CMake configuration... These tests fail with the same error:
JITModule.cpp:232 triggered by user code at : Condition failed: result != 0: HalideJITMemoryManager: unable to find address for __sync_lock_test_and_set_1
It's probably worth fixing the underlying cause. @abadams said "it's trying to lower atomics in a way that won't work. I suspect this is a real problem we'll need to figure out at some point."
attn: @steven-johnson assigning to you since Andrew said you would know the most about this.
When I last looked at this I thought the 'right' fix was to update our code to use the newer atomic primitives instead of the older sync ones, but that may or may not be what is needed here.
It looks like we're skipping these for the Makefile builds, but the problem shows up in the CMake configuration... These tests fail with the same error:
It's probably worth fixing the underlying cause. @abadams said "it's trying to lower atomics in a way that won't work. I suspect this is a real problem we'll need to figure out at some point."
attn: @steven-johnson assigning to you since Andrew said you would know the most about this.