> On Dec 28, 2018, at 9:50 AM, Mathew, Cherry G. <che...@zyx.in> wrote: > > That seems like a dependency burden on people developing new features. I > can't tell if modload is the only way to get NVMM to run. If it is, then ok. > If not, then there's the option of provisionally turning it off until we fix > the underlying problem.
Uh, no :-) This would have been exactly the same situation if you had changed some other symbol to be a weak-alias. NVMM was using an exported symbol, no different from how a module might use, say, kmem_zalloc(), and you changed the semantics of that symbol in a way that broke a module using it. -- thorpej