As noted in #449, on 1.10, we intermittently encounter
Error During Test at C:\Users\runneradmin\.julia\packages\EnzymeTestUtils\fNmYk\src\test_forward.jl:71
Got exception outside of a @test
KeyError: key (ℂ^2 ⊗ (ℂ^3)') ← (ℂ^2 ⊗ (ℂ^3)' ⊗ (ℂ^4)') not found
Stacktrace:
[1] getindex
@ .\dict.jl:498 [inlined]
[2] _unsafe_getindex
@ C:\Users\runneradmin\.julia\packages\LRUCache\ZH7qB\src\LRUCache.jl:189In the multiplication tests. This seems to only occur for Vtr, and what's especially strange about it is right before the "offending" line in LRUCache.jl, haskey(lru, key) (with this key) returned true while a lock was held. Some Enzyme friends are looking into this, but I'm also opening an issue as @lkdvos asked so that we can track this and reference it in the test file.
The error never occurs if we set the CacheStyle of degeneracystructure to NoCache or set the cache size of the GlobalLRUCache to 0 (not surprising since there is then no cache to miss on).
Removing one space from the product in the domain is seemingly enough to resolve this error, making me think it's some kind of weird race condition, but I've been unsuccessful in figuring out more so far.
As noted in #449, on 1.10, we intermittently encounter
In the multiplication tests. This seems to only occur for
Vtr, and what's especially strange about it is right before the "offending" line inLRUCache.jl,haskey(lru, key)(with this key) returnedtruewhile a lock was held. Some Enzyme friends are looking into this, but I'm also opening an issue as @lkdvos asked so that we can track this and reference it in the test file.The error never occurs if we set the
CacheStyleofdegeneracystructuretoNoCacheor set the cache size of theGlobalLRUCacheto 0 (not surprising since there is then no cache to miss on).Removing one space from the product in the domain is seemingly enough to resolve this error, making me think it's some kind of weird race condition, but I've been unsuccessful in figuring out more so far.