Ralf, > > According to a comment on another thread from Ralf, this has been observed > > in the wild only once. We can simplify the code and remove that comment. > > Why not just use the ADDIU and be done with it? > > > > There are many locking and atomic primitives that don't have any such error > > checking. What makes the read lock so special that it needs this extra > > protection? > > Because I was desparate to find a use for the signed add ;-) > > Honestly, it's nice to have such a safeguard if it's available at no > runtime overhead at all but these days are such nice lock debugging tools > that the loss won't be missed. So (cut'n'paste): > > Why not just use the ADDIU and be done with it? Given David's comment I meant to defer to you on this as the originator of this bit, but since you have spoken, I think we've come to a conclusion now. :) Maciej
