Path: utzoo!attcan!uunet!cs.utexas.edu!rutgers!rochester!pt.cs.cmu.edu!wb1.cs.cmu.edu!avie From: avie@wb1.cs.cmu.edu (Avadis Tevanian) Newsgroups: comp.sys.next Subject: Re: Next Bugs Message-ID: <6247@pt.cs.cmu.edu> Date: 24 Sep 89 08:34:29 GMT References: <2420@ucsfcca.ucsf.edu> Organization: NeXT, Inc. Lines: 249 In article <2420@ucsfcca.ucsf.edu> jst@cca.ucsf.edu.UUCP (Joe Stong) writes: >I got private mail from Avadis Tevanian, who appears to be a software >manager at NeXT. As it was private, I will not include it, but I will >mention the content: It did NOT contain either helpful suggestions or >sympathy about my situation in maintaining a NeXT box. It merely >re-iterated observations that I had already made, and told me what I CANNOT >do with the NeXT box (namely remote system administration). I'd much rather >be told what I *CAN* do to improve my situation. I don't have much choice >in doing remote system administration. > >I would expect better public relations from a NeXT employee! This last line really hurts. I'm trying as hard as possible to provide reasonable answers to people in what little time I have available. In any case, when I wrote to Joe, I basically was interested in knowing if he was reporting 1.0 bugs or was going through his laundry list of 0.9 bugs. Joe is indeed still running 0.9. Here is the text of the message I sent to Joe: >You seem to be posting lots of bugs to comp.sys.next that are related to >0.9. Is this the case or are you finding them in 1.0 as well? > >Also, you cannot expect to administer a NeXT easily through an rlogin >connection from somewhere else! At a minimum, as you have discovered, >you need a NeXT to read the documentation (in WriteNow format). Also, >in 1.0, most system administration is done through applications which >require the window system. > > Avie It seemed appropriate to mention to Joe that what he was trying to do was next to impossible. When he mentioned that he uses "strings" it seemed appropriate to mention that he needs a NeXT to read it (or at least print it out). In any case, Joe has posted a long list of bugs, suggestions and complaints. Let me try to address those that make sense in this forum. >I'm indeed working with the machine with only the online documentation, >and through a rlogin on the ethernet. I don't yet know how to >get reasonable printouts of .wn (WriteNow) documents. Sometimes I >"strings" them. My understanding is that all the printed documents >are availiable on-line (Are they?). The way to print out WriteNow documents is to run WriteNow on a NeXT machine (or even a Mac will work if you can transfer the files). Not all printed documents are on-line (e.g., User Documentation is not online). Is this stuff fixed in 1.0? #The manual page for "passwd" mentions nothing about NetInfo. There #should be cross-references to niutil, niload, and nidump, and nu. #All sorts of wierd interactions occur. Users are generally expected to use Preferences to change their password. System Managers can use UserManager if they wish. niutil, niload, ... are provided to assist in managing a heterogeous network (e.g., get or put your database on machines without NetInfo) and are documented elsewhere. #chfn and chsh don't work. This should be stated up front, when the #programs run, and not after you've typed in several lines of input #which it subsequently refuses to admit to NetInfo. We just removed these from the 1.0 release since we didn't have time to make them work using NetInfo. Again, UserManager (or NetInfoManager) can be used to change this information. #There should be a program that gets all the files synchronized with #each other. Well, niload and nidump attempt to do this. I recommend that one keep things synchronized by just using the Administration Apps (NetInfoManager can handle just about anything related to NetInfo). In those cases where you'd like to edit a file, the best bet is to nidump, edit and niload. #There needs to be a document about NetInfo in general. Included online in 1.0. #/bin/passwd only updates NetInfo. It doesn't touch /etc/passwd or # /etc/passwd.dir or /etc/passwd.pag or /usr/adm/nu.passwd This is a feature. Don't be surprised if those files aren't even on our next software release. #It is possible for more than one entry to exist in NetInfo for users. #We had it happen here. The output from nidump had an error message #concatenated with the root passwd entry. The error message should have #gone out stderr and not stdout, and should have had a newline following it. #We're still not sure how the multiple entries were created. I need more info to understand this one. I think nidump has been fixed to use stderr, but I don't have any way to verify it at this moment. #Niutil will not -destroy a blank entry. What ARE those reference #numbers for, anyway (on the niutil -list listings)? There's no #documentation on how to use them. We had to resort to the #console administration program to do the job, which was annoying. The reference numbers are just NetInfo directory tags. You don't really use them unless you are a NetInfo wizard and want to know what they are. #nu is obnoxious about not allowing one to enter multiple superuser #accounts. We use them so that different people have different superuser #passwords. (It doesn't allow a uid of 0 to be entered). nu is obnoxious is several ways. Its a tradeoff between letting experienceds users do things and preventing not-so-experienced users from getting things screwed up. nu errs on the sid of the naive user. #The entry of the Sri-Nic hosts file into NetInfo has been going on for #a couple of hours now, and still isn't done. Is this really the way to #do things? I believe that this aspect of NetInfo has been sped up in 1.0; however, the better way to get the entire Internet namespace is to run BIND (we do within NeXT, for example). #We have avoided running the Yellow Pages, and don't know what complexity and #bugs this would bring up. If you don't need it, don't bother running it. >Is there a sensible way for me to get at the .wn documents on an ascii >terminal? No, you really need to use a NeXT to run WriteNow. # We are trying to back up files on our optical disk. # # Karpinsi's attempts to use "dump" resulted in various hung copies of dump. You can use dump so long as you don't go beyond the size of the OD. dump gets confused when you run out of space on the OD. I don't know why you are getting hung copies of dump, we've tried it and it works fine (at least in 1.0). # There is no online documentation (unix man page) for the od device driver. There is nothing you need to know about for the od device driver, do you have specific questions? # It would be nice if you're accessing the od from a standard unix terminal # (via rlogin) that it wouldn't put a dialogue box up on the main screen # asking to insert a new disk. Well, you need to go to the machine anyway to insert the disk :-). I agree though that a message on your rlogin'ed terminal also seems appropriate though. # We have a disk that allows itself to be inserted and written upon. # It dumps tar with an I/O error about 41 megabytes into the 250MB # surface. Do you have more info on the I/O error? You may have a bad disk. # We want to re-cycle the 0.8 distribution disk. The system won't let us # put it in, it says something like "wrong volume". How do we re-label # the disk if it won't even let us insert it. # Sep 21 14:28:43 ccnext vmunix: Please insert new disk for volume 0 # Sep 21 14:28:43 ccnext vmunix: (press 'n' key if volume is not available) If you request a disk to be initialized (disk -i or from workspace) and this happens the only thing I can think of is a 0.9 bug when dealing with multiple volumes. I've initialzed many disks and used multiple volumes on 1.0 with no problems (0.9 was very preliminary, 1.0 is much more solid). # The "su" command doesn't seem to run ANY .cshrc or .login, neither the # one in my home directory nor root's home. Where should I put the .cshrc? # I'm su'ing to a secondary root account: # jstr:xxx:0:1:jdkjfkjdkf:/home/jst:/bin/csh The classic pilot error problem here is to have the wrong permissions on the .cshrc and .login. The csh checks this to prevent security problems. If everything is set up right, it will run the .cshrc from the home directory of the user being su'ed to. # There is various documentation in WriteNow format (.wn) files that I # cannot read, because I'm using a standard ascii terminal. Is there a # way to convert this stuff so I can read it? Again, use a NeXT and print it out. If you still want it online, you can use WriteNow to save as an ascii file, but you still need to use a NeXT to do that. # I discovered the "disk" command, only through the NetNews group, # comp.sys.next . It has an interestingly misleading help message, from # which I might conclude that typing "disk -i /dev/rod0a" is a way to # bring it up in interactive mode. Pretty frightening, -i is INITIALIZE, # and not interactive. Correct, disk -i is initialize. I don't know why its frightening. # Getting into the "disk" command, we discover that ALL the bad blocks # are taken on this floptical cartridge. It continues to read and # write on bad areas of the cartridge, but it takes a huge amount of time # 10-40 seconds instead of 1/3second to write 8 sectors, and generates # lots of errors. Reinitialize the disk under 1.0 (which uses a new remap strategy). If the bad block table still fills up, then throw away the media, it is bad. # The documentation about assigning badblocks indicates we should use # block numbers out of /usr/adm/messages. We get block numbers there # that are sometimes larger than 250MB. # Sep 22 14:43:29 ccnext vmunix: od0a: write failed (ECC) # block 250448 phys block 250459 (19802:0:11) The disk does actually have more than 250MB on it. Although, 250459 is NOT larger that 250MB. In any case, badblock remaping is done automatically for you. # The disk program is confusing about writing patterns. Why does it # ask, "random pattern" in the write command, when you actually use # the "set" command to set the pattern to be written? Because it only writes the pattern you set if you say no to the "random pattern" questoin. # Why is there no major/minor number device assigned to the font and # back porches of the od? How about doing programs like disk with # a script front end and some hardware interface programs, so a smart # user could begin to decipher things without the OS source? Most information like this is accessed via an ioctl interface. Noone should ever need to play around with this stuff. And if you have an application that needs this info, you should contact NeXT about getting the info you need. # How do you increase the size of the badblock space on a floptical # cartridge? Should we throw this one away? Back to-how do we # recycle an old distribution cartridge? You don't, if you have a suspicious OD, reinitialize under 1.0, if it still doesn't work, throw it away. ># disk /dev/rod0a >disk>bulk > >Crashes the system, reliably. "Golly" folks, what's happening? :-( Fixed in 1.0. -- Avadis Tevanian, Jr. (Avie) Manager, Systems Software NeXT, Inc. avie@NeXT.COM