On 6/4/26 1:59 PM, Catalin Iacob wrote:
For learning purposes, do you mind quickly sharing your rationale for combining 1 and 2? To me they are in the same area but still separate things: one removes dead code, the other increases encapsulation. Each is a small but worthwhile cleanup, they can be applied in any order and each can be applied without the other. It's true that, while patch 2 without patch 1 works, it would increase confusion but I find that an argument to order 2 after 1, not to say they are one thing. I'd also find it hard to describe what that combined patch does except by calling out two things: "remove dead code and move the declaration". Your suggested "unexport" seems more confusing than clarifying, I'd argue that today the symbol is not exported since that ifdef always compiles the export out.
"Unexport" is commonly used terminology in the Linux kernel for removing EXPORT_SYMBOL*(). I proposed to combine patches 1 and 2 because it is normal that the declaration of a function is moved from a public to a private header file if it is unexported. If patches 1 and 2 are combined, the most important change is removal of EXPORT_SYMBOL_GPL(). So I think that "Unexport scsi_device_from_queue()" would still be an appropriate description for such a patch. Thanks, Bart.
