louie@trantor.umd.edu (Louis A. Mamakos) (12/22/88)
We just got a few NeXT machines in. Pretty neat; I have high hopes for the released version of the software. There are a few problems with the 0.8 release, especially with printing. Anyway, here's a list of bugs and problems that I have jotted down which playing with the machine thus far: It would sure be nice if there was a tool or shell script to set up sendmail, /etc/named rather than ypsrv -i. It should be easy to replace gethostbyname() in the shared libc.a, but I don't know how to do this. -- No /usr/lib/aliases. No `newaliases' program. sendmail complains of `dbm: no open database' on SMTP connection when referencing user names. -- No kernel objects, so it is impossible to rebuild kernel with a different configuration (e.g. changing value of tickadj) -- No man page for `otool', or how to use it (I guess) to rebuild shared libraries. -- finger sez: finger: /usr/adm/lastlog open error -- The .cshrc shouldn't add `.' to the path if the shell is a super-user shell. -- How do you set the erase character to ^H while rlogin'ing in? -- tset/termcap(3) library doesn't use TERMPATH env variable. -- I'm real impressed with the performence of the ethernet interface. Try doing some FTPs in binary mode.. Also, has anyone figured out how to open a Terminal or Shell app with more than 24 lines? It even has GNU emacs 18.52... Louis A. Mamakos WA3YMH Internet: louie@TRANTOR.UMD.EDU University of Maryland, Computer Science Center - Systems Programming
jdi@camelot..Berkeley.EDU (John Irwin) (12/22/88)
In article <3048@haven.umd.edu> louie@trantor.umd.edu (Louis A. Mamakos) writes: ... > >No kernel objects, so it is impossible to rebuild kernel with a different >configuration (e.g. changing value of tickadj) >-- ... Speaking of which, has anyone else noticed that when booting the machine prints "Allocating 16 buffers (.12 Mb)"? On a Sun (OS<4.0) the "16" is the total number of memory disk buffers. I sure hope this isn't the case in Mach... Anyone know if Mach uses the standard 4.[23]BSD buffer mechanism? If it's really the total disk buffer pool this could be the cause for the somewhat disappointing I/O performance. -- John
bet@dukeac.UUCP (Bennett Todd) (12/23/88)
In article <3048@haven.umd.edu> louie@trantor.umd.edu (Louis A. Mamakos) writes: >tset/termcap(3) library doesn't use TERMPATH env variable. Excuse my ignorance, please, but what is the TERMPATH environment variable supposed to contain? I knew about TERM and TERMCAP of course, and the gratuitously incompatible TERMINFO. Never heard of TERMPATH. What UNIX does it come from? -Bennett
louie@trantor.umd.edu (Louis A. Mamakos) (12/23/88)
The TERMPATH environment variable appeared in post 4.3 BSD systems, and probably first saw the light of day in the 4.3 BSD-tahoe release. It is real handly; it allows you to specify a path of termcap files. For instance, mine looks like: TERMPATH=$HOME/mytermcap:/etc/termcap I can have my private TERMCAP, still have access to all of the stuff in the system termcap. The .cshrc sets up TERMPATH, then invokes tset to set TERM and the TERMCAP environment variables once its picked a terminal type. On another topic: anyone figured out how to make the NeXT use the domain name system without the yellow pages stuff? If I was real clever and had documentation, I could rebuild the shared libc.a and replace gethostbyname() and gethostbyaddr(); of course this depends upon the applications *using* the shared libc.a. Louis A. Mamakos WA3YMH Internet: louie@TRANTOR.UMD.EDU University of Maryland, Computer Science Center - Systems Programming
gh3@mentor.cc.purdue.edu (Gerrit Huizenga) (12/23/88)
In article <8474@pasteur.Berkeley.EDU> jdi@camelot.UUCP (John Irwin) writes: > >Speaking of which, has anyone else noticed that when booting the machine prints >"Allocating 16 buffers (.12 Mb)"? On a Sun (OS<4.0) the "16" is the total >number of memory disk buffers. I sure hope this isn't the case in Mach... >Anyone know if Mach uses the standard 4.[23]BSD buffer mechanism? > >If it's really the total disk buffer pool this could be the cause for the >somewhat disappointing I/O performance. > > -- John While going through a set up exercise with the NeXT machines I noticed the same thing and asked one of the guys from NeXT about this. He said that files are memory mapped and in general there is little need for a pool of standard bufs. This is a side-effect of MACH rather than NeXT's implementation if I understand correctly. If he is correct, this shouldn't have any direct effect on performance. Gerrit Huizenga, Purdue University Computing Center gh3@mentor.cc.purdue.edu