On Thu, 31 Mar 2022 13:35:34 +1100, faeychild wrote:
> On 31/3/22 04:04, David W. Hodgins wrote:
>
>
>> Upgrading from release to release+1 is tested to the best of our
>> ability. We can't
>> test every combination of hardware, partitioning choices, software
>> combinations,
>> but do the best we can. Third party software may also cause problems.
>
> Yep
>
> Each release invites me to install my printer/scanner driver
> I know that I should have taken notes but by the time I've banged my
> head on the keyboard all evening I am just glad to have it FINALLY working
>
> And, of course, I'll remember next time, wont I?
>
> AND AND the previous time for M8 I decided to network the printer
>
> Got real close to the edge of madness there.

IIRC you have a Brother MFC-9340CDW

To create an install script you just create/paste the commands you
enter at the terminal.

It should be no problem for you to write a script to do 90% of the install.
Use wget to pull down the Brother install script, un zip it, install the
Mageia scanner rpms, run the Brother printer script start cups and httpd,
and use xmessage to pop up cups commands for printer setup in cups admin,
add printer ip to /etc/hosts.

FYI, you can use 'expect' to answer Brother console questions to further
automate the process.

I can post my expect script if you like. Your script would have
autoexpect bash linux-brprinter-installer $_printer
which would automagically answer all the installer questions.

Any time you want to automate your install you can start a new thread
and I can help you code a script.

On 1/4/22 01:50, TJ wrote:

> Interesting. I have three printers, all HP. The oldest is a Deskjet
> 5650, almost 20 years old, but it still works. Newest is a 2018 Envy
> Photo 7858, purchased as refurbished to replace an Officejet that went
> belly up. I needed the scanner, not the printer, but all-in-ones are a
> LOT easier to find, and cheaper, than stand-alone scanners. In between
> is an old color Laserjet, a gift my nephew rescued when it was cast off
> by the office where he worked.
>
> Easy to install with MCC, only a few minutes each. The newest one is the
> only one that's wireless, so I used it to test the networking setup of
> the recent hplip update. The printer set itself up with my router, so
> MCC again had no problems configuring it as a network printer on a
> couple of my other computers.
>
> TJ

I suspect it's me

Brother provides a script which downloads and installs all the drivers.

This all seems terrific until the last bit when the script will ask if a
test print is required.

This doesn't work because the printer is yet to be installed.
This is silly and irritating

The printer is then installed with MCC or cups, usually OK.

The scanner must not be installed by MCC or everything is fubar.
The scanner must be installed by xsane or it does for me.

My printer installs with MCC demands such information as "Device URI"

Which I discovered to be "socket://192.168.20.2:9100" This is an IP
address and a port. So why call it "URI"

Something else that you were clearly expected to know beforehand.

This does not result in a calm experience. But maybe I am doing it backwards

Fortunately I have taken notes and I expect Mageia 9 to go a little
better. Like all thing, it's easy when you know how.

--
faeychild
Running plasmashell 5.20.4 on 5.15.32-desktop-1.mga8 kernel.
Mageia release 8 (Official) for x86_64 installed via Mageia-8-x86_64-DVD.iso

On 1/4/22 02:44, Bit Twister wrote:

>
> I can post my expect script if you like. Your script would have
> autoexpect bash linux-brprinter-installer $_printer
> which would automagically answer all the installer questions.
>
> Any time you want to automate your install you can start a new thread
> and I can help you code a script.

It presently "aint broke" and it wont change until Mageia 9.
Then we may be up for it again. But I have my notes and the magic URI
entry. Which is just an IP address and port number: but if you don't
know it, you're stuffed. Maybe it's fully automated and can be left
blank. I don't think I ever tried that,

I just sat there and yelled at MCC " What's a fscking URI you son of a
bitch"

One day when I'm feeling invincible I will trash the printer install and
play about with it. A learning experience. But not to be taken on during
the heat of a new release

regards

--
faeychild
Running plasmashell 5.20.4 on 5.15.32-desktop-1.mga8 kernel.
Mageia release 8 (Official) for x86_64 installed via Mageia-8-x86_64-DVD.iso

On 2022-03-31, faeychild <faeychild@nomail.afraid.org> wrote:
> On 1/4/22 02:44, Bit Twister wrote:
>
>>
>> I can post my expect script if you like. Your script would have
>> autoexpect bash linux-brprinter-installer $_printer
>> which would automagically answer all the installer questions.
>>
>> Any time you want to automate your install you can start a new thread
>> and I can help you code a script.
>
>
>
> It presently "aint broke" and it wont change until Mageia 9.
> Then we may be up for it again. But I have my notes and the magic URI
> entry. Which is just an IP address and port number: but if you don't
> know it, you're stuffed. Maybe it's fully automated and can be left
> blank. I don't think I ever tried that,
>
> I just sat there and yelled at MCC " What's a fscking URI you son of a
> bitch"
>
> One day when I'm feeling invincible I will trash the printer install and
> play about with it. A learning experience. But not to be taken on during
> the heat of a new release

Which is clearly why you should get the notes and script up and running
well before the heat of a new release arrives.

>
> regards
>

On 2022-03-30, William Unruh <unruh@invalid.ca> wrote:
>
> c) Update by transfering a Mga8 /etc/urpmi/urpmi.cfg in, running
> urpmi.update -a, and then urpmi --auto-select
> Advantage: Again preserving configurations. But again the worry that
> some names have changed sufficiently to not be updated (eg gimp
> renamed to gimp2 and thus the update is invisible to the system)
> It can be done remotely.

OK, I did c) and the system ran, until I rebooted. Then I got dumped
into the systemd error bin. Of course no explanation of why. Just ^D to
continue ( which just got my error again) or give password for root, and
look at journalctl. Nothing there to indicate what it was that caused
the problem, except scattered through the boot log were errors saying
that various directories could not be mounted. They were all nfs, and
trying to ping I discovered that the network had not come up. Now, the
system was all on the local machine, so why in the world would inability
to mount nfs crash the system? I finally went into /etc/fstab, commented
out all the nfs entries, and sure enough the boot finished. Logged in as
root, and the network now seemed to be up (why would the system try to
mount nfs files when the network was not up?)

So now I have to figure out how to harden the system Commenting out all
the nfs mounts is not an option, since they are needed for that computer
to carry out its job. I am now putting in noauto option into those nfs
mounts, but of course that means after the reboot I will have to by hand
mount all of those nfs mounts one by one, since noauto means that mount
-a does not work for those mounts. (would a clean install have been
better? No. because I would then have to recreate the /etc/fstab file
from scratch.)

Why in the world is the system not smart enough to ignore errors in nfs
mounted files, or set up the system to finish its boot and then try to
run mount -a after the network has come up?

It is things like that that make upgrading the horror that it is, and
make people very gun-shy about upgrading.

On Thu, 31 Mar 2022 21:56:59 -0400, William Unruh <unruh@invalid.ca> wrote:
> root, and the network now seemed to be up (why would the system try to
> mount nfs files when the network was not up?)

Because the system has no way of knowing which mounts are critical from the
user's point of view.

Instead of commenting out the fstab entries, add the nofail option to every
entry that isn't critical. For example, I have a file system used for backups.
I don't want the system to drop to the recovery shell if it fails to mount, so
in fstab I have ...
LABEL=aback /aback ext4 defaults,noatime,nofail 1 2

It still gets mounted, but if it fails for any reason, it doesn't stop the boot.

And just fyi, this is not a new option. It was the same in Mageia 7. The only
difference is that now the network failed to come on that system.

Regards, Dave Hodgins

On 1/4/22 12:43, William Unruh wrote:

> Which is clearly why you should get the notes and script up and running
> well before the heat of a new release arrives.

The wisdom of hindsight and the arrogance of pigheaded all clash to the
determent of me.

I've done it before and I'll do it again - lazy and stupid

regards

--
faeychild
Running plasmashell 5.20.4 on 5.15.32-desktop-1.mga8 kernel.
Mageia release 8 (Official) for x86_64 installed via Mageia-8-x86_64-DVD.iso

On 31/3/22 01:42, Aragorn wrote:

>
> Timeshift.
>

I've had a quick look, Aragorn

Does it run independently or does it backup the live system

should I read the man page :-)

--
faeychild
Running plasmashell 5.20.4 on 5.15.32-desktop-1.mga8 kernel.
Mageia release 8 (Official) for x86_64 installed via Mageia-8-x86_64-DVD.iso

On 2022-04-01, David W. Hodgins <dwhodgins@nomail.afraid.org> wrote:
> On Thu, 31 Mar 2022 21:56:59 -0400, William Unruh <unruh@invalid.ca> wrote:
>> root, and the network now seemed to be up (why would the system try to
>> mount nfs files when the network was not up?)
>
> Because the system has no way of knowing which mounts are critical from the
> user's point of view.
>
> Instead of commenting out the fstab entries, add the nofail option to every
> entry that isn't critical. For example, I have a file system used for backups.
> I don't want the system to drop to the recovery shell if it fails to mount, so
> in fstab I have ...
> LABEL=aback /aback ext4 defaults,noatime,nofail 1 2

Hm. Every mount had nofail. Unfortunately on looking more closely two of
the entries had
..... nofail,nfs async,rw,rsize=8192,wsize=8192,soft,bg,intr 0 0

and
..... nofail,nfs rw,soft 0 0

Not quite the right place to put that option I guess.

>
> It still gets mounted, but if it fails for any reason, it doesn't stop the boot.
>
> And just fyi, this is not a new option. It was the same in Mageia 7. The only
> difference is that now the network failed to come on that system.

Yes, which is why I had nofail already in the file. Doesn't help if you
do not put it in the right place.

Thanks for making me look again.

>
> Regards, Dave Hodgins

On Fri, 1 Apr 2022 01:56:59 -0000 (UTC), William Unruh wrote:
> On 2022-03-30, William Unruh <unruh@invalid.ca> wrote:
>>
>> c) Update by transfering a Mga8 /etc/urpmi/urpmi.cfg in, running
>> urpmi.update -a, and then urpmi --auto-select
>> Advantage: Again preserving configurations. But again the worry that
>> some names have changed sufficiently to not be updated (eg gimp
>> renamed to gimp2 and thus the update is invisible to the system)
>> It can be done remotely.
>
> OK, I did c) and the system ran, until I rebooted.

PS: I had a mysql update which clobbered my MythTV database.
My install_updates scripts now checks for any mysql updates and
stops mysqld if so.

> Then I got dumped
> into the systemd error bin. Of course no explanation of why. Just ^D to
> continue ( which just got my error again) or give password for root, and
> look at journalctl. Nothing there to indicate what it was that caused
> the problem, except scattered through the boot log were errors saying
> that various directories could not be mounted. They were all nfs, and
> trying to ping I discovered that the network had not come up.

Yep, been there done that, have the hat and T-shirt.

That is why having automated install scripts save time in the long run.
I have 4 major top level scripts that call all the other scripts.

At one time if the network failed systemd would consume a VERY large
amount of cpu trying to bring the network back up. It was so bad there
would be several seconds for each keyboard entry to echo at the root terminal.

I now run my ck_network script at the start and end of my top level scripts.
I also run in via hourly cron. Saves time wondering some network app is
no longer working.

That really came in handy when I was getting my network change scripts
working since I have configured my systems to use systemd-networkd service.

> Now, the
> system was all on the local machine, so why in the world would inability
> to mount nfs crash the system? I finally went into /etc/fstab, commented
> out all the nfs entries, and sure enough the boot finished. Logged in as
> root, and the network now seemed to be up (why would the system try to
> mount nfs files when the network was not up?)
>
> So now I have to figure out how to harden the system Commenting out all
> the nfs mounts is not an option, since they are needed for that computer
> to carry out its job. I am now putting in noauto option into those nfs
> mounts, but of course that means after the reboot I will have to by hand
> mount all of those nfs mounts one by one, since noauto means that mount
> -a does not work for those mounts. (would a clean install have been
> better? No. because I would then have to recreate the /etc/fstab file
> from scratch.)

Hehehe, yup. I have a fix_fstab script to change the uuid to label and
add all partitions with labels. I use lsblk to get desired information
to populate fstab. Look at this snippet

$ lsblk -o NAME,TYPE,FSTYPE,MOUNTPOINT,SIZE,FSAVAIL,FSUSED,LABEL,PARTLABEL

sda disk 931.5G
├─sda2 part ext4 / 42G 23.1G 15.8G mga8 mga8
├─sda3 part ext4 40.8G mga7 mga7
├─sda4 part ext4 40.4G cauldron cauldron

sdb disk 931.5G
├─sdb1 part swap [SWAP] 8G swap swap
├─sdb2 part ext4 20G bk_up bk_up

]$ grep -E "mga8|swap|bk_up|ga7|cauldron" /etc/fstab
LABEL=mga8 / ext4 relatime,acl 1 1
PARTLABEL=swap swap swap defaults,nofail 0 0
LABEL=cauldron /cauldron ext4 users,noauto,relatime,acl 1 2
LABEL=bk_up /bk_up ext4 users,noauto,relatime,acl 1 2
LABEL=cauldron_bkup /cauldron_bkup ext4 users,noauto,relatime,acl 1 2

I use the label as the mount point. For any lurkers I used gparted to
set LABEL and PARTLABEL. gparted will not let you change/set values if
partition is mounted. Helps to have a rescue cd handy to work on a system
that is or needs to be offline. I user one from https://www.system-rescue.org/

> Why in the world is the system not smart enough to ignore errors in nfs
> mounted files, or set up the system to finish its boot and then try to
> run mount -a after the network has come up?

Careful about you want/ask for. Sounds like you need a custom systemd service.

Target node may not be up and you would still fail to boot.
Better to have a ck_mount script to do the mount -a and check for failures.

> It is things like that that make upgrading the horror that it is, and
> make people very gun-shy about upgrading.

People doing upgrades better have a working fallback method just in case
the upgrade is no good or fails.

I do clean installs in a new partition. Worse comes to worse I can boot
previous release/install. Also is handy for comparing config files between
releases for changes.

On Fri, 1 Apr 2022 08:58:03 +1100, faeychild wrote:
> On 1/4/22 01:50, TJ wrote:
>

>
> Brother provides a script which downloads and installs all the drivers.

Yep, and lots of other stuff is also done. Spent a few hours getting
my install_printer to install all the Mageia packages for scanner and
whatnot, remove previous Brother install rpms and whatnot.
All that work is because all my scripts have a remove argument to remove
or reset back to day one condition.

I have the Advanced Intrusion Detection Environment​ (aide) rpm ]installed.
Run the aidecheck --init prior to install and aide.check --check afterwards
to find everything that changed.

> This all seems terrific until the last bit when the script will ask if a
> test print is required.
> This doesn't work because the printer is yet to be installed.
> This is silly and irritating

Hmmmm, the few times I forgot to answer N to print test page worked for me.

>
> The printer is then installed with MCC or cups, usually OK.
>
> The scanner must not be installed by MCC or everything is fubar.
> The scanner must be installed by xsane or it does for me.

I have
$ grep urpmi install_printer
x_urpmi simple-scan ! Simple scanning utility​
x_urpmi sane-frontends ! Graphical frontend to SANE​
x_urpmi gocr ! OCR Optical Character Recognition program​
x_urpmi xsane ! Frontend for the SANE scanner interface​
x_urpmi saned ! local and remote scanner, digital, video device access
x_urpmi xsane-gimp ! GIMP plug-in which provides the SANE scanner interface​
x_urpmi psutils ! PostScript utilities​
x_urpmi cups-common ! Common Unix Printing System - Common stuff​
x_urpmi cups ! Common Unix Printing System - Server package​
rpms installed.

FYI: x_urpmi is a wrapper around urpmi. It checks if rpm is already
installed and skips if so.

On 01.04.2022 at 14:21, faeychild scribbled:

> On 31/3/22 01:42, Aragorn wrote:
>
> > Timeshift.
>
> I've had a quick look, Aragorn
>
> Does it run independently or does it backup the live system

It can be used as a GTK-based GUI application from within the running
system, or as a command-line tool, and if it's present on the live
USB/DVD, then you can even restore your backups from there, in case
your installed system is completely b0rk3n.

You can opt to have it periodically run — for _making_ backups
only, of course — via cron, or to start it manually from within your
system menu (or from the command line). It is however polkit-aware and
will ask you for a password, so it runs with root privileges.

Mind you — and it's important to know this — the medium upon which it
must store its backups is not set in the configuration as a
directory/mountpoint, but as a _block device_ — e.g. /dev/sdb2. It will
automatically mount said block device (if needed) to
/run/timeshift/backup, even if it's already mounted somewhere else in
the tree. The backups themselves are stored under a directory called
"timeshift" in the root directory of the volume that it is set to back
up to.

> should I read the man page :-)

Always. :p

--
With respect,
= Aragorn

On 3/31/22 18:27, faeychild wrote:
> On 1/4/22 02:44, Bit Twister wrote:
>
>>
>> I can post my expect script if you like. Your script would have
>>          autoexpect bash linux-brprinter-installer $_printer
>> which would automagically answer all the installer questions.
>>
>> Any time you want to automate your install you can start a new thread
>> and I can help you code a script.
>
>
>
> It presently "aint broke" and it wont change until Mageia 9.
> Then we may be up for it again. But I have my notes and the magic URI
> entry. Which is just an IP address and port number: but if you don't
> know it, you're stuffed. Maybe it's fully automated and can be left
> blank. I don't think I ever tried that,
>
> I just sat there and yelled at MCC " What's a fscking URI you son of a
> bitch"
>
> One day when I'm feeling invincible I will trash the printer install and
> play about with it. A learning experience. But not to be taken on during
> the heat of a new release
>
> regards
>
Yelling at MCC never works. Trust me on this. Don't ask how I know, just
trust me.

More than you wanted to know about "URI:"

https://en.wikipedia.org/wiki/Uniform_Resource_Identifier

This was a month ago, but as I remember it the procedure I used for my
HP printer was this:

1) Used the printer's front panel menu/functions to establish a
connection to my network on my Linksys wireless router.
2) Booted into Mageia, ran MCC and from there asked to configure a
printer. This installed system-config-printer, which has
task-printing-hp as one of its dependencies.
3) When system-config-printer came up, clicked on "Add." The printer was
detected automatically. I clicked on it, then said to use the hplip
driver. Any needed information was gathered for me, and the system
installed both printer and scanner.

I used an alternative method on another machine where
system-config-printer and task-printing-hp had already been installed:

As root, I ran "hp-setup <network IP of the printer>" Again, I did not
have to provide any further information that I recall now, and both
printer and scanner were installed.

Easy-peasy.

TJ

On 1/4/22 16:05, Bit Twister wrote:

>
>> This all seems terrific until the last bit when the script will ask if a
>> test print is required.
>> This doesn't work because the printer is yet to be installed.
>> This is silly and irritating
>
> Hmmmm, the few times I forgot to answer N to print test page worked for me.

OK I am in error somewhere

And looking forward to mageia 9
Oh Wow

regards

--
faeychild
Running plasmashell 5.20.4 on 5.15.32-desktop-1.mga8 kernel.
Mageia release 8 (Official) for x86_64 installed via Mageia-8-x86_64-DVD.iso

On 2/4/22 00:24, TJ wrote:

> Yelling at MCC never works. Trust me on this. Don't ask how I know, just
> trust me.

I have noticed this. Is it hard of hearing or just uncaring
>
> More than you wanted to know about "URI:"
>
> https://en.wikipedia.org/wiki/Uniform_Resource_Identifier
>
> This was a month ago, but as I remember it the procedure I used for my
> HP printer was this:
>
> 1) Used the printer's front panel menu/functions to establish a
> connection to my network on my Linksys wireless router.
> 2) Booted into Mageia, ran MCC and from there asked to configure a
> printer. This installed system-config-printer, which has
> task-printing-hp as one of its dependencies.
> 3) When system-config-printer came up, clicked on "Add." The printer was
> detected automatically. I clicked on it, then said to use the hplip
> driver. Any needed information was gathered for me, and the system
> installed both printer and scanner.
>
> I used an alternative method on another machine where
> system-config-printer and task-printing-hp had already been installed:
>
> As root, I ran "hp-setup <network IP of the printer>" Again, I did not
> have to provide any further information that I recall now, and both
> printer and scanner were installed.

So you already knew of the secret guild code to run "hp-setup". You were
lucky. :-)
The hp task printing stuff must installed by default and I have Brother
printer

--
faeychild
Running plasmashell 5.20.4 on 5.15.32-desktop-1.mga8 kernel.
Mageia release 8 (Official) for x86_64 installed via Mageia-8-x86_64-DVD.iso

On 4/1/22 15:38, faeychild wrote:
> On 2/4/22 00:24, TJ wrote:
>
>> As root, I ran "hp-setup <network IP of the printer>" Again, I did not
>> have to provide any further information that I recall now, and both
>> printer and scanner were installed.
>
> So you already knew of the secret guild code to run "hp-setup". You were
> lucky. :-)

Well, no. I just did my homework first. DuckDuckGo is my friend.

> The hp task printing stuff must installed by default and I have  Brother
> printer
>
>
Yes, in Mageia, task-printing-hp is a dependency of
system-config-printer. The reason has something to do with
auto-detecting HP printers. See
https://bugs.mageia.org/show_bug.cgi?id=9902
for a lively discussion about whether this should be necessary if one
doesn't own an HP printer.

TJ

On 2022-03-30, David W. Hodgins <dwhodgins@nomail.afraid.org> wrote:
> On Wed, 30 Mar 2022 00:07:19 -0400, William Unruh <unruh@invalid.ca> wrote:
>> Thre are three ways I can think of upgrading say from Mga7 to 8.
>> a) Wipe 7 and Install 8.
><snip>
>> b) Upgrade in place from the cdrom/usb stick.
>> c) Update by transfering a Mga8 /etc/urpmi/urpmi.cfg in, running
>> urpmi.update -a, and then urpmi --auto-select
><snip>
>> What are the other disadvnatages of the three choices? Any advice?
>
> This is overlooking the most common method, initiating the upgrade using mgaapplet.
> It's like option c above, but done without having to manually alter the urpmi.cfg
> file.
>
> There's also "urpmi.removemedia -a", followed by adding the media. That can be done
> using urpmi.addmedia, as per
> https://wiki.mageia.org/en/How_to_choose_the_right_Mageia_upgrade_method#How_to_do_a_simple_upgrade_in_text_mode

That web page or the method has a lacuuna. I run the test. Each time I
do so, it downloads all of the files again (more than 1 hr on my Gbit
fibre cable) making the test a real pain to use. It also does not say
that it has run out of room. It just says that the package needs 27MB to
install. And I have 3GB on / partition. I finally figured out that this
meant it had run out of room on the test run.
Then, when I finally made enough room ( I had 17GB hidden under the
mount point of my /local) and it said I could install, it seemed again
to go ahead and download all of files that it just downloaded-- a real
waste of time. I tried looking at the options for uprmi but could not
see anything which sid to just use the cached files in the cache
directory /newlocal/rpms (that was where I had to put the cache since /
was too full).

Ie, that web page really needs some editing to make it more idiot proof.
Eg, why in the world save the downloads if it is going to ignore them
anyway.

>
> See that site in the wiki for the pros/cons of each

On Sat, 9 Apr 2022 22:06:48 -0000 (UTC), William Unruh wrote:
> On 2022-03-30, David W. Hodgins <dwhodgins@nomail.afraid.org> wrote:
>> On Wed, 30 Mar 2022 00:07:19 -0400, William Unruh <unruh@invalid.ca> wrote:
>>> Thre are three ways I can think of upgrading say from Mga7 to 8.
>>> a) Wipe 7 and Install 8.
>><snip>
>>> b) Upgrade in place from the cdrom/usb stick.
>>> c) Update by transfering a Mga8 /etc/urpmi/urpmi.cfg in, running
>>> urpmi.update -a, and then urpmi --auto-select
>><snip>
>>> What are the other disadvnatages of the three choices? Any advice?
>>
>> This is overlooking the most common method, initiating the upgrade using mgaapplet.
>> It's like option c above, but done without having to manually alter the urpmi.cfg
>> file.
>>
>> There's also "urpmi.removemedia -a", followed by adding the media. That can be done
>> using urpmi.addmedia, as per
>> https://wiki.mageia.org/en/How_to_choose_the_right_Mageia_upgrade_method#How_to_do_a_simple_upgrade_in_text_mode
>
> That web page or the method has a lacuuna. I run the test. Each time I
> do so, it downloads all of the files again (more than 1 hr on my Gbit
> fibre cable) making the test a real pain to use. It also does not say
> that it has run out of room. It just says that the package needs 27MB to
> install. And I have 3GB on / partition. I finally figured out that this
> meant it had run out of room on the test run.
> Then, when I finally made enough room ( I had 17GB hidden under the
> mount point of my /local) and it said I could install, it seemed again
> to go ahead and download all of files that it just downloaded-- a real
> waste of time. I tried looking at the options for uprmi but could not
> see anything which sid to just use the cached files in the cache
> directory /newlocal/rpms (that was where I had to put the cache since /
> was too full).

There is no time saved on test, or not, as far as download time is concerned.

I do have to disagree about using the downloaded files. My pull_updates
script checks the node and if not "wb" then rsync's the cached rpms from 'wb' then
do the urpmi to get sync*.cz and already sees the cached rpm files
and does not download them again.

On Sat, 09 Apr 2022 18:06:48 -0400, William Unruh <unruh@invalid.ca> wrote:
> Ie, that web page really needs some editing to make it more idiot proof.
> Eg, why in the world save the downloads if it is going to ignore them
> anyway.

The files in /var/cache/urpmi/rpms should only be deleted if they have actually
been installed (i.e. without the --test), or if "urpmi --clean" is run.

It should not download them again, and hasn't when I've tested it. If there isn't
enough room, in /var/cache/urpmi/rpms, and you're using a directory such as
/newlocal/rpms, then replace the directory /var/cache/urpmi/rpms with a symlink
to /newlocal/rpms.

I've added a note to https://wiki.mageia.org/en/How_to_choose_the_right_Mageia_upgrade_method#How_to_do_a_simple_upgrade_in_text_mode

Regards, Dave Hodgins

On 2022-04-09, Bit Twister <BitTwister@mouse-potato.com> wrote:
> On Sat, 9 Apr 2022 22:06:48 -0000 (UTC), William Unruh wrote:
>> On 2022-03-30, David W. Hodgins <dwhodgins@nomail.afraid.org> wrote:
>>> On Wed, 30 Mar 2022 00:07:19 -0400, William Unruh <unruh@invalid.ca> wrote:
>>>> Thre are three ways I can think of upgrading say from Mga7 to 8.
>>>> a) Wipe 7 and Install 8.
>>><snip>
>>>> b) Upgrade in place from the cdrom/usb stick.
>>>> c) Update by transfering a Mga8 /etc/urpmi/urpmi.cfg in, running
>>>> urpmi.update -a, and then urpmi --auto-select
>>><snip>
>>>> What are the other disadvnatages of the three choices? Any advice?
>>>
>>> This is overlooking the most common method, initiating the upgrade using mgaapplet.
>>> It's like option c above, but done without having to manually alter the urpmi.cfg
>>> file.
>>>
>>> There's also "urpmi.removemedia -a", followed by adding the media. That can be done
>>> using urpmi.addmedia, as per
>>> https://wiki.mageia.org/en/How_to_choose_the_right_Mageia_upgrade_method#How_to_do_a_simple_upgrade_in_text_mode
>>
>> That web page or the method has a lacuuna. I run the test. Each time I
>> do so, it downloads all of the files again (more than 1 hr on my Gbit
>> fibre cable) making the test a real pain to use. It also does not say
>> that it has run out of room. It just says that the package needs 27MB to
>> install. And I have 3GB on / partition. I finally figured out that this
>> meant it had run out of room on the test run.
>> Then, when I finally made enough room ( I had 17GB hidden under the
>> mount point of my /local) and it said I could install, it seemed again
>> to go ahead and download all of files that it just downloaded-- a real
>> waste of time. I tried looking at the options for uprmi but could not
>> see anything which sid to just use the cached files in the cache
>> directory /newlocal/rpms (that was where I had to put the cache since /
>> was too full).
>
> There is no time saved on test, or not, as far as download time is concerned.

Which is stupid. The files are all downloaded and tested that they will
(at least superficially) install. I have now downloaded about 50GB,
instead of 5 from princeton. That overloads their server, and is really
bad manners.

>
> I do have to disagree about using the downloaded files. My pull_updates
> script checks the node and if not "wb" then rsync's the cached rpms from 'wb' then
> do the urpmi to get sync*.cz and already sees the cached rpm files
> and does not download them again.

I do agree rsync would be better, but the default is curl, and wget is
easy. But using rsync can be rather tricky getting the right path.
man urpmi.conf
gives no hint about how to use rsync. And rsync can be tricky.

Can I simply put downloader: rsync
into the first stanza and replace http:// with rsync://
in the URL? EG
rsync://mirror.math.princeton.edu/pub/mageia/distrib/7/x86_64/media/core/release
in /etc/urpmi/urpmi.conf?

On Sat, 09 Apr 2022 18:55:44 -0400, William Unruh <unruh@invalid.ca> wrote:
> On 2022-04-09, Bit Twister <BitTwister@mouse-potato.com> wrote:
>> There is no time saved on test, or not, as far as download time is concerned.
>
> Which is stupid. The files are all downloaded and tested that they will
> (at least superficially) install. I have now downloaded about 50GB,
> instead of 5 from princeton. That overloads their server, and is really
> bad manners.

There is no time saved by using the test option, compared to skipping the step with
the --test option. It doesn't download them again, if they are in
/var/cache/urpmi/rpms. Note if you get a corrupted download, you must manually delete
the corrupted file, or use urpmi --clean, which deletes all files in that directory.

Regards, Dave Hodgins

On Sat, 9 Apr 2022 22:55:44 -0000 (UTC), William Unruh wrote:

<snip>
>> There is no time saved on test, or not, as far as download time is concerned.
>
> Which is stupid. The files are all downloaded and tested that they will
> (at least superficially) install. I have now downloaded about 50GB,
> instead of 5 from princeton. That overloads their server, and is really
> bad manners.

Hmmm, Off hand I do not remember 50 gig downloads unless doing a new release.
If those are normal updates then I suggest you should be installing updates
more often. Just checked, third time today and just now updated four packages.

>
>
>>
>> I do have to disagree about using the downloaded files. My pull_updates
>> script checks the node and if not "wb" then rsync's the cached rpms from 'wb' then
>> do the urpmi to get sync*.cz and already sees the cached rpm files
>> and does not download them again.
>
> I do agree rsync would be better, but the default is curl, and wget is
> easy. But using rsync can be rather tricky getting the right path.
> man urpmi.conf
> gives no hint about how to use rsync. And rsync can be tricky.

My rsync is just between local nodes. wb being my web browsing node.
I use wget for pulling anything from mirrors/sites.

On 2022-04-09, Bit Twister <BitTwister@mouse-potato.com> wrote:
> On Sat, 9 Apr 2022 22:55:44 -0000 (UTC), William Unruh wrote:
>
> <snip>
>>> There is no time saved on test, or not, as far as download time is concerned.
>>
>> Which is stupid. The files are all downloaded and tested that they will
>> (at least superficially) install. I have now downloaded about 50GB,
>> instead of 5 from princeton. That overloads their server, and is really
>> bad manners.
>
> Hmmm, Off hand I do not remember 50 gig downloads unless doing a new release.
> If those are normal updates then I suggest you should be installing updates
> more often. Just checked, third time today and just now updated four packages.

Thus are multiple reloads of the same files (about 5G worth each time,
but many times)

>
>>
>>
>>>
>>> I do have to disagree about using the downloaded files. My pull_updates
>>> script checks the node and if not "wb" then rsync's the cached rpms from 'wb' then
>>> do the urpmi to get sync*.cz and already sees the cached rpm files
>>> and does not download them again.
>>
>> I do agree rsync would be better, but the default is curl, and wget is
>> easy. But using rsync can be rather tricky getting the right path.
>> man urpmi.conf
>> gives no hint about how to use rsync. And rsync can be tricky.
>
> My rsync is just between local nodes. wb being my web browsing node.
> I use wget for pulling anything from mirrors/sites.

I got it to work, more of less as I suggested. I changes http: in
urpmi.conf to rsync: , put in a "downloader: rsync" into the very first
{ }
section, (and for jameswhitby site, removed the "pub" as the first
sub-directory) Only jameswhitby and princeton seem to be high speed and
support rsync in N America.

Does anyone know where a source is for Mga6? The sites I have looked at
are all 7,8,cauldron only.

On Sat, 09 Apr 2022 21:32:29 -0400, William Unruh <unruh@invalid.ca> wrote:
> Thus are multiple reloads of the same files (about 5G worth each time,
> but many times)

That is not my experience.

> Does anyone know where a source is for Mga6? The sites I have looked at
> are all 7,8,cauldron only.

https://distrib-coffee.ipsl.jussieu.fr/pub/linux/Mageia-archive/distrib/

There may be other mirrors that have the archive, but if so, I'm not aware of them.

Regards, Dave Hodgins

On 2022-04-10, David W. Hodgins <dwhodgins@nomail.afraid.org> wrote:
> On Sat, 09 Apr 2022 21:32:29 -0400, William Unruh <unruh@invalid.ca> wrote:
>> Thus are multiple reloads of the same files (about 5G worth each time,
>> but many times)
>
> That is not my experience.

I know. I was using an alternative cache (/newlocal/rpms) so that might
have made the difference. Your suggestion of linking
/var/cache/urpmi/rpms to that directory is probably good, but it occured
after the last, installation download of 3000 packages had finished (after many earlier
ones as I tried to get the test to pass) and the files were being
installed.

>
>> Does anyone know where a source is for Mga6? The sites I have looked at
>> are all 7,8,cauldron only.
>
> https://distrib-coffee.ipsl.jussieu.fr/pub/linux/Mageia-archive/distrib/
>
> There may be other mirrors that have the archive, but if so, I'm not aware of them.
>
> Regards, Dave Hodgins