hanche@imf.unit.no (Harald Hanche-Olsen) (05/26/90)
Have a look at the following, folks! vifsla 90 % ls Makefile README.more Tangle.c.ORIG entangle.a88k* untangle.m68k* Makelog Tangle.c ToDo entangle.m68k* README Tangle.c.DIFF debate untangle.a88k* vifsla 91 % mv README.more NEWS mv: NEWS: rename: No such file or directory vifsla 92 % ls Makefile Tangle.c ToDo entangle.m68k* Makelog Tangle.c.DIFF debate untangle.a88k* README Tangle.c.ORIG entangle.a88k* untangle.m68k* vifsla 93 % Yup, I swear to you, the above is copied directly out of my xterm where this happened. "mv" complained that the target file did not exist, and voila! It did not exist! Unfortunately, neither did the original... Wanna see it again? Sorry, but I don't know how to reproduce this error. Still, that the file system is able to perform feats like that sort of scares me... It would have been reasonable, wouldn't it, to expect "mv" FIRST to create a new link, and THEN, IF SUCCESSFUL, to remove the old one? Doesn't look like it... This time the file that was lost was of no importance at all. But let it serve as awarning to us all, to take backup seriously. And the folks at HP/Apollo might want to look for a bug in the code for "rm". (Details: The exchange happened on a diskless DN3000 booted from a DN3500, but the disappearing file resided on a DN10000. All involved machines run SR10.2). - Harald Hanche-Olsen <hanche@imf.unit.no> Division of Mathematical Sciences The Norwegian Institute of Technology N-7034 Trondheim, NORWAY