On Sat, 4 Apr 2009, Tony Abernethy wrote: > Any differences in efficiency between cp and mv > with be swamped by any time taken thinking about it.
Huh? the difference in efficiency isn't the issue, the issue is the state of the disc directory "/". cp /bsd.foo /bsd Suppose the / fs is the old, standard ffs and is mounted async. It may be "some time" before the file buffers are flushed (synced). A crash during that interval can result in a corrupted /bsd. This, I think is the old bugaboo. mv will be a call to "rename(2)", and in most sane implementations of ffs, the resulting changed directory entry (i.e. the changed inode) will be flushed to disk immediately. There was a time when disk writes were much slower than now. Still, we should consider the time taken by cp to be "long", and since I believe that cp is "interruptable", cp may be paged out for higher priority activities. This being the case, we cannot guarantee that cp will complete in certain time, or if it will complete at all. Cp makes many system calls, and can be interrupted or even swapped out. This is a distinct possiblilty on a busy machine or one that is "swapping". OpenBSD is not magic -- and its behavior when swapping is actually miserable. (1GHz dual Pentium III: five-ten second delays refreshing the screen with mozilla when that bloated, memory-leaking, crash-prone pig has consumed 400 MB of core and is eating into swap -- this even when X itself is offloaded to an X-terminal). so anything with /bsd as the destination of a cp(3) is "dangerous". Dave _______________________________________________ Openbsd-newbies mailing list [email protected] http://mailman.theapt.org/listinfo/openbsd-newbies
