Crash: Assertion failed: (vec.back() == currentBlock)

[mhauru]

MWE:

using Bijectors
using Enzyme
using Random, LinearAlgebra

Enzyme.API.runtimeActivity!(true)

_topd(x) = x * x' + I
d = 4
b = Bijectors.PDVecBijector()
z = randn(d, d)
x = _topd(z)
f(x) = sum(transform(b, _topd(reshape(x, d, d))))
Enzyme.gradient(Enzyme.Reverse, f, vec(z))

Output:

Assertion failed: (vec.back() == currentBlock), function addReverseBlock, file /workspace/srcdir/Enzyme/enzyme/Enzyme/GradientUtils.cpp, line 8952.

[75577] signal (6): Abort trap: 6
in expression starting at /Users/mhauru/projects/Enzyme-mwes/current_block_assertion_fail_crash/mwe.jl:15
__pthread_kill at /usr/lib/system/libsystem_kernel.dylib (unknown line)
Allocations: 57218916 (Pool: 57141228; Big: 77688); GC: 76
Abort trap: 6

Version info:

(current_block_assertion_fai...) pkg> status Bijectors
Status `~/projects/Enzyme-mwes/current_block_assertion_fail_crash/Project.toml`
  [76274a88] Bijectors v0.13.14

(current_block_assertion_fai...) pkg> status Enzyme
Status `~/projects/Enzyme-mwes/current_block_assertion_fail_crash/Project.toml`
  [7da242da] Enzyme v0.12.19

julia> versioninfo()
Julia Version 1.10.4
Commit 48d4fd48430 (2024-06-04 10:41 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: macOS (arm64-apple-darwin22.4.0)
  CPU: 10 × Apple M1 Pro
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-15.0.7 (ORCJIT, apple-m1)
Threads: 1 default, 0 interactive, 1 GC (on 8 virtual cores)

[wsmoses]

Can you paste the whole error log and ideally a MWE without additional package dependencies?

IIRC it looked likt the issue was coming from an syrk call

[wsmoses]

Fixed by pending jll bump here: JuliaPackaging/Yggdrasil#8973