RISC-V: JALR `rd == 0` does not emit a branch

[emesare]

However, I encountered another problem: "functions" that are called with jalr rd, rs1, imm usually end with jr rd. Obviously, the control-flow is the target "function" is unresolved, since rd does not have defined value in the context of the function.

As a matter of fact, this target "function" (ending with jr rd) is most probably just one or multiple basic blocks that are shared between multiple calling functions for space optimization.

Is there any way to tell BN to treat these target "functions" as such ? (i.e. not functions, but basic blocks to reattach to different parent functions). Doing this manually would be the equivalent of "Append function tail" feature in IDA ; that would be great is BN would handle this automatically (BN usually has no problem with shared basic blocks IIRC)

Or maybe my patch is incomplete, and this is already handled.

Anyway, thanks in advance

Forwarded for @jeanmicheldeva (original message)

[emesare]

While the first idea would be to emit an unresolved branch this would be ill-advised as that same instruction is used for function calls, some changes in analysis would be required to emit an unresolved branch that functions as a call in certain scenarios. I have yet to look into exactly what needs to change but I think it's pretty straightforward.