I suspect a race condition due to the nature of the panic in the CI log above -- a nil receiver seems to have been dereferenced after a conditional nil check should have executed -- this seems to indicate that it was concurrently modified (if gccgo stack traces can be trusted).
Uniter cleanups are called from a separate goroutine, so this does seem possible. Unfortunately, I cannot personally reproduce the panic on my machine.
http:// reviews. vapour. ws/r/2677/
I suspect a race condition due to the nature of the panic in the CI log above -- a nil receiver seems to have been dereferenced after a conditional nil check should have executed -- this seems to indicate that it was concurrently modified (if gccgo stack traces can be trusted).
Uniter cleanups are called from a separate goroutine, so this does seem possible. Unfortunately, I cannot personally reproduce the panic on my machine.