Thanks Ricky - yes, it seems that the new files have ".new", and if something fails with  the "old" (which are unchanged) , you have to review the '.new" and see if there anre any substantive, or structural or functioal changes, and include them to the old (or copy the custimization from old to ".new,", and then rename the ".new" to lose the extension. Is that the general idea?

A spare machine is always a good idea. I have three, for self, wife and kid, all networked, and all sharing files and printers under both Linux, and that other, well-known popular OS. A fourth one to play with, a true "sandbox" would be great.
George

Ricky - thank you. You have always been very helpful. I tried to be as descriptive as possible, since I feel that an OS upgrade is not a trivial task (where I just was at a large internal IBM application (at IBM itself), they are planning for almost two years, just to go to the next release of the RDBMS (and not even the OS). I must add that that it is a very major new release, however).
Anyway, you say:
"Now what you should to is simply upgrade, then check what things are not functioning properly, then try to replace the old config file of that stuff with the new one ie. your old one. if then it still not works, then match the config parameters of the both files and do the changes as required."
Let me just ask one question<g>. An upgrade will lay down  his own (new) config files with the ".new" extension. So the existing configuration files will stay in place; they are not moved or renamed (of course we are encouraged to back them up first, and I have done that). That was my understanding, that the new ones are called ".new", and the old ones remain, and you might have to include changes from the old to the ".new", which will later need  to have the ".new" extension removed, of course. Patrick V supplies a script in item 6 of UPGRADE.TXT to help with this process.

You say that S/Ware upgrade will "rename the old one "
So, I'm afraid I am confused on this point. Are the existing configuration files renamed or not? If so, to what names?

Let's assume we can clear this matter up.

The newly installed config files  will function correctly, or they will not, of course. If not "And what changes are required there ? The answer is simple and already answered "?
That is the $64,000 question of course, and sadly, does not appear so simple to this Linux newbie. For example, the most obscure customization I had to do was to mimes.conv and mime.types in /etc/cups. Those two took me weeks to discover, just to fianlly get printing working! Who could know such an obscure thing except a guru?, to enable raw print support in an obscure CUPS file (not even in cupsd.conf)? Now I do not even know if these files are changed (clobbered) by the upgrade (technically, they are not directly under /etc., but under /etc/cups; they not are they called ".conf", but they are certainly "configuration" files) If they are clobbered, it's only by a miracle of God that I would remember to change them. The same would be true of printers.conf, the ppd's, /etc/samba/smb.conf, etc., etc......and a multitude of customizations that we have all done, over the years to get our systems to whwere we want them to be.

But it is possible I am missing something? Maybe he does not touch these files. The thing is, I do not know what is changed, what is restored by the upgrade to "ground zero", so to speak (I realize this is a very mixed metaphor). All I do know is what is written:
"Fix your config files.  Some of the config files in /etc are going to need your attention." from UPGRADE.TXT
To this newbie, this is less than enlightening documentation.

Ricky, forget about my 25 years, I'll change my signature, I appreciate all your replies to my questions!
George

Hi group.
I am about  to upgrade Slacware from 9.1 to 10.0, and am pretty confused about  what Patrick has written in UPGRADE.TXT , item 6, re "fix your config files"...that the new ones come in with extesion of .new, the old are saved as .bak. and that you "may need to fill these in with information from your old config files and move them over".

Similarly, another paper on this process is along the same lines:

"Merge the new configuration files with your old ones. The newly installed configuration files will have the file extension ".new" Most, if not
all, of the new configuration files will be located in your /etc directory. You can generate a list of these files
like so:
# find /etc -name \*.new
Fold any changes you made to the previous versions of the
configuration files into the new files, as necessary. Then
rename the new configuration file to the name of the old
configuration file (the same name, without the .new extension, conveniently.)"

What does it mean to "fold any changes you made to the previous versions of the configuration files into the new files, as necessary".

How does one know which files have changed over the lifetime of an installation? How can one know this? Does one simply copy the old (.bak) over the new (.new)? It doesn't sound like it. Or identify the private customizations one has made over the year(s), and re-modify the .new's, by hand, one by one and then rename thqam back? This is clearly impossible. Is that what is being said? i.e. that the structure of the new .conf's is unchanged and we just copy the (customized) backups and overlay the .new's? Or the structure is changed, and we ned to "remodify" all the customizations made in the past?
Maybe so! I just don't know.
Forgive me for this possibly moronic question, but while I am a DB guy, I have been a Systems Programmer on and off over 25 years on and off on some of the world's largest and most complex computers (Honeywell, IBM mainframes) and I don't have a clue what is being said here, with the all-important config files. Surely a user cannot "lose" all his changes and customizations, and have to identify them and re-insert them into the . new config files? Take CUPS, for example. I had to make changes long ago to files with  names like mimes.something...conv, types, to enable raw octets; I don't know and cannot remember.  All I know is that it took me weeks to find out that it had to be done, and I made the changes. Will these files be changed back to some pristine state? And I have to find them again, find the changes, and "fill these in ....."
What am I missing here?
Thanks
George

[very]

Hi Group,
I have gotten self into a bit of a mess, and would very greatly appreciate some help. I am finding that small upgrades can lead to spending dozens of hours, and ending up in a worse spot that by doing nothing. Right now, I have no windowing functionality in Linux at all.

Anyway, I decided to upgrade my KDE to the current  release (3.4.1) on my Slackware 9.1 system. This seemed innocent enough; I d/loaded the .tgz files and installed with upgradepkg.
The install messages were fine, but startx failed, could not start kdeinit, and the reason was inability to load shared file libidn.so.11. Seeking help on the web (from another machine's browser), others had encountered this, and recommended install of another package libidn (GNU Int. Domain Name lib), and that is what I did.
New message from startx : couldn’t find libXerama.so.1. More help on the web advised  to upgrade X itself. I dutifully downloaded the XFree86 4.5.0 binaries, ensuring that I had  right binary distribution: it comes with an aid “Xinstall.sh -check” which I ran, and was told I must download  Linux-ix86-glibc23 . Again, I did this, installing all 11 mandatory files, and a few of the optional ones (of which there are 13).
Again the install messages (from X) were fine, but startx failed, could not start kdeinit, and the reasons appear in two categories  
a) /lib/libc.so.6 version  ‘GLIBC_2.3.4’ not found, (required by /opt/kde/lib/libkdecore.so.4), and a second category of message:
b)couldn’t find init font path element /usr/X11R6/lib/X11/fonts, removing from list, and later, and presumably related:
xset: bad path element (#96), removing from list; possible causes are
i)   Dir doesn’t exist or wrong permission
ii)   Missing fonts.dir
iii)   Incorrect font server address or syntax

usr/X11R6/lib/X11/fonts does exist  rw- r--r--,  and has a fonts.dir of 2 bytes wih a value of 0 therein.

So that is where I am at. Clearly I need to do something with glibc, but I would appreciate guidance. I don’t want to keep endlessly installing and endlessly failing, and getting into ever deeper waters.

And I am mystified by the fonts business, whether it is even from an optional X package (like Xfsrv, in which case perhaps it can be uninstalled), or a mandatory one.
I realize I have to attack this piece by piece, and would greatly appreciate help.
Thanks
George

Ricky - again thanks. No you are not confusing me! Please bear with me in this message...it's not as complex as it is long!!! In fact, the question is probably quite simple.

"What you need to do is install other OS, make entry to your current lilo.conf file for the Kernel image for new OS and then update your lilo".

This is what I thought..one LILO.CONF, with two (or more) image entries. Then yes, run LILO, and copy the bootsector image back to C:\.

Here is current fdisk - l:
Device Boot      Start         End      Blocks   Id  System
/dev/hda1               1           4       32098+  de  Dell Utility
/dev/hda2   *           5        1314    10522575    7  HPFS/NTFS
/dev/hda3            1315        9732    67617585    f  W95 Ext'd (LBA)
/dev/hda5            1315        7163    46982061    7  HPFS/NTFS
/dev/hda6            7164        9712    20474811   83  Linux
/dev/hda7            9713        9732      160618+  82  Linux swap

So, I will free some space and create a new ext2 partition (with PartitionMagic, or whatever) and this will probably be called /dev/hda8,  install the new LINUX there, and modify the existing LILO.conf  (currently on /dev/hda6) to include this new image.

##Proposed  new lilo.conf with two LINUX's on separate partitions
boot = /dev/hda6
compact      
prompt
timeout = 50
# Linux bootable partition config begins
  image = /boot/vmlinuz
  label = SWLinux2.4.22
  root = /dev/hda6
  read-only # Non-UMSDOS filesystems should be mounted read-only for checking
# Linux bootable partition config ends
#Here will be the new LINUX
image = /boot/vmlinuz.new
root = /dev/hda8   <=================new LINUX partition
 label= RedHatLinux2.4.22
 read-only

So far so good? Is this right?

Now, here is the catch. If this setup works fine, my goal is to remove the "old" Linux on /dev/hda6, reclaim the space, and leave only one Linux, the "new" one (RedHat). So the /dev/hda6 partition will disappear.

I guess I would then have to do this:
Boot to the "new Linux" (the second one in the lilo.conf file, above), and from within there, create a new lilo.conf with only one (1) entry, like:

#Proposed  new lilo.conf after removeal of old LINUX partition
boot = /dev/hda8   <===========the new LINUX home
compact      
prompt
timeout = 50
# Linux bootable partition config begins
  image = /boot/vmlinuz
  label = RedHatLinux2.4.22
  root = /dev/hda8   <=================as above
  read-only # Non-UMSDOS filesystems should be mounted read-only for checking
# Linux bootable partition config ends

Then run lilo, copy the bootsector to floppy, and transfer it to C:\

Ricky, please let me know if this approach seems reasonable.

There is a reason for all this, which started the  thread - Slackware is not "validated" by IBM's DB2 DBMS, and so I have had enormous trouble with RPM-based package install and upgrades (RedHat is both RPM-based, and a "validated" distro by IBM).
Finally, the reason for the two-step approach is simple conservatism:  I do now have a working DB2 UDB/Slackware Linux, and do not want to just trash it,  and try a brand new LINUX. This way, I can "fall back"  - if necessary - to what I already have, if the new install causes problems.

Finally, (again), there is another way, right......? Install the new RedHat on the new partition /dev/hda8, let the RedHat install procedure create a Lilo there with the correct parms (I am guessing that he will do that!!); and I then manually  add the "old" Slackware Linux as the additional image to that lilo.conf, and run lilo  (and then copy /dev/hda8 bootsect to C:\).

This way, I boot directly to the new LILO on /dev/hda8, with the option to launch the "old" SWare Linux on /dev/hda6?

It seems in both cases, I must install RedHat, modify lilo.conf (on one or the other LINUX's) , run lilo, copy bootsect  to floppy, and then to C:\.

Perhaps it's a case of "6 of one, half a dozen of the other".

What do you think? I guess the main question is: is my logic on the right track?

Thanks
George (and apologies for the length)

Rohit
Ok, but how then to interpret  the messgae:
"May 30 18:26:07 george inetd[5614]: auth/tcp (2): bind: Address already in use"
Which address, exactly? The port (51000)? Certainly that port is NOT in use as shown by NETSTAT. Perhaps it is referring to something else? I am at a loss...I am more familiar with the IBM world, where each and every message(for all sub-systems, products, etc.)  has an identifier, and one can go off to the appropriate Error Messages Manual, with suggested remedies. They might vary in usefulness, of course, but this message from inetd directs me nowhere!
Thanks
George

Thanks rohitb
What is so interesting is that the port reserved for this daemon is not in use, as shown by NETSTAT, not is it used by another procees in /etc/services. I am still mystified by the precise meaning of the SYSLOG message.
THanks
George

Hi group
I am trying to run a daemon which uses a port reserved for it in /etc/services, and after several minutes, it fails, with:

/opt/IBM/db2/V8.1/bin$ db2icd start
"Initializing startup.
a few minutes passes ....
Unable to start DB2 Information center daemon."

SYSLOG indicates the following:

May 30 18:26:07 george inetd[5614]: auth/tcp (2): bind: Address already in use
May 30 18:26:07 george inetd[5614]: finger/tcp (2): bind: Address already in use
May 30 18:26:07 george inetd[5614]: ftp/tcp (2): bind: Address already in use
May 30 18:26:07 george inetd[5614]: time/tcp (2): bind: Address already in use

Am running Slackware LINUX with 2.4.22 kernel

I am presuming the message means that the port is in use (by another process). How can I found out who is currently using which ports, please.
Thanks
George Peters

What I think .. that when you will install other linux .. it will not overwrite the MBR as you can choose it to be resided on first sector of the root partition of that linux. ..

So you have to also create a file of bootsector of new linux then again you have to add entry for that in boot.ini

and really Thank you! for your second brief post.. however we always encourage people to be more descriptive as they can

Ricky - many thanks. Is this saying that on the new LINUX partition I will have a new LILO, and its own boot sector to copy to WIN boot.ini as a third boot option?  (1=WIN XP, 2=Slackware Linux, 3=RedHat etc)? I was thinking that if LILO is a multi-boot loader, I could somehow get the new LINUX stanza in the existing lilo.conf (or, alternatively, have the new LINUX create his own LILO, and I would add the existing Slackware stanza to its .conf), and so when LILO comes up, have him presnt me with two LINUX OS's to choose from. Perhaps my understanding of LILO is faulty?
George

What will be (was?)  the result of necropolois' install of his second LINUX (BSD 5.1) into a separate LINUX partition, given that he will already have a  Linux Bootloader already  in place, from the Debian install (am not familar with that bootloader) ? Will the second  install accomodate itself to the existing Debian bootloader in some way?
I have posted a similar question (but mrde long-winded) elsewhere here recently.
Thanks

I apologize for this cross-posting (and the wordiness), but this  posting probably belongs in this thread:

I have a question re: chained multi-Linux boot loading, starting from WIN NT BootLoader, and here is some background. I am an old-timer, but not a UNIX/LINUX expert....

Using advice given here, last year I created LINUX partition on a WIN XP machine (perhaps with Partition Magic, it's not
important), installed SLACKWARE 9, and WIN now offers me dual boot options on startup, the second option being to
boot Linux. This second option derives from me making a manual entry to WIN boot.ini, pointing to a small file containing a copy of the Linux Boot Sector on the Linux Boot Partition (/dev/hdfa6, as it happens). This Sector was copied from the
Linux Boot partition with:
dd if=/dev/hda6 bs=512 count=1 of=bootsect.lnx, as documented by Ricky, and I think "aaa" in this Forum. It works
perfectly, with WIN NT Bootloader enabling the LILO option, which currently has only one OS defined , booting by default (or ENTER key) to the single SLACKWARE LINUX. {For those new to this world, and wanting to keep WIN in place as I was, I strongly recommend following this procedure, by the way}.

Now, I need to install a second LINUX, while maintaining the current dual boot process that is in place. (Very simply, I am working with IBM's DB2 UDB V8.2, and its install scripts etc. assume "standard" RPM-based package management, and SLACKWARE's seems to be "non-standard" (indeed, it is not RPM based, but it does support the RPM command...imperfectly...This has been the subject of postings and inquiries, and even an IBM DB2 DeveloperWorks article, but I have found that they are
incomplete. For those who are interested, the only way around is to convert all the .rpm packages to native format .tar.gz
(or .tgz) with a SLACKWARE utility, and install the packages manually ("installpkg") , thereby bypassing the entire IBM
install procedure, which I am not too pleased about. This has caused me unending pain, a great learning curve...and a
desire to be more "standard"). I have become the package management expert I never wanted to be!

So I have acquired another LINUX (REDHAT Professional Workstation, which was the only REDHAT in the store!).
I have huge book by Sobell, and I have all the RedHat manuals, but I am speculating that a) they will probably assume
only one LINUX (let alone a pre-existing dual-boot setup) and b) regardless, a group such as this has more real-world
experience. I figure my question cannot be new, or difficult.

I am more than happy to keep LILO.
What is the best way to go about this install please? And to keep the current setup in place?
In short, to install another LINUX, on a presumed new partition (again, Partition Magic, whatever), and have it made known to LILO ? Or, perhaps a new REDHAT install might create a new LILO (and .conf) and put himself there (and perhaps lose what I have)? I don't know. Maybe it doesn't matter. So long as one can modify the lilo.conf correctly, execute lilo, and then copy the boot sector to WIN as described above? That's OK too, so long as I can continue boot to LILO from WIN (as I do now), and identify what will now be two (2) LINUX OS's. I am just not familiar with these multiple levels of indirection on OS's on which am no expert. I am reluctant to meddle too much, since I now do have something that works, and you know what they say ("If it aint broke ......)
IBM's DB2 is my livelihood, on MVS, OS/2, WIN and now flavors of UNIX/LINUX, so this is not a hobby!

Perhaps I was too wordy above.  In a nutshell, I want to add a second installation of another LINUX OS, which is currently managed by LILO (SLACKWARE), which itself is launched from the dual boot feature of WIN NT (XP Pro) bootloader.
Is there anything "special" I must do?
Thanks
George

Hi Group
I have a question re: chained multi-Linux boot loading, starting from WIN NT BootLoader, and here is some background.  I am an old-timer, but not a UNIX/LINUX expert....

Using advice given here, last year I created LINUX partition on a WIN XP machine (perhaps with Partition Magic, it's not  
important), installed SLACKWARE 9, and WIN now offers me dual boot options on startup, the second option being to  
boot Linux. This second option derives from me making a manual entry to WIN boot.ini, pointing to a small file containing a copy of the Linux Boot Sector on the Linux Boot Partition (/dev/hdfa6, as it happens). This Sector was copied  from the  
Linux Boot partition with:
dd if=/dev/hda6 bs=512 count=1 of=bootsect.lnx, as documented by Ricky, and I think "aaa" in this Forum. It works  
perfectly, with WIN NT Bootloader enabling  the LILO option, which currently has only one OS defined , booting by default (or ENTER key) to the single SLACKWARE LINUX. {For those new to this world, and wanting to keep WIN in place as I was, I strongly recommend following this procedure, by the way}.

Now, I need to install a second LINUX, while maintaining the current dual boot process that is in place. (Very simply, I am working with IBM's DB2 UDB V8.2, and its install  scripts etc. assume "standard" RPM-based package management, and SLACKWARE's seems to be "non-standard" (indeed, it is not RPM based, but it does support the RPM command...imperfectly...This has been the subject of postings and inquiries, and even an IBM DB2 DeveloperWorks article, but I have found that they are  
incomplete. For those who are interested, the only way around is to convert all the .rpm packages to native format .tar.gz  
(or .tgz)  with a SLACKWARE utility, and install the packages manually ("installpkg") , thereby bypassing the entire IBM  
install procedure, which I am not too pleased about. This has caused me unending pain,  a great learning curve...and a  
desire to be more "standard").  I have become the package management expert I never wanted to be!

So I have acquired  another LINUX (REDHAT Professional Workstation, which was the only REDHAT in the store!).
I have huge book by Sobell, and I have all the RedHat manuals, but I am speculating that  a) they will probably assume  
only one LINUX (let alone a pre-existing dual-boot setup) and b) regardless, a group such as this has more real-world  
experience. I figure my question cannot be new, or difficult.

I am more than happy to keep LILO.
What is the best way to go about this install please? And to keep the current setup in place?
In short,  to install another LINUX, on a presumed new partition (again, Partition Magic, whatever), and have it made known to  LILO ? Or, perhaps a new REDHAT install might create a new LILO (and .conf) and put himself there (and perhaps lose what I have)? I don't know. Maybe it doesn't matter. So long as one can modify the lilo.conf correctly, execute lilo, and then copy the boot sector to WIN as described above? That's OK too, so long as I can continue  boot to LILO from WIN (as I do now), and identify what will now be two (2) LINUX OS's. I am just not familiar with these multiple levels of indirection on OS's on which am no expert. I am reluctant to meddle too much, since I now do have something that works, and you know what they say ("If it aint broke ......)
IBM's DB2 is my livelihood, on MVS, OS/2, WIN and now flavors of UNIX/LINUX, so this is not a hobby!
Many Thanks
George Peters

Ricky
You wrote:
" After copying start windoz and write the following line to boot.ini found on C:\ of windoz." (referring to the C:\bootsect.lnx= ......line)

Are you saying that we over-write the entire boot.ini file with this single line? Or that we include it in the existing file?
Thanks
George

   Recommend LS
   Linux User Around