root@blakex.RAIDERNET.COM (Blake McBride) (03/24/91)
There seems to be some sort of IO problem with ISC UNIX when it writes to a DOS partition. If you mount a DOS (3.3 30MB DOS) partition and then write to it - either through UNIX or VPIX - the DOS file system, on occasion, becomes corrupt. Yes, I can copy 6 relatively small files without a problem, and yes, even some large files usually without a problem. However, I have found that if I am copying a large number of relatively small files (20K) one-at-a-time, approximately 1 out of 50 doesn't make it. By "doesn't make it" I mean that the file seems to copy successfully but after the copy you can check the file with "ls -l" and find that it has zero bytes. Then you can unmount the DOS file system, down the system, reboot under DOS (3.3), and then do a chkdsk. You will find that the DOS file system is corrupted and there are missing clusters. This problem shows itself quite consistently when you try to copy a very large number of files (~25MB) with a number of sub-directories using either xcopy in VPIX or tar in UNIX. This problem is extremely serious! As a professional software developer for DOS applications I need to depend on the file system. Here is a scenario which has happened to me: I spend several months developing a DOS application under UNIX & VPIX. I do this because of the significantly better development tools under UNIX and the larger market of DOS. There are several reasons why I, at times, must keep my application on a mounted DOS partition (other ISC bugs & DOS protect mode debugger only works under DOS) which I won't go into now. I go to modify a module which has been completed for several months and find that it has 0 bytes! Since all of my file copys/saves appeared successful I had no way of knowing that one of my source files contents were gone. And, since the file got clobbered several week ago all my backups have a 0 length file. I LOST over a week of work!!!!!!! I have been in contact with ISC & Multi User Systems about this problem on and off for over a year. I am able to duplicate the problem on an associates machine, so its not my machine. ISC usually tells me that they can't duplicate the problem. On occasion they have said that A) other people have complained about this problem, and B) they can't duplicate the problem *reliably*. This problem is a major problem to me. I am writing this note for two reasons. 1) to find out if others are experiencing this problem, and 2) to attempt to get ISC to address the problem. -- Blake McBride Home (615) 790-8521 3020 Liberty Hills Drive Work (615) 790-1088 Franklin, TN 37064 root@blakex.raidernet.com U.S.A. ...!uunet!mjbtn!raider!blakex!root