Vincent Lefevre wrote: > Bob Proulx wrote: > > Vincent Lefevre wrote: > > > touch: setting times of `blah/cd': Permission denied > > > though blah/cd doesn't exist. That's inconsistent. > > > > Permission denied is not the same as not creating non-existing files. > > If the permission is denied then the file already exists. If you find > > otherwise please describe an exact test case to recreate it. > > Reported as bug 439033.
I repeat, permission denied is not the same as not creating non-existing files. Read that three times before proceeding any further. Repeat until you believe it. If you don't believe it then do not exit the loop. You are losing credibility at a rapid rate when you report a bug immedately after having read that it is not a bug and in reply to a sentence that describes why it is not a bug. > > But it *is* documented! > > OK, I hadn't looked at it. Then why is the behavior on touch not > documented? (You explained the behavior by the limitation of utime, > but I recall that the touch man page doesn't have any reference to > utime.) The use of utime(2) per se is not relevant as such. It is relevant that there is no way to do what you are asking. I described the implementation to try to help in the understanding. > > You have taken this discussion full circle. Touch has no need to > > document anything special about symlinks since it does not do anything > > particular special with regards to symlinks. > > So, why the documentation concerning chmod since it doesn't do any > special treatment? It shouldn't be necessary. I agree with that. I think it was only mentioned there because the POSIX docs mention that it used to behave differently than it does now. So really it probably got documented when the behavior changed but I don't know the history of this one. > And why isn't the special treatment documented in the stat(1) man page? > Note: the fact that the man page mentions a --dereference option > doesn't mean anything as some options are no-op (or default, if > you prefer). The stat(1) command is new and a recent addition. It does not follow the same style and is not at the same level of age as other older commands. Bob -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
