peter@hydrovax.nmt.edu (02/28/89)
Hi again, Unfortunately I can not provide examples of the many problems we are having with AIX because I am at school 80 miles from our site. I have installed and administered the system since its installation though. Someone mentioned a painless upgrade to 2.2.1 Our upgrade to 2.2.1 failed miserably due to several bad disks. It was nice of IBM to come out and rectify the situation, but we still had down time due to it. The version that they DID install would not execute shell scripts or binaries properly because for some reason it sets the current working directory to '/' (yes root). Make fails because my files are not in '/'. I have very recently installed a new version (more disks borrowed from IBM). It seems to be ok now. When Make *could* find the files that I was compiling, it didn't work on any imbedded commands. It would either fail without explanation or give an error number that wasn't in the manual. The asynch driver is still broken. Terminals have to be run at 19200 or AIX throws away output characters. I have tried numerous flow control and bit/parity combinations, to no avail. I would really like to hang a 2400 baud modem on the system so I can call in from school, but the system (getty) won't send nl at 2400. It also locks up after I reply to the login prompt. IBM sent out a tech. He couldn't figure it out either, but verified that the systems at IBM (Albuquerque) have the same problem. Of course this may be a hardware bug and BSD might behave the same. The system doesn't provide job control or sockets, which is more a drawback of sysV than AIX specifically. The console mimics job control using the utilities found in the Useability Services, but it is a poor substitute and only good for the person using console. Curses is broken under AIX. I will try and bring an example to post when I return from work this weekend. VS/FORTRAN has bugs. Again I do not want to cite an example until I can bring verified source code to post. I haven't looked under 2.2.1, but the fortran compiler has a poorly written shell script for a front end. It's the three pass variety and ALWAYS linked (even if there were warnings and/or errors) when it was finished. We re-wrote it, but it is a pain none the less. Our earlier version of the OS came with F77. This one doesn't. IBM says we have to purchase it now. The VS/FORTRAN does have its saving graces. VAX mode (although buggy) has drastically reduced the time it took to port some applications from the VAX. The archive/library program is buggy and won't behave the same way two times in a row. This is a particularly nasty problem because Make can't handle an excessive (again I'm not prepared to cite numbers) number of file names to link. It just crashes. There are no man pages, and I have yet to get the 'help' command to yeild anything resembleing an informative answer. Mostly all I get is a '... not found' message. UUCP is broken. Anything I try to run says that it wants some earlier version of the VRM. A lot of files referenced in the manual aren't even there. I *know* it's installed properly because I went back and did it again. Nroff flags that are in the manual do nothing. (Again I'll bring examples next week). DUMP is described wrong in the manual. We had to use some unusual magic cookies on the command line to restore our file systems after 2.2.1 scrambled it. We now avoid the problem by giving dump different flags. I think part of our problem may have been our ignorance. There are several functions in the BSD compatability libraries that do not perform in the manner described in the manual pages. There are other routines that I suspect are just stubs. In the same room with the RT is a VAX 11/730 running MORE/BSD 4.3. The RT runs circles (many many circles) around that machine speed wise, but our programmers claim they get better productivity from the 730 even though it's slow. BSD came with C, PASCAL, Fortran, Ingres, and a host of tools that make life easy. AIX requires that we purchase each of these, yet is more expensive for the basic system than my entire BSD distibution. The BSD versions of the compilers all work and work well. The IBM compilers are buggy and hard to use. I say that the AIX system is unproductive because there is a learning curve, even for people who know sysV. We have two other sysV machines that we do administrative/development work on. We should have been able to come up to speed much quicker than we were able to. I don't beleive that people should have to program/work around pitfalls in a machines architecture and operating system. Especially since it costs money every time we make a port to a new platform. There is a list of problems sitting on my desk. I will document them better and email them to anyone who's interested. Bye, Peter Blemel ----
usenet@cps3xx.UUCP (Usenet file owner) (02/28/89)
in article <2036@nmtsun.nmt.edu$, peter@hydrovax.nmt.edu says:
$ Someone mentioned a painless upgrade to 2.2.1
I upgraded our Model 135 from AIX 2.2 to 2.2.1 with no problems at
all.
$ Our upgrade to 2.2.1 failed miserably due to several bad disks. It was
$ nice of IBM to come out and rectify the situation, but we still had
$ down time due to it. The version that they DID install would not execute
$ shell scripts or binaries properly because for some reason it sets the
$ current working directory to '/' (yes root). Make fails because my files
$ are not in '/'. I have very recently installed a new version (more disks
$ borrowed from IBM). It seems to be ok now. When Make *could* find the files
$ that I was compiling, it didn't work on any imbedded commands. It would
$ either fail without explanation or give an error number that wasn't in the
$ manual.
I've seen none of these problems occur on our system. Our 2.2.1 upgrade
did come with one update diskette for the system. Perhaps you got the
release from before this update diskette and the update fixed the
problems you mentioned.
$ The asynch driver is still broken. Terminals have to be run at 19200 or
$ AIX throws away output characters. I have tried numerous flow control
$ and bit/parity combinations, to no avail. I would really like to hang
$ a 2400 baud modem on the system so I can call in from school, but the
$ system (getty) won't send nl at 2400. It also locks up after I reply to the
$ login prompt. IBM sent out a tech. He couldn't figure it out either,
$ but verified that the systems at IBM (Albuquerque) have the same problem.
$ Of course this may be a hardware bug and BSD might behave the same.
I have two differnet modems hanging off of the builtin RS232C I can log
onto both of them just fine and run UUCP out of them fine. The asynch
driver does go off to outer space after the machine has been running for
about a week and a half. The fix: reboot.
$ The system doesn't provide job control or sockets, which is more a drawback
$ of sysV than AIX specifically. The console mimics job control using the
$ utilities found in the Useability Services, but it is a poor substitute
$ and only good for the person using console.
No job control is a pain. No sockets? Then why on page 2-646 of the AIX
Tech Ref manual is there "sockets library"? AIX 2.2.1 has added
HoneyDanber UUCP, dbx, sendmail, and symbolic links. All of which 2.1.1
didn't have. I suspect we'll be seeing job control soon.
$ Curses is broken under AIX. I will try and bring an example to post when
$ I return from work this weekend.
Is this just one feature of curses which is broken? Our main application
that we use in house is curses intensive and runs on 160 RT's with no
problems.
$ VS/FORTRAN has bugs. Again I do not want to cite an example until I can
$ bring verified source code to post. I haven't looked under 2.2.1, but the
$ fortran compiler has a poorly written shell script for a front end. It's the
$ three pass variety and ALWAYS linked (even if there were warnings and/or
$ errors) when it was finished. We re-wrote it, but it is a pain none the less.
$ Our earlier version of the OS came with F77. This one doesn't. IBM says
$ we have to purchase it now. The VS/FORTRAN does have its saving graces.
$ VAX mode (although buggy) has drastically reduced the time it took to port
$ some applications from the VAX.
No comment as we don't use Fortran.
$ The archive/library program is buggy and won't behave the same way two times
$ in a row. This is a particularly nasty problem because Make can't handle
$ an excessive (again I'm not prepared to cite numbers) number of file names
$ to link. It just crashes.
I've never seen this occur. Perhaps an example?
$ There are no man pages, and I have yet to get the 'help' command to yeild
$ anything resembleing an informative answer. Mostly all I get is a '... not
$ found' message.
Looking through the Commands Reference manual I saw a page on 'man'. At
the bottom of the page there was a note that man was available
optionally. I checked with our IBM rep and:
IBM AIX Online Publications RPQP91026 $50
They cover the contents of the Commands Reference and Tech. Ref.
$ UUCP is broken. Anything I try to run says that it wants some earlier
$ version of the VRM. A lot of files referenced in the manual aren't even
$ there. I *know* it's installed properly because I went back and did it again.
Strange, I've never had any problem with UUCP on the RT. I've always
thought that it ran quite solid.
$ Nroff flags that are in the manual do nothing. (Again I'll bring examples
$ next week).
$
$ DUMP is described wrong in the manual. We had to use some unusual magic
$ cookies on the command line to restore our file systems after 2.2.1
$ scrambled it. We now avoid the problem by giving dump different flags. I
$ think part of our problem may have been our ignorance.
$
$ There are several functions in the BSD compatability libraries that do
$ not perform in the manner described in the manual pages. There are other
$ routines that I suspect are just stubs.
$
$ In the same room with the RT is a VAX 11/730 running MORE/BSD 4.3. The RT
$ runs circles (many many circles) around that machine speed wise, but our
$ programmers claim they get better productivity from the 730 even though it's
$ slow. BSD came with C, PASCAL, Fortran, Ingres, and a host of tools that make
$ life easy. AIX requires that we purchase each of these, yet is more expensive
$ for the basic system than my entire BSD distibution. The BSD versions of
$ the compilers all work and work well. The IBM compilers are buggy and hard
$ to use.
You can't expect IBM to do business the same way a University does. Plus
programmers at IBM make far more than graduate students do.
$ I say that the AIX system is unproductive because there is a learning curve,
$ even for people who know sysV. We have two other sysV machines that we
$ do administrative/development work on. We should have been able to come
$ up to speed much quicker than we were able to. I don't beleive that people
$ should have to program/work around pitfalls in a machines architecture and
$ operating system. Especially since it costs money every time we make a
$ port to a new platform.
I came to the RT from a SUN and Ultrix world and it only took about two
days to get used to some of the AIX/SysV differences. It took me the
longest to break myself of the 'more' habit and type 'pg'.
$ There is a list of problems sitting on my desk. I will document them better
$ and email them to anyone who's interested.
$
$ Bye,
$
$ Peter Blemel
Sounds like you've had some problems. But I'd say don't give up as it
looks like you're more the exception than the rule.
John H. Lawitzke UUCP: Work: ...rutgers!mailrus!frith!fciiho!jhl
Michigan Farm Bureau ...decvax!purdue!mailrus!frith!fciiho!jhl
Insurance Group ...uunet!frith!fciiho!jhl
Home: ...uunet!frith!fciiho!ipecac!jhlroot@blender.UUCP (Super user) (03/01/89)
In article <2036@nmtsun.nmt.edu>, peter@hydrovax.nmt.edu writes: > The asynch driver is still broken. Terminals have to be run at 19200 or > AIX throws away output characters. I have tried numerous flow control > and bit/parity combinations, to no avail. I would really like to hang > a 2400 baud modem on the system so I can call in from school, but the > system (getty) won't send nl at 2400. It also locks up after I reply to the > login prompt. IBM sent out a tech. He couldn't figure it out either, > but verified that the systems at IBM (Albuquerque) have the same problem. > Of course this may be a hardware bug and BSD might behave the same. > While I have nothing to say about the other bugs you've mentioned since I haven't experienced them or even worked in the environment to have the opportunity to experience them, I do have something to say about the above. We have about 12-15 clients with various models of the RT (ie:25, 125, 115, 135) as well as a model 25 (gag) at work. At least one of each machine is running 2.2.1 and these machines are running mostly IBM3151 terminals but each one has at least one modem on it. About a half and half split between 1200 baud and 2400 baud. They work fine and dandy. Since we do our support remotely over these modems they're used quite often and we've never experienced any troubles with them aside from Operator errors. "What? It's supposed to be plugged into 'line'? Not 'phone'?". In fact we do a fair bit of uploading and downloading either via straight ascii or an xmodem type thing. We've never experienced lost characters. The settings I use (or the relevant ones anyways) in devices are the following: rts 2400 (1200) pt none bpc 8 pro dtr dvam 0 (I know it says "0 for local, 1 for modem" but who cares?) tt ibm3151 (or wy50, vt100 (procomm plus)) Those are the only things I change when I hook up the modem. Once I've got the modem hooked up I generally run up ATE and change the configuration (hayes compat) to work with the RT. Otherwise it won't work too well... ie: "ATE0M0S0=1&C1Q1&W" I think the operative one there is "E0" since if the modem is echoing back to the RT, and the RT is echoing back to the modem, and the modem is echoing back to the machine..... well, you know what I mean... On another note, Some of the machines have AT's or clones hooked up to them running Procomm+ at 9600 baud and they work fine too. Forgive me if some or all of what I've told you here is intuitively obvious but it's what I've been doing for every one of our clients machines and they work fine. (I won't say flawlessly.) I would however like to see BSD run on an RT. Especially a 135 and XWindows on a Megapel. That would seem impressive since it's something I've never seen. For those of you who work with this everyday I say PPPPPHHHHHTTTTT!!! -- herb@blender.UUCP Did I tell you I'm forgetful?
njs@scifi.UUCP (Nicholas J. Simicich) (03/02/89)
In article <2036@nmtsun.nmt.edu> peter@hydrovax.nmt.edu writes: (.....) >Our upgrade to 2.2.1 failed miserably due to several bad disks. It was >nice of IBM to come out and rectify the situation, but we still had >down time due to it. The version that they DID install would not execute >shell scripts or binaries properly because for some reason it sets the >current working directory to '/' (yes root). Make fails because my files >are not in '/'. I have very recently installed a new version (more disks >borrowed from IBM). It seems to be ok now. When Make *could* find the files >that I was compiling, it didn't work on any imbedded commands. It would >either fail without explanation or give an error number that wasn't in the >manual. Hmmm...what is your default shell? I've seen this when people try to do Makes and don't specify SHELL=.... Probably, though, this is something that used to work, and it is a bug. (.....) >The system doesn't provide job control or sockets, which is more a drawback >of sysV than AIX specifically. The console mimics job control using the >utilities found in the Useability Services, but it is a poor substitute >and only good for the person using console. As someone else pointed out, sockets are part of AIX. Most of the TCP/IP user programs are socket based these days, many run under inetd, and so forth. >VS/FORTRAN has bugs. Again I do not want to cite an example until I can >bring verified source code to post. I haven't looked under 2.2.1, but the >fortran compiler has a poorly written shell script for a front end. It's the >three pass variety and ALWAYS linked (even if there were warnings and/or >errors) when it was finished. We re-wrote it, but it is a pain none the less. >Our earlier version of the OS came with F77. This one doesn't. IBM says >we have to purchase it now. The VS/FORTRAN does have its saving graces. >VAX mode (although buggy) has drastically reduced the time it took to port >some applications from the VAX. Under AIX, as far as I know, Fortran was always extra cost. We're glad to hear that it has bugs, as if it didn't, software engineers the world over would have to rethink their assumptions. :-) I can only encourage you to contact the defect support people and complain. It is certain that you will get more attention than I would. (.....) >There are no man pages, and I have yet to get the 'help' command to yeild >anything resembleing an informative answer. Mostly all I get is a '... not >found' message. Man pages cost $50/machine extra, from what I remember. Contact your salesman. >UUCP is broken. Anything I try to run says that it wants some earlier >version of the VRM. A lot of files referenced in the manual aren't even >there. I *know* it's installed properly because I went back and did it again. In 2.2, the old UUCP was replaced with HoneyDanBer UUCP. If you have an old manual, well, lots of the files (L.sys) won't be there. I had a real early copy of 2.2, and got HDB UUCP working by using the Nutshell handbooks. (.....) >In the same room with the RT is a VAX 11/730 running MORE/BSD 4.3. The RT >runs circles (many many circles) around that machine speed wise, but our >programmers claim they get better productivity from the 730 even though it's >slow. BSD came with C, PASCAL, Fortran, Ingres, and a host of tools that make >life easy. AIX requires that we purchase each of these, yet is more expensive >for the basic system than my entire BSD distibution. The BSD versions of >the compilers all work and work well. The IBM compilers are buggy and hard >to use. There are two C compilers for AIX. Portable C is free, the other C compiler costs money. I do everything with Portable C, personally. (.....) >There is a list of problems sitting on my desk. I will document them better >and email them to anyone who's interested. I encourage you to call defect support with information about these problems, where the problems are defects. Where they are not defects, but design problems, I encourage you to report them to your branch folks, so that they will know what you need, and can feed your requirements back into the process. -- Nick Simicich --- uunet!bywater!scifi!njs --- njs@ibm.com (Internet)