Re: le panneau d'annonce planté en suède

[Basile Starynkevitch]

On 1/27/19 9:13 PM, Bernard Schoenacker wrote:

bonjour,

j'ai trouvé cette image :

https://media.mstdn.io/mstdn-media/media_attachments/files/004/798/189/original/17881511899d00ac.jpg

désolé pour le troll, mais pour une fois que c'est
pas un panneau avec kro$oft ...



Dans le même genre, les panneaux à l'intérieur du bus 91-06 que je 
prends pour aller au travail à Massy Palaiseau (vers NanoInnov, arrêt 
Corbeville) direction Saclay -en banlieue parisienne sud- affichent 
régulièrement une invite GRUB.



Ce n'est pas étonnant: Linux est très utilisé dans le monde des systèmes 
embarqués non critiques.



Librement

--
Basile STARYNKEVITCH   == http://starynkevitch.net/Basile
opinions are mine only - les opinions sont seulement miennes
Bourg La Reine, France

Re: Need help with Play on Linux

[Matthew Crews]

On 1/27/19 8:26 PM, Jiangsu Kumquat wrote:
> So, I did all that and then I tried once again to get Spotify to install
> without errors, but It still can't find the .dll file.
> 
> I posted a picture:
> https://postimg.cc/wRLyFTyb
> and added a comment about this in the PlayOnLinux forum under Spotify:
> https://www.playonlinux.com/en/app-386.html
> 
> I think I give up. Other people have been having this issue for a while.
> Thanks oodles for your help. I did not know about adding the i386 to
> dpkg before you told me about it.

I just looked, and Spotify apparently has a native Linux client. So I
guess Wine/PoL are not needed anymore?

https://www.spotify.com/us/download/linux/

You can either use the .deb package they provide, or you can use a snap.
Either way should yield better results.

Cheers,

-Matt

Re: Need help with Play on Linux

[Jiangsu Kumquat]

 So, I did all that and then I tried once again to get Spotify to install
without errors, but It still can't find the .dll file.

I posted a picture:
https://postimg.cc/wRLyFTyb
and added a comment about this in the PlayOnLinux forum under Spotify:
https://www.playonlinux.com/en/app-386.html

I think I give up. Other people have been having this issue for a while.
Thanks oodles for your help. I did not know about adding the i386 to dpkg
before you told me about it.

On Fri, Jan 25, 2019 at 3:03 PM Matthew Crews 
wrote:

> Let's take this a step at a time.
>
> On 1/25/19 11:55 AM, Jiangsu Kumquat wrote:
> > Here's some stats about the computer I'm running Play on Linux with:
> >
> > OS: stretch 64bit
> > CPU: Intel® Pentium(R) 4 CPU 3.00GHz × 2
> > RAM:  2.0 GiB
> > Graphics: Intel® 945G (built-in video on the motherboard)
> > I think I have located the driver for this video card but I'm not sure
> > how to install it. I tried to install it as a regular program, but the
> > setup program says that my computer does not meet the requirements to
> > install it.
>
> The Linux Kernel has built-in drivers for your graphics. No additional
> drivers needed.
>
> > I just added i386 to dpkg and installed wine32. This seemed to resolve a
> > lot of problems.
>
> Good!
>
> > I successfully got the installer for a windows app to run. However it
> > did not install correctly...
> > I'm trying to install Spotify client.
> >
> > I installed "mono-complete" and "mono-complete:i386" but the installer
> > still could not find mono.
> > So I had the play on Linux installer install mono.
> >
> > *It complained that I do not have "wine-gecko" installed. *I cannot seem
> > to resolve this.
>
> Thats odd, PlayOnLinux should have prompted you to install wine-gecko.
>
> > Spotify says that it could not find WINHTTP.dll and won't run.
> >
> >
> > However, I did successfully manually install and run this software with
> > default wine... although it doesn't require a lot of stuff to run. (It
> > will run in DOS apparently) It's an audio spectrum analyzer.
> >
> >
> https://www.softpedia.com/get/Multimedia/Audio/Other-AUDIO-Tools/Audio-SpectrumView.shtml
> >
> > Also, on the play on Linux main window, the capture plugin gives me this
> > error:
> >
> > 01/25/19 09:03:52 - [main] Fatal: Please install alsa-base
> >
> > However, alsa-base is not available in stretch.
>
> Have you tried installing Spotify via vanilla Wine, instead of PlayOnLinux?
>
> > The attached log file is for Spotify. It doesn't say it in the file, but
> > I am getting an error message that it cannot find WINHTTP.dll.
>
> The only thing I can see that *might* cause a problem is that you are
> installing Wine 1.7.35. Granted, this is the version that is in Stretch,
> but it is a very old version. Perhaps try installing it from Stretch
> Backports, then reinstall Wine, and see if that helps. Also, Spotify
> appears to be problematic with Wine outright.
>
> https://appdb.winehq.org/objectManager.php?sClass=application=8514
>
> To install from backports, first add backports to your repos.
>
> # echo "deb http://deb.debian.org/debian stretch-backports main" | tee
> /etc/apt/sources.list.d/stretch-backports
> # apt update
>
> Then install (or in this case, re-install) Wine
>
> # apt install \
>   wine/stretch-backports \
>   wine32/stretch-backports \
>   wine64/stretch-backports \
>   libwine/stretch-backports \
>   libwine:i386/stretch-backports \
>   fonts-wine/stretch-backports
>
> This will install Wine 3.0.3 from backports, and should give improved
> compatibility.
>

Re: why lavrec isn't in package mjpegtools?

[Long Wind]

Thank David!
in changelog.Debian:
Update shlibs file for libmjpegtools-1.10-0.
- liblavrec removed for now (depends on V4L1, doesn't build). 

On Monday, January 28, 2019 9:40 AM, David Wright 
 wrote:
 

 On Sun 27 Jan 2019 at 22:13:08 (+), Long Wind wrote:
> lavrec's manual is in mjpegtoolsbut binary isn't

No idea. Well, just one. Search for lavrec in the
changelog.Debian.gz file, which might give a hint.

Cheers,
David.



   

Re: system stops

[David Wright]

On Sun 27 Jan 2019 at 20:23:38 (-0500), Cindy-Sue Causey wrote:
> On 1/27/19, Cindy-Sue Causey  wrote:
> >
> > AND THAT... just reminded me that there was a thread where someone
> > here said something I'd never remembered hearing before... There's a
> > spot related to I THINK initramfs or something LIKE that
> > (initrd.img??) where we can accidentally be carrying over a wrong UUID
> > (or possibly PARTUUID, LABEL, etc) declaration.
> >
> > That would, for example, happen if we copied over a live session from
> > one partition to another. It's well known that we immediately need to
> > change declarations in /etc/fstab to represent the new partition. This
> > other, though, just might answer some questions like this that have
> > otherwise long gone unsolved in the past..
> >
> > OR NOT. *grin*
> 
> 
> Sorry, I did it AGAIN. Left out the rationale for even bringing that up.
> 
> Whichever kind Debian User brought that topic up in last couple months
> said that it's a place where our systems would be tapping into when
> bringing things up to speed. If the wrong declaration is sitting in
> that position, our systems will stall while they're scratching their
> heads trying to figure out what to do next when they can't find that
> wrong UUID (or possibly PARTUUID or LABEL) that was accidentally
> carried over from another partition or suchly..

Sounds like you're referring to /etc/initramfs-tools/conf.d/resume
which needs to be present in the initramfs as well as the filesystem.

Cheers,
David.

Re: repositorio

[Nitebirdz]

On Sun, Jan 27, 2019 at 09:02:49PM -0400, Fraciel Arevalo wrote:
> buenas noches amigos de debian mi distribusion es debian 9 y quiero ponerla a
> tono con todo repositorios y paquetes pero cada ves que uso un comando como
> apt-get y aptitude no logro que se realise la operasion y alega dependencias
> incumplidas si pueden orientarme sera de mucha ayuda gracias
> 

Perdona si ya sabes lo que explico a continuación, pero ignoro qué tan
familiarizado puedas estar con Linux en general (y Debian en concreto). 

¿Has probado a actualizar corriendo los siguientes comandos como usuario
root?

  # apt update

  # apt upgrade

Si es así, ¿cuáles son los errores que estás viendo? 

Por si sirve de ayuda, yo recientemente tuve un problema similar al tratar
de actualizar mi portátil con Debian Stretch debido a un asunto de
dependencias con paquetes de Nvidia. Mis errores eran algo parecido a lo
que se muestra en esta página:

http://forums.debian.net/viewtopic.php?f=5=138150=0

Al final, para solucionarlos tuve que correr el siguiente comando:

  # apt dist-upgrade

No obstante, antes de que probases algo así convendría saber qué errores
te estás encontrando, pues puede ser que la raíz de tu problema sea bien
distinta. 

-- 
Nitebirdz
http://www.sacredchaos.com/

Re: why lavrec isn't in package mjpegtools?

[David Wright]

On Sun 27 Jan 2019 at 22:13:08 (+), Long Wind wrote:
> lavrec's manual is in mjpegtoolsbut binary isn't

No idea. Well, just one. Search for lavrec in the
changelog.Debian.gz file, which might give a hint.

Cheers,
David.

Re: repositorio

[qorg11]

has tratado # apt -f install ?

El 28 de enero de 2019 2:02:49 CET, Fraciel Arevalo 
 escribió:
>buenas noches amigos de debian mi distribusion es debian 9 y quiero
>ponerla
>a tono con todo repositorios y paquetes pero cada ves que uso un
>comando
>como apt-get y aptitude no logro que se realise la operasion y alega
>dependencias incumplidas si pueden orientarme sera de mucha ayuda
>gracias

-- 
Enviado desde mi dispositivo Android con K-9 Mail. Por favor, disculpa mi 
brevedad.

Re: system stops

[Cindy-Sue Causey]

On 1/27/19, Cindy-Sue Causey  wrote:
>
> AND THAT... just reminded me that there was a thread where someone
> here said something I'd never remembered hearing before... There's a
> spot related to I THINK initramfs or something LIKE that
> (initrd.img??) where we can accidentally be carrying over a wrong UUID
> (or possibly PARTUUID, LABEL, etc) declaration.
>
> That would, for example, happen if we copied over a live session from
> one partition to another. It's well known that we immediately need to
> change declarations in /etc/fstab to represent the new partition. This
> other, though, just might answer some questions like this that have
> otherwise long gone unsolved in the past..
>
> OR NOT. *grin*


Sorry, I did it AGAIN. Left out the rationale for even bringing that up.

Whichever kind Debian User brought that topic up in last couple months
said that it's a place where our systems would be tapping into when
bringing things up to speed. If the wrong declaration is sitting in
that position, our systems will stall while they're scratching their
heads trying to figure out what to do next when they can't find that
wrong UUID (or possibly PARTUUID or LABEL) that was accidentally
carried over from another partition or suchly..

Cindy :)
-- 
Cindy-Sue Causey
Talking Rock, Pickens County, Georgia, USA

* runs with birdseed *

repositorio

[Fraciel Arevalo]

buenas noches amigos de debian mi distribusion es debian 9 y quiero ponerla
a tono con todo repositorios y paquetes pero cada ves que uso un comando
como apt-get y aptitude no logro que se realise la operasion y alega
dependencias incumplidas si pueden orientarme sera de mucha ayuda gracias

Re: system stops

[Cindy-Sue Causey]

On 1/27/19, Jens Holzhäuser  wrote:
> On Sun, Jan 27, 2019 at 09:11:33PM +0100, BELAHCENE Abdelkader wrote:
>> Sometimes (maybe often) when I leave the system for a times without
>> touching it, when I come back, the system is frozen , juste the pointer
>> of
>> mouse can move, but nothing else,  keyboard doesn't respond,  even
>> ctrl+alt+F1 , or F2,  ...
>
>
>> Jan 27 20:29:52 mx kernel: [11477.409506] PM: suspend exit
> [...]
>> Jan 27 20:29:52 mx kernel: [11476.619289] sd 0:0:0:0: [sda] Starting disk
> [...]
>> Jan 27 20:29:52 mx kernel: [11476.091515] ACPI: Waking up from system
>> sleep state S3
> [...]
>> Jan 27 20:29:52 mx kernel: [11476.082190] ACPI: Low-level resume complete
>
> Looks like your system is waking up from being suspended/hibernated.
>
> Try disabling suspension/hibernation and see if that prevents the issue.
> Something with the power management and system resume might not be
> working correctly.


That's what I was thinking, too. Then, if it looks like all of that is
set for it to just sit there instead of shutting down, it's ok to say,
nope, it's still doing it

Because mine does, too. I'll go off and do a few chores then come back
to it. It's obvious that it has still gone into some level of
bystanding mode in spite of my own settings. It takes its absolute
sweet time coming back out of it, too.

I had to change mine to not go into sleep/hibernation at all because
it just would not recover too frequently. Hitting that power button
was repeatedly the only way to recover in my case, too.

AND THAT... just reminded me that there was a thread where someone
here said something I'd never remembered hearing before... There's a
spot related to I THINK initramfs or something LIKE that
(initrd.img??) where we can accidentally be carrying over a wrong UUID
(or possibly PARTUUID, LABEL, etc) declaration.

That would, for example, happen if we copied over a live session from
one partition to another. It's well known that we immediately need to
change declarations in /etc/fstab to represent the new partition. This
other, though, just might answer some questions like this that have
otherwise long gone unsolved in the past..

OR NOT. *grin*

As an aside, can we address that by simply updating initramfs via that
command that was just provided in that one thread a couple days ago?

Maybe that's something we're supposed to be doing after copying over
to a new partition, and some of us just hadn't seen that memo yet? Or
would that not help because we still need to manually change that
particular "declaration" first before we update... if it's even about
update-initramfs in the first place? :)

One other thing I've noticed lately is about things like shutting
programs down to bring them back up and having it just sit there for
the longest time before things finally happen. I started watching
"free -m" during those times IF there was enough memory available to
even run that command, GRIN.

Time and again I've sat here and watched it SLOWLY pull things out of
swap and put it back I PRESUME into "ram"... and THEN it FINALLY shuts
the program on down. :)

While what I just wrote is about shutting programs down, I've imagined
it playing some potentially similar part in spending the same amount
of time watching it come back up from a hibernation it was never
expected to take in the first place. :D

Cindy :)
-- 
Cindy-Sue Causey
Talking Rock, Pickens County, Georgia, USA

* runs with birdseed *

Re: system stops

[David Christensen]

On 1/27/19 12:11 PM, BELAHCENE Abdelkader wrote:

Sometimes (maybe often) when I leave the system for a times without
touching it, when I come back, the system is frozen , juste the pointer of
mouse can move, but nothing else,  keyboard doesn't respond,  even
ctrl+alt+F1 , or F2,  ...
So the only  thing todo is stop button.


Install the package:

openssh-server


Configure your host and/or network so that you can log in over the network.


David

Re: Debian em Blade com storage

[Leandro Guimarães Faria Corcete DUTRA]

Le samedi 26 janvier 2019 à 11:32 +0100, Helio Loureiro a écrit :
> Use openstack ou kubernetes se for possível trabalhar só com
> containers.
> 
> OpenStack tem suporte à containers também mas nunca testei.  Dá
> trabalho instalar e configurar mas o resultado compensa.
> 
> Promox é coisa pequena.  Serve pra solução caseira.

Que loucura é essa de recomendar sistemas sem saber para que se os
usarão?


-- 
skype:leandro.gfc.dutra?chat  Yahoo!: ymsgr:sendIM?lgcdutra
+55 (61) 3546 7191  gTalk: xmpp:leand...@jabber.org
+55 (61) 9302 2691ICQ/AIM: aim:GoIM?screenname=61287803
BRAZIL GMT−3  MSN: msnim:chat?contact=lean...@dutra.fastmail.fm

Re: Can a recipients rights under GNU GPL be revoked?

[rhkramer]

On Sunday, January 27, 2019 07:24:17 PM rhkra...@gmail.com wrote:
> Resending to the list -- I didn't notice that Ivan had sent this to me
> only, and my reply, of course, then went to him only.
> 
> On Sunday, January 27, 2019 10:06:46 AM Ivan Ivanov wrote:
> > Yes: The linux devs can rescind their license grant. GPLv2 is a bare
> > license and is revocable by the grantor. Search for "vsnsdualce" "gpl"
> > online to find his messages which prove that, he is a lawyer and has
> > investigated this subject very well. I am CC'ing him in case you'd
> > like to request more information. So if you didn't like the Code of
> > Conduct covertly accepted behind the scenes against your will, and
> > maybe some other questionable political decisions in technical
> > projects

...
 
> I *might* go read some of the stuff by vsnsdualce, but the Weboob situation
> is not an example of a (free or GPL) license being rescinded.  (You didn't
> quite say it was, but one could infer that is what you are trying to say
> by its inclusion in the same paragraph.)

Ok, I went and read a few things by "vsnsdualce" re the GPL, in particular:

http://readlist.com/lists/gentoo.org/gentoo-user/42/213256.html

And from that, I went to:

https://copyleft.org/guide/comprehensive-gpl-guidech8.html#x11-540007.4

It seems clear that this is one of those things that I talked about in my 
previous last paragraph (the aside, still quoted below) -- vsnsdualce is 
stating his opinion / taking a position that is in opposition to the postions 
/ opinions of other lawyers.

I don't know how far he is willing to go to try to confirm his position, but 
until a court case or something similar (and probably appeals) decides the 
issue, there are two opinions.

If I had to guess / be which would prevail, I would bet on the side of 
copyleft.org who, in a way are the successors (mcow) to the original author(s) 
of the GPL.  (And Bradley Kuhn is a lawyer -- my older mind can't remember if 
he was the lawyer who argued (and lost) a previous free software case (don't 
remember the details) in front of the US Supreme Court.  (Sometimes referred 
to as "the supremes", but they don't really sing that well (well, to be fair, 
I guess I never heard them sing ;-)

Even if he was the guy that lost that case (I'm fairly sure it was someone 
else), he is certainly a very experienced lawyer, and very familiar with the 
issues around this license.  I would trust his opinion more that I would 
"vsnsdualce"'s.

...

> 
> Just another aside: One of my takes on lawyers is that they interpret laws
> and take legal positions for various reasons, often to further their own
> or their client's interests, and then are willing to fight the legal
> battle that may ensue.  A lawyer expressing an opinion does not make that
> opinion correct / legal.

Re: Has NFS changed recently?

[Mike Kupfer]

John wrote:

> Mail is delivered into a mbox on the Debian server, and the disk is
> mounted on the user machine via NFS so the mail can be read into emacs
> from a simple file.  This has worked for many years.  But now with the
> jessie->stretch upgrade it fails with emacs barfing.

Can you give more details, like the error message(s) from Emacs?

mike

Re: Can a recipients rights under GNU GPL be revoked?

[rhkramer]

Resending to the list -- I didn't notice that Ivan had sent this to me only, 
and my reply, of course, then went to him only.

On Sunday, January 27, 2019 10:06:46 AM Ivan Ivanov wrote:
> Yes: The linux devs can rescind their license grant. GPLv2 is a bare
> license and is revocable by the grantor. Search for "vsnsdualce" "gpl"
> online to find his messages which prove that, he is a lawyer and has
> investigated this subject very well. I am CC'ing him in case you'd
> like to request more information. So if you didn't like the Code of
> Conduct covertly accepted behind the scenes against your will, and
> maybe some other questionable political decisions in technical
> projects 

> (e.g. the recent removal of useful "weboob" package which
> have been a part of Debian for 8 years but got removed just because
> some mad SJWs suddenly got offended at its' name) - well you know what
> to do, and maybe vsnsdualce will be happy to help with your case free
> of charge.

I *might* go read some of the stuff by vsnsdualce, but the Weboob situation is 
not an example of a (free or GPL) license being rescinded.  (You didn't quite 
say it was, but one could infer that is what you are trying to say by its 
inclusion in the same paragraph.)

Whatever license and rights conveyed by that license still exist, but Debian 
(not the copyright owner) has decided no longer to include that in what they 
distribute.

You can still get the Weboob package from other sources (unless they all 
disappear) and use the Weboob package in accordance with the license terms for 
the package you find.

Just another aside: One of my takes on lawyers is that they interpret laws and 
take legal positions for various reasons, often to further their own or their 
client's interests, and then are willing to fight the legal battle that may 
ensue.  A lawyer expressing an opinion does not make that opinion correct / 
legal.

Re: cups "Filter failed" | filter rastertopdf stops with status 1 | local printing works; remote printing not

[Brian]

On Sun 27 Jan 2019 at 20:49:36 +0100, toog...@mailbox.org wrote:

> Hey!
> 
> i have a laptop and one remote server/computer. The remote server is 
> bananian, i
> hope that doesn't matter for this case.
> 
> I have  configured cups with hplip to print things on the remote server. That
> means, i can go to https://remote-server/printers/printername in my laptop's
> webbrowser and print a testpage there. However, when i visit
> https://localhost:631 (on the laptop), add the printer of the remote-server
> manually there (url is ipp://172.16.2.4:631/printers/HP_Officejet_6600) and
> print e.g. the testpage, the job fails with "Filter failed".

[...]

Please confirm that /var/log/cups/error_log (LogLevel warn) is from the
client and say *exactly* what you did when you added the printer of the
remote-server with localhost:631.

-- 
Brian.

Re: Can a recipients rights under GNU GPL be revoked?

[John Hasler]

Richard writes:
> IANAL, but ok. But what defines a 'legitimate owner'? No ownership
> rights are generally transferred with a piece of software, right?

A copy under USA copyright law is a tangible object.  When you buy a CD
or other embodiment of a work you own that object and have all the
rights of ownership.  Ownership of a copy --a tangible object-- is
entirely different from ownership of the right to make and distribute
additional copies of an object embodying material protected by
copyright.  Copyright law does not limit your rights in that object.
You own it.

You do not own the right to make and distribute copies of that object.
However, USA copyright law automatically grants you the right to make
any transient or temporary copies that might be necessary in order to
use it.  You have the right to transfer ownership of the object to
someone else, but if you do so you must, of course, destroy the
aforementioned temporary copies.

> ... and I understand that if there's no consideration paid, there's no
> legal contract, right?

"Contract"?  If someone makes a gift of an object to you expecting no
compensation, you still own it.
-- 
John Hasler 
jhas...@newsguy.com
Elmwood, WI USA

Re: system stops

[Jens Holzhäuser]

On Sun, Jan 27, 2019 at 09:11:33PM +0100, BELAHCENE Abdelkader wrote:
> Sometimes (maybe often) when I leave the system for a times without
> touching it, when I come back, the system is frozen , juste the pointer of
> mouse can move, but nothing else,  keyboard doesn't respond,  even
> ctrl+alt+F1 , or F2,  ...


> Jan 27 20:29:52 mx kernel: [11477.409506] PM: suspend exit
[...]
> Jan 27 20:29:52 mx kernel: [11476.619289] sd 0:0:0:0: [sda] Starting disk
[...]
> Jan 27 20:29:52 mx kernel: [11476.091515] ACPI: Waking up from system sleep 
> state S3
[...]
> Jan 27 20:29:52 mx kernel: [11476.082190] ACPI: Low-level resume complete

Looks like your system is waking up from being suspended/hibernated.

Try disabling suspension/hibernation and see if that prevents the issue.
Something with the power management and system resume might not be
working correctly.

I am not very familiar with Linux hibernation, so there's not much more
I can help with here.

Jens

Re: Disque Dur SSD

[Gaëtan Perrier]

Le dimanche 27 janvier 2019 à 12:24 +0100, Sébastien Dinot a écrit :
> Du coup, lorsque j'effectue des compilations intensives d'un gros projet
> en C++ comme c'est le cas actuellement, le chiffrement est pénalisant
> (le volume du projet - 13 Go de données produites - faisant que tous les
> fichiers générés ne peuvent être cachés en mémoire). Mais vu que la
> compilation ex-nihilo dudit projet prend 58 minutes, ce ne sont pas les
> 2 minutes d'écriture des données sur disque qui ralentissent mon
> travail...

Quel est l'intérêt de faire de la compilation sur une partition chiffrée ?
Normalement une très grande partie des données de compilation se trouvent
effacées à la fin de celle-ci. Donc les chiffrer me semble peu pertinent, non ?

Gaëtan


signature.asc
Description: This is a digitally signed message part

Re: Disque Dur SSD

[Sébastien Dinot]

Pascal Hambourg a écrit :
> Pour quelle(s) raison(s) ? Le découpage en blocs qui implique une
> lecture-modification-écriture en cas de modification partielle d'un
> bloc (comme les bandes en RAID 5) ?

Je ne suis pas un spécialiste de ces questions, mais je sais que la
limite vient de la capacité de chiffrement du processeur (les dernières
générations ayant fait d'énormes progrès) et l'expérience montre que le
chiffrement d'un flux est bien plus lent que son déchiffrement.

> Une telle chute en écriture me surprend quand même beaucoup. Comment
> as-tu testé pour obtenir ces valeurs ?

Pour la lecture, je copie un gros fichier dans le répertoire /tmp,
sachant qu'il s'agit d'une partition tmpfs (donc montée en mémoire) :

  sync ; time ( cp debian-9.7.0-amd64-DVD-1.iso /tmp ; sync )

NB ; Ce test n'est consistant que lors du premier essai car les fois
 suivantes, le fichier est en cache et les performances apparentes
 dépassent les performances théoriques du disque.

Pour l'écriture, je duplique ce fichier sur disque :

  sync ; time ( cp /tmp/debian-9.7.0-amd64-DVD-1.iso ~/ ; sync )

Au passage, j'ai effectué deux autres tests ce soir, avec d'autres
fichiers de plusieurs Go, et je constate une certaine variabilité :

Écriture : entre  82 et 109 Mo/s
Lecture :  entre 724 et 830 Mo/s

> Je n'ai pas ce genre de matériel mais j'ai fait un test sur ma petite
> machine avec un volume chiffré LUKS et j'obtiens une chute de 15% en
> lecture et 30% en écriture séquentielle brute avec dd.

La vitesse d'écriture d'un disque dur SATA mécanique plafonnant aux
alentours des 120 Mo/s environ, la capacité de chiffrement du processeur
n'est pas le facteur limitatif car le processeur chiffre les données
plus vite que le disque ne peut les enregistrer (enfin, si l'on
considère un processeur multi-cœur pas trop poussif et ancien).

Lorsque j'utilisais un disque mécanique, la compilation été ralentie
parce qu'un cœur était largement occupé par le chiffrement, mais les
écritures sur disque n'était pas ralenties par le processeur.

Sébastien

-- 
Sébastien Dinot, sebastien.di...@free.fr
http://sebastien.dinot.free.fr/
Ne goûtez pas au logiciel libre, vous ne pourriez plus vous en passer !

Re: Disque Dur SSD

[Gaëtan Perrier]

Au début d'Unix les SSD n'existaient pas ...
Le swap reste utile ne serait-ce que pour l'hibernation d'une machine, il me
semble.

Le vendredi 25 janvier 2019 à 08:56 +0100, aishen a écrit :
> Pas seulement, au début linux/unix on mettait de la swap pour avoir une 
> mémoire disque plus rapide, donc le ssd était presque de la RAM, il n'y 
> a pas d'intérêt de ce côté, mais les ssd on un cycle (grand) de 
> copy/effacer dont il ne vaut pas mieux aller au delà. De toute façon 
> importe chacun fait ce qu'il veut mais ne devrait pas être traité 
> d'absurde ou idiot... Comme disent les enfants, idiot lui-même celui qui 
> le prononce !
> Le 25/01/2019 à 08:38, Daniel Caillibaud a écrit :
> > Le 25/01/19 à 07:35, aishen  a écrit :
> > > on n'a pas besoin de swap sur un ssd linux,
> > 
> > Quel rapport entre le besoin de swap (manque de RAM du système) et la
> > nature du disque ?
> > 
> > > d'après les avis il est
> > > même déconseillé d'en mettre pour éviter des lectures/écritures
> > > constantes qui "usent" la mémoire des ssd
> > Quels avis ? Je suis assez curieux de voir qui peut bien raconter ça…
> > 
> > Le swap, c'est fait pour que le système ne crash pas quand il manque de
> > RAM, il utilise alors le disque.
> > 
> > Et dans ce cas, c'est quand même beaucoup mieux d'avoir un disque rapide…
> > (Cf discussion "Partitionnement d'un serveur web" initiée le 14/01)
> > 
> > Je suis donc de l'avis de Pascal Hambourg, ça me parait idiot de pas mettre
> > de swap parce qu'on a un disque ssd.
> > 
> > Ça peut se justifier de ne pas mettre de swap, mais ça n'a rien à voir avec
> > ssd ou pas.
> > 



signature.asc
Description: This is a digitally signed message part

Re: Has NFS changed recently?

[deloptes]

John wrote:

> nice idea but the uids are the same across the network; the gid differ
> but that should not matter as the access is 600.  I can create and
> delete files read them etc as expected.
> 
> The files are mounted with nfs3
> 
> snout  -fstype=nfs,nfsvers=3,rw,intr,rsize=8192,wsize=8192,user,suid
> 172.16.4.42:/

so if it is not nfs, might be something else with the mail?

Re: le panneau d'annonce planté en suède

[ajh-valmer]

On Sunday 27 January 2019 21:13:49 Bernard Schoenacker wrote:
> j'ai trouvé cette image :
>media.mstdn.io/mstdn-media/media_attachments/files/004/798/189/original/17881511899d00ac.jpg
> désolé pour le troll, mais pour une fois que c'est pas un panneau avec
> kro$oft ... 

En tout cas, la compagnie ferroviaire suédoise est sous GNU/Linux, bravo,
mais pas notre compagnie SNCF, sous m$-windows.

Re: Has NFS changed recently?

[Étienne Mollier]

On 1/27/19 7:35 PM, John wrote:
> I run a mixed Debian/openSuSE /android LAN an after an upgrade of one
> component from jessie to stretch I have hit a big problem.
> 
> Mail is delivered into a mbox on the Debian server, and the disk is
> mounted on the user machine via NFS so the mail can be read into emacs
> from a simple file.  This has worked for many years.  But now with the
> jessie->stretch upgrade it fails with emacs barfing.  I know that the
> emacs did not change so I am guessing that the issue is with NFS.  Any
> possibility of this?
> 
> ==John ffitch

Good Day John,

Mail+NFS+inconsistent file content... maybe it is unrelated,
but this reminds me of something read some time ago:

https://utcc.utoronto.ca/~cks/space/blog/linux/KernelNFSPageBug

A few articles later, it turned out the NFS protocol didn't
support that kind operating mode in the first place:

https://utcc.utoronto.ca/~cks/space/blog/tech/ExplicitSpecImportance

If this is related, it doesn't look like something that
will be "fixed" anytime soon, well not in NFS kernel side
at least.  Could be emacs, or any other process holding the
"mbox" in read-write mode, and causing these kernel cache
inconsistencies across machines.

Kind Regards
-- 
Étienne Mollier 

PS: To answer to the question in the Subject header, yes.
NFS changed between Jessie and Stretch.

Re: Partition information as text file?

[David Wright]

On Sat 26 Jan 2019 at 15:10:55 (-0600), Richard Owlett wrote:
> On 01/26/2019 01:32 PM, Felix Miata wrote:
> > Richard Owlett composed on 2019-01-26 08:32 (UTC-0600):
> > 
> > > I am attempting to create a spreadsheet to document the content of
> > > multiple disks of multiple machines.
> > 
> > > Gparted displays the desired information.
> > > *HOWEVER* I see no way to capture the information.
> > 
> > > At the command line using "lsblk -o NAME,FSTYPE,LABEL /dev/sdb" gives
> > > most of the desired information.
> > 
> > > It omits partition size, used space, and unused space.
> > 
> > > Suggestions?
[…]
> > Sometimes I append output from lsblk or parted -l.
> > 
> > hdparm and smartctl might also provide some of what you're looking for.
> 
> I'll attempt to redefine my problem.
> 
> I have:
>   multiple machines
> each having
>   multiple disks
> each having
>   multiple partitions.
> 
> I wish to inventory the above "conglomeration".
> 
> I wish to to answer the question(s):
>   How big is each
>   How much is available

It appears that you're really interested in the filesystems'
information rather than the partitions', with the exception of the
filesystem LABELs, which you have said elsewhere you use as
indications of the filesystems' contents.

So it looks as if   df --output -x tmpfs -x devtmpfs   gives you all
you want (and more) with the exception of LABELs. It seems sensible
to use   lsblk -o NAME,LABEL -l   to get these because AFAICT it
automatically handles the business of selecting e2label/dosfslabel/etc
as appropriate and gets them all in a heap.

With judicious use of head, tail and sort, it would be fairly simple
to get the two listings to correspond well enough for entry into a
spreadsheet (I don't know what you meant by 'generic'), making
final adjustments (df omits the device and partitions like swap) to
line things up.

> OWL now DUCKS fer cover ;/

No need.

Cheers,
David.

Re: Can a recipients rights under GNU GPL be revoked?

[rhkramer]

On Sunday, January 27, 2019 03:45:14 PM Ivan Ivanov wrote:
> IANAL either but "vsnsdualce" is a lawyer and provided a lot of
> information on how the GPL could be revoked. Maybe the laws, e.g. USA
> laws, are more powerful than your EULAs and EULA-like stuff.
> 
> If you really would like to find out the truth instead of drinking
> kool-aid calming yourself that "nah it can't be revoked because I
> wouldn't like that" then please search for "vsnsdualce" "gpl"
> messages.

You cc'ed me (which I responded to you privately about), but now I'll expound 
to say that was not my position -- I merely pointed out that one (?) way to 
seemingly get around the do not revoke meme would be, as an owner of the 
software (not the owner of a copy of the software or owner of a license to use 
the software (unless that license allowed what I'm going to mention)) is to 
modify the software and issue the new version under different terms.

To all: please don't cc me unless there is some good special reason -- it 
makes me feel that you are somehow personally attributing to me (or arguing 
with me?) about something in the post. 

regards,
Randy Kramer
 
> My personal point of view, which may be incorrect because IANAL, is
> that the forced introduction of Code of Conducts was quite similar to
> one-sided modification of license agreement and should render it
> invalid.
> 
> Best regards,
> Ivan Ivanov
> 

Re: IPv6 static config in /etc/network/interfaces ignored

[Claudio M]

Hi,

On Sun, Jan 27, 2019 at 5:55 PM Pascal Hambourg 
wrote:

> Le 27/01/2019 à 16:29, Claudio M a écrit :
> >>
> >> auto eth0
> >> iface eth0 inet static
> >>address a.b.c.d
> >>netmask 255.255.255.224
> >>gateway c.d.e.f
> >>up route add -net a.b.c.x netmask 255.255.255.224 gw c.d.e.x dev eth0
>
> AFAIK, the gateway must be directly reachable on eth0. But c.d.e.x does
> not seem to belong to a.b.c.d/255.255.255.224.
>

It is, I just replaced the IPs with random(ish) letters


> >>up ip addr add f.g.h.i/32 dev eth0
> >>down ip addr del  f.g.h.i /32 dev eth0
>
> Hackish. You can create a second inet stanza instead.
>

Good to know, thanks, doesn't really pose a problem in this case (I set it
up like my server provider suggested, just in case...).

>>post-up /etc/network/iptables.sh
>
> Is there a chance that this script returns a non zero value ?
>

Yes (also see below)

*post-up command*
> *Run command after bringing the interface up. If this command fails
> then ifup aborts, refraining from marking the interface as configured (even
> though it has really been configured), **prints an error message**, and
> exits with status 0. This behavior may change in the future. *
>


> >> iface eth0 inet6 static
> >>address 2a01:::::2
> >>netmask 64
> >>gateway fe80::1
>
> What is the output of
>
>   service networking status
>

● networking.service - Raise network interfaces
   Loaded: loaded (/lib/systemd/system/networking.service; enabled; vendor
preset: enabled)
   Active: failed (Result: timeout) since Sun 2019-01-27 20:45:15 CET; 1h
1min ago
 Docs: man:interfaces(5)
  Process: 489 ExecStart=/sbin/ifup -a --read-environment (code=killed,
signal=TERM)
  Process: 480 ExecStartPre=/bin/sh -c [ "$CONFIGURE_INTERFACES" != "no" ]
&& [ -n "$(ifquery --read-environment --list --exclude=lo)" ] && udevadm
settle (code=exited, status=0/SUCCESS)
 Main PID: 489 (code=killed, signal=TERM)
  CPU: 484ms

Jan 27 20:40:15 my-server systemd[1]: Starting Raise network interfaces...
Jan 27 20:40:20 my-server ifup[489]: Waiting for DAD... Done
Jan 27 20:45:15 my-server systemd[1]: networking.service: Start operation
timed out. Terminating.
Jan 27 20:45:15 my-server systemd[1]: networking.service: Main process
exited, code=killed, status=15/TERM
Jan 27 20:45:15 my-server systemd[1]: Failed to start Raise network
interfaces.
Jan 27 20:45:15 my-server systemd[1]: networking.service: Unit entered
failed state.
Jan 27 20:45:15 my-server systemd[1]: networking.service: Failed with
result 'timeout'.


>
> Did you try to swap the inet and inet6 stanzas ?
>
> I did, and it solved the assignment problem, but I guess  it's the
iptables.sh script that's causing this.
I suppose I could try to replace it with nftables (we're going there
anyway...), but I guess in any case I need to start looking at a proper
firewall persistence thing (iptables-persistent or its netfilter
equivalent) rather than a shell script :/
Any advice?

@Ric Moore, nope, definitely eth0 (and @tomas is right, they're not devices
:) )

Thanks,

Claudio

Re: Partition information as text file?

[Richard Hector]

On 28/01/19 9:36 AM, Pascal Hambourg wrote:
>>> Usually, all of a partition is used. If the partition contains a
>>> filesystem, swap area, RAID member or LVM physical volume, these data
>>> structures use all the partition space.
>>
>> Not necessarily - eg if you've extended the partition and not the
>> filesystem, it doesn't.
> 
> Of course there are exceptions. This is why I wrote "usually".
> But this space is not accounted as available/free in the partition
> table, nor is it accounted as used/unused in the filesystem. Gparted
> would not display it. Unless it is a transient state, it is just wasted
> space.

I made the point because it might be useful, in this case or others, to
detect such wasted space (eg by comparing the partition table with the
filesystem metadata)

Richard




signature.asc
Description: OpenPGP digital signature

Re: Partition information as text file?

[Pascal Hambourg]

Le 27/01/2019 à 21:08, Richard Hector a écrit :

On 27/01/19 4:45 AM, Pascal Hambourg wrote:

Le 26/01/2019 à 16:12, Richard Owlett a écrit :


But I still need to know how much of each partition is used/unused.


Usually, all of a partition is used. If the partition contains a
filesystem, swap area, RAID member or LVM physical volume, these data
structures use all the partition space.


Not necessarily - eg if you've extended the partition and not the
filesystem, it doesn't.


Of course there are exceptions. This is why I wrote "usually".
But this space is not accounted as available/free in the partition 
table, nor is it accounted as used/unused in the filesystem. Gparted 
would not display it. Unless it is a transient state, it is just wasted 
space.



It's probably possible to specify the fs size at creation
time, too


Yes, some filesystem creation tools allow to specify the size.

system stops

[BELAHCENE Abdelkader]

HI,
uname -a  gives
Linux mx 4.19.0-1-amd64 #1 SMP Debian 4.19.5-2~mx17+1 (2018-12-12) x86_64
GNU/Linux
Sometimes (maybe often) when I leave the system for a times without
touching it, when I come back, the system is frozen , juste the pointer of
mouse can move, but nothing else,  keyboard doesn't respond,  even
ctrl+alt+F1 , or F2,  ...
So the only  thing todo is stop button.
Here is some lines from messages (last line before the shutdown), hope that
someone can help me by reading this  log before the stopping system ( at
20:29).

Jan 27 20:29:52 mx kernel: [11479.565256] Bluetooth: hci0: BCM: Patch
brcm/BCM20702A1-0489-e032.hcd
not found
Jan 27 20:29:52 mx kernel: [11479.565248] bluetooth hci0: Direct firmware
load for brcm/BCM20702A1-0
489-e032.hcd failed with error -2
Jan 27 20:29:52 mx kernel: [11478.719457] ata1.00: configured for UDMA/133
Jan 27 20:29:52 mx kernel: [11478.513494] ata1: SATA link up 3.0 Gbps
(SStatus 123 SControl 300)
Jan 27 20:29:52 mx kernel: [11477.409506] PM: suspend exit
Jan 27 20:29:52 mx kernel: [11477.374637] Bluetooth: hci0: BCM20702A1
(001.002.014) build 
Jan 27 20:29:52 mx kernel: [11477.373634] Bluetooth: hci0: BCM20702A
Jan 27 20:29:52 mx kernel: [11477.356632] Bluetooth: hci0: BCM: features
0x07
Jan 27 20:29:52 mx kernel: [11477.355640] Bluetooth: hci0: BCM: chip id 63
Jan 27 20:29:52 mx kernel: [11477.245215] Restarting tasks ... done.
Jan 27 20:29:52 mx kernel: [11477.245213] OOM killer enabled.
Jan 27 20:29:52 mx kernel: [11477.164547] usb 1-1.4: reset high-speed USB
device number 5 using ehci
-pci
Jan 27 20:29:52 mx kernel: [11477.106796] ata3.00: configured for UDMA/100
Jan 27 20:29:52 mx kernel: [11477.105193] ata3: SATA link up 1.5 Gbps
(SStatus 113 SControl 300)
Jan 27 20:29:52 mx kernel: [11477.022519] usb 1-1.3: reset full-speed USB
device number 4 using ehci
-pci
Jan 27 20:29:52 mx kernel: [11477.018509] usb 4-1.6: reset high-speed USB
device number 3 using ehci
-pci
Jan 27 20:29:52 mx kernel: [11476.837625] ACPI: button: The lid device is
not compliant to SW_LID.
Jan 27 20:29:52 mx kernel: [11476.619289] sd 0:0:0:0: [sda] Starting disk
Jan 27 20:29:52 mx kernel: [11476.608519] ACPI: EC: event unblocked
Jan 27 20:29:52 mx kernel: [11476.594662] ACPI: EC: interrupt unblocked
Jan 27 20:29:52 mx kernel: [11476.091515] ACPI: Waking up from system sleep
state S3
Jan 27 20:29:52 mx kernel: [11476.088916] CPU1 is up
Jan 27 20:29:52 mx kernel: [11476.088582]  cache: parent cpu1 should not be
sleeping
Jan 27 20:29:52 mx kernel: [11476.083485] smpboot: Booting Node 0 Processor
1 APIC 0x2
Jan 27 20:29:52 mx kernel: [11476.083484] x86: Booting SMP configuration:
Jan 27 20:29:52 mx kernel: [11476.083445] Enabling non-boot CPUs ...
Jan 27 20:29:52 mx kernel: [11476.082238] PM: Restoring platform NVS memory
Jan 27 20:29:52 mx kernel: [11476.082238] ACPI: EC: EC started
Jan 27 20:29:52 mx kernel: [11476.082190] ACPI: Low-level resume complete

Sorry if the list is long.
Thanks a lot
regards

Re: Has NFS changed recently?

[John]

> Check the ownership? I recall at some point of time there was something with
> the id mapping - also check if you may need to enforce nfs version 3

nice idea but the uids are the same across the network; the gid differ
but that should not matter as the access is 600.  I can create and
delete files read them etc as expected.

The files are mounted with nfs3

snout  -fstype=nfs,nfsvers=3,rw,intr,rsize=8192,wsize=8192,user,suid 
172.16.4.42:/

==John ffitch

le panneau d'annonce planté en suède

[Bernard Schoenacker]

bonjour,

j'ai trouvé cette image :

https://media.mstdn.io/mstdn-media/media_attachments/files/004/798/189/original/17881511899d00ac.jpg

désolé pour le troll, mais pour une fois que c'est 
pas un panneau avec kro$oft ...

merci

slt
bernard

Re: Partition information as text file?

[Richard Hector]

On 27/01/19 4:45 AM, Pascal Hambourg wrote:
> Le 26/01/2019 à 16:12, Richard Owlett a écrit :
>> On 01/26/2019 08:36 AM, Jude DaShiell wrote:
>>> lsblk >>lsblk.txt
>>
>> I had misinterpreted "SIZE  size of the device" in the response
>> to "lsblk --help". I did not equate "device" to "partition".
>>
>> But I still need to know how much of each partition is used/unused.
> 
> Usually, all of a partition is used. If the partition contains a
> filesystem, swap area, RAID member or LVM physical volume, these data
> structures use all the partition space.

Not necessarily - eg if you've extended the partition and not the
filesystem, it doesn't. Or if you've shrunk the filesystem and not the
partition. It's probably possible to specify the fs size at creation
time, too, but I've never come across a need for that.

Richard



signature.asc
Description: OpenPGP digital signature

Re: Can a recipients rights under GNU GPL be revoked?

[rhkramer]

On Sunday, January 27, 2019 02:41:49 PM Richard Hector wrote:
> On 28/01/19 3:32 AM, John Hasler wrote:
> > Note that under USA law the right to *use* (including the right to make
> > such transient and temporary copies as might be required for effective
> > use) a copy of a program of which one is a legitimate owner is automatic
> > and requires no license.
> 
> IANAL, but ok. But what defines a 'legitimate owner'? No ownership
> rights are generally transferred with a piece of software, right?

IANAL either (aside: when I first saw that acronym, a long time ago, I thought 
(from the context) that someone was trying to say they were not a liar -- to a 
certain extent, I sometime consider the two words synonymous ;-)

Anyway, I don't think "legitimate owner" (of the software) is quite the right 
phrase, it is more like "legitimate owner of a copy of the software" and / or 
the "legitimate owner of a license (or right) to use the software".

(And, yes, consideration is required (afaik), but it is not necessarily 
financial -- it could be (I think) agreement to uphold the terms of the 
license.

Re: Can a recipients rights under GNU GPL be revoked?

[Ben Finney]

rhkra...@gmail.com writes:

> […] I believe that the original author of a package could do something
> like create further modifications to the code and create a non-free
> version of the code.

Yes. The _Copyleft and the GNU General Public License_ guide
https://copyleft.org/guide/> addresses that possibility:

 The GPL is irrevocable in the sense that once a copyright holder
 grants rights for someone to copy, modify and redistribute the
 software under terms of the GPL, they cannot later revoke that
 grant. Since the GPL has no provision allowing the copyright holder
 to take such a prerogative, the license is granted as long as the
 copyright remains in effect. The copyright holders have the right
 to relicense the same work under different licenses […], or to stop
 distributing the GPLv2’d version (assuming GPLv2 §3(b) was never
 used), but they may not revoke the rights under GPLv2 already
 granted.

 In fact, when an entity loses their right to copy, modify and
 distribute GPL’d software, it is because of their *own actions*,
 not that of the copyright holder. The copyright holder does not
 decide when GPLv2 §4 termination occurs (if ever); rather, the
 actions of the licensee determine that.

 
https://copyleft.org/guide/comprehensive-gpl-guidech8.html#x11-510007.1>

So, the copyright holders can make new releases without granting GPL
freedoms. But they have no way to revoke the GPL freedoms already
granted to a person in a specific past release of the work.

-- 
 \“Fascism is capitalism plus murder.” —Upton Sinclair |
  `\   |
_o__)  |
Ben Finney

Re: Can a recipients rights under GNU GPL be revoked?

[Ben Finney]

rhkra...@gmail.com writes:

> […] I believe that the original author of a package could do something
> like create further modifications to the code and create a non-free
> version of the code.

Yes. The _Copyleft and the GNU General Public License_ guide
https://copyleft.org/guide/> addresses that possibility:

 The GPL is irrevocable in the sense that once a copyright holder
 grants rights for someone to copy, modify and redistribute the
 software under terms of the GPL, they cannot later revoke that
 grant. Since the GPL has no provision allowing the copyright holder
 to take such a prerogative, the license is granted as long as the
 copyright remains in effect. The copyright holders have the right
 to relicense the same work under different licenses […], or to stop
 distributing the GPLv2’d version (assuming GPLv2 §3(b) was never
 used), but they may not revoke the rights under GPLv2 already
 granted.

 In fact, when an entity loses their right to copy, modify and
 distribute GPL’d software, it is because of their *own actions*,
 not that of the copyright holder. The copyright holder does not
 decide when GPLv2 §4 termination occurs (if ever); rather, the
 actions of the licensee determine that.

 
https://copyleft.org/guide/comprehensive-gpl-guidech8.html#x11-510007.1>

So, the copyright holder can make new releases without granting GPL
freedoms. But they have no way to revoke the GPL freedoms already
granted to a person in a specific past release of the work.

-- 
 \“Fascism is capitalism plus murder.” —Upton Sinclair |
  `\   |
_o__)  |
Ben Finney

cups "Filter failed" | filter rastertopdf stops with status 1 | local printing works; remote printing not

[toogley]

Hey!

i have a laptop and one remote server/computer. The remote server is bananian, i
hope that doesn't matter for this case.

I have  configured cups with hplip to print things on the remote server. That
means, i can go to https://remote-server/printers/printername in my laptop's
webbrowser and print a testpage there. However, when i visit
https://localhost:631 (on the laptop), add the printer of the remote-server
manually there (url is ipp://172.16.2.4:631/printers/HP_Officejet_6600) and
print e.g. the testpage, the job fails with "Filter failed".

What did i try:

* Interestingly, LogLevel debug doesn't contain the error, only LogLevel warn. 
However, i added both log files below.
* I found
https://thenumberzero.blogspot.com/2014/02/how-to-fix-hp-printer-filter-failed.html
- i did not complete what's suggested there
* hp-check says, cups is not running or the package is missing (both on the
server as well as on my laptop), but i use the webinterface of cups
successfully. And i can even print from the remote server's 
 
* https://www.cups.org/doc/api-raster.html doesn't contain something status
related (the filter exited with status 1)
* googling "Unable to determine number of pages, page count" doesn't help. But 
i think that is only a result of the failed filter.
* I found https://lists.debian.org/debian-user/2017/03/msg00234.html but the 
filter fail for a different reason there.

used filters


D [27/Jan/2019:20:38:14 +0100] [Job 13] 4 filters for job:
D [27/Jan/2019:20:38:14 +0100] [Job 13] rastertopwg 
(application/vnd.cups-raster to image/pwg-raster, cost 100)
D [27/Jan/2019:20:38:14 +0100] [Job 13] rastertopdf (image/pwg-raster to 
application/pdf, cost 32)
D [27/Jan/2019:20:38:14 +0100] [Job 13] pdftopdf (application/pdf to 
application/vnd.cups-pdf, cost 66)
D [27/Jan/2019:20:38:14 +0100] [Job 13] foomatic-rip (application/vnd.cups-pdf 
to printer/HP_Officejet_6600, cost 0)

failed filters
--

D [27/Jan/2019:20:38:14 +0100] [Job 13] PID 2562 
(/usr/lib/cups/filter/rastertopdf) stopped with status 1.



lpstat -t on the remote server
--

# lpstat -t
scheduler is running
system default destination: HP_Officejet_6600
device for HP_Officejet_6600: hp:/usb/Officejet_6600?serial=CN48C9R1BQ05RN
HP_Officejet_6600 accepting requests since Sun 27 Jan 2019 06:22:03 PM CET
printer HP_Officejet_6600 is idle.  enabled since Sun 27 Jan 2019 06:22:03 PM
CET
HP_Officejet_6600-11unknown 311296   Sun 27 Jan 2019 05:36:33 PM CET

lpstat -t on the laptop (contains also a local usb config)
-

system default destination: Officejet-6600
device for Officejet-6600: hp:/usb/Officejet_6600?serial=CN48C9R1BQ05RN
device for remote_printer: ipp://172.16.2.4:631/printers/HP_Officejet_6600
Officejet-6600 accepting requests since Sat 19 Jan 2019 06:14:19 PM CET
remote_printer accepting requests since Sun 27 Jan 2019 05:36:29 PM CET
printer Officejet-6600 is idle.  enabled since Sat 19 Jan 2019 06:14:19 PM CET
printer remote_printer now printing remote_printer-32.  enabled since Sun 27 Jan
2019 05:36:29 PM CET
remote_printer-32   toogley   1024   Sun 27 Jan 2019 05:36:29 PM CET



/var/log/cups/error_log (LogLevel warn)
--

E [27/Jan/2019:20:38:14 +0100] [Job 13] Job stopped due to filter errors; 
please consult the error_log file for details.
D [27/Jan/2019:20:38:14 +0100] [Job 13] The following messages were recorded 
from 08:38:12 PM to 08:38:14 PM
D [27/Jan/2019:20:38:14 +0100] [Job 13] Adding start banner page "none".
D [27/Jan/2019:20:38:14 +0100] [Job 13] Queued on "HP_Officejet_6600" by 
"anonymous".
D [27/Jan/2019:20:38:14 +0100] [Job 13] File of type 
application/vnd.cups-raster queued by "anonymous".
D [27/Jan/2019:20:38:14 +0100] [Job 13] Adding end banner page "none".
D [27/Jan/2019:20:38:14 +0100] [Job 13] time-at-processing=1548617893
D [27/Jan/2019:20:38:14 +0100] [Job 13] 4 filters for job:
D [27/Jan/2019:20:38:14 +0100] [Job 13] rastertopwg 
(application/vnd.cups-raster to image/pwg-raster, cost 100)
D [27/Jan/2019:20:38:14 +0100] [Job 13] rastertopdf (image/pwg-raster to 
application/pdf, cost 32)
D [27/Jan/2019:20:38:14 +0100] [Job 13] pdftopdf (application/pdf to 
application/vnd.cups-pdf, cost 66)
D [27/Jan/2019:20:38:14 +0100] [Job 13] foomatic-rip (application/vnd.cups-pdf 
to printer/HP_Officejet_6600, cost 0)
D [27/Jan/2019:20:38:14 +0100] [Job 13] job-sheets=none,none
D [27/Jan/2019:20:38:14 +0100] [Job 13] argv[0]="HP_Officejet_6600"
D [27/Jan/2019:20:38:14 +0100] [Job 13] argv[1]="13"
D [27/Jan/2019:20:38:14 +0100] [Job 13] argv[2]="anonymous"
D [27/Jan/2019:20:38:14 +0100] [Job 13] argv[3]="Test Page"
D [27/Jan/2019:20:38:14 +0100] [Job 13] argv[4]="1"
D [27/Jan/2019:20:38:14 +0100] [Job 13] 
argv[5]="job-uuid=urn:uuid:5612d4e9-dbf9-30ad-462b-f47e9daab90a 
job-originating-host-name=172.16.2.7 date-time-at-creation= 
date-time-at-processing= 

Re: Can a recipients rights under GNU GPL be revoked?

[Richard Hector]

On 28/01/19 3:32 AM, John Hasler wrote:
> Note that under USA law the right to *use* (including the right to make
> such transient and temporary copies as might be required for effective
> use) a copy of a program of which one is a legitimate owner is automatic
> and requires no license.

IANAL, but ok. But what defines a 'legitimate owner'? No ownership
rights are generally transferred with a piece of software, right?

>  Thus if someone gives (or sells) a copy

... and I understand that if there's no consideration paid, there's no
legal contract, right?

I'm paraphrasing from recent comments by someone who claims to be a
lawyer, posting on The Register. They also claim that promissory
estoppel claims are generally hard to win.

NB I'm in New Zealand, so whatever laws apply here may well be different
anyway.

Richard



signature.asc
Description: OpenPGP digital signature

Re: Has NFS changed recently?

[deloptes]

John wrote:

> I run a mixed Debian/openSuSE /android LAN an after an upgrade of one
> component from jessie to stretch I have hit a big problem.
> 
> Mail is delivered into a mbox on the Debian server, and the disk is
> mounted on the user machine via NFS so the mail can be read into emacs
> from a simple file.  This has worked for many years.  But now with the
> jessie->stretch upgrade it fails with emacs barfing.  I know that the
> emacs did not change so I am guessing that the issue is with NFS.  Any
> possibility of this?

Check the ownership? I recall at some point of time there was something with
the id mapping - also check if you may need to enforce nfs version 3

regards

Has NFS changed recently?

[John]

I run a mixed Debian/openSuSE /android LAN an after an upgrade of one
component from jessie to stretch I have hit a big problem.

Mail is delivered into a mbox on the Debian server, and the disk is
mounted on the user machine via NFS so the mail can be read into emacs
from a simple file.  This has worked for many years.  But now with the
jessie->stretch upgrade it fails with emacs barfing.  I know that the
emacs did not change so I am guessing that the issue is with NFS.  Any
possibility of this?

==John ffitch

Re: Debian Bug Squashing Party (BSP) in Venlo, the Netherlands, 12 and 13 January 2018

[Joost van Baal-Ilić]

Hi,

The Debian Bug Squashing Party earlier this month in Venlo, Netherlands was
productive and fun.  Thanks a lot to Transceptor Technology, part of the
insign.it Group, for hosting[1] us, sponsoring food and being very hospitable
(you might know this company as the people behind the SiriDB[2] time series
database server).

Some of the results:

 Forwarded bugs -- Important bugs (1 bug)

 Pending Upload bugs: 3 bugs
 -- Serious (policy violations or makes package unfit for release) (2 bugs)
 -- Normal bugs (1 bug)
 Resolved bugs: 27 bugs
 -- Grave functionality bugs (1 bug)
 -- Serious (policy violations or makes package unfit for release) (14 bugs)
 -- Important bugs (2 bugs)
 -- Normal bugs (3 bugs)
 -- Wishlist items (7 bugs)

Next to squashing bugs, we spend time discussing upcoming FOSDEM conference in
Brussels, discussing the "Proposal: Repository for fast-paced package
backports" [3], discussing autopkgtest and migrations, and doing some PGP
keysigning.

Participating were Dekkers, elbrus, ivodd, joostvb, mechtilde, michael, Myon,
Natureshadow, stappers, stew and 2 others.

More details at our list of tagged bugs[4], at the wiki[5] and at the gobby
document at gobby.debian.org : BSP/2019/01-Venlo.

Bye,
Joost

[1] 
https://www.insign.it/nieuws/debian-bug-squashing-party-met-transceptor-technology
[2] https://packages.debian.org/siridb-server
[3] 
https://lists.debian.org/msgid-search/20181225204607.gd32...@portux.naturalnet.de
[4] 
https://bugs.debian.org/cgi-bin/pkgreport.cgi?users=debian-rele...@lists.debian.org;tag=bsp-2019-01-nl-venlo
[5] https://wiki.debian.org/BSP/2019/01/nl/Venlo


signature.asc
Description: Digital signature

Re: Disque Dur SSD

[Pascal Hambourg]

Le 27/01/2019 à 12:24, Sébastien Dinot a écrit :


Le chiffrement symétrique est réputé bien plus performant que le
chiffrement assymétrique, mais symétrique ne veut nullement dire que
chiffrement et déchiffrement ont le même cout. Dans la pratique, le
chiffrement se révèle bien plus couteux que le déchiffrement.


Pour quelle(s) raison(s) ? Le découpage en blocs qui implique une 
lecture-modification-écriture en cas de modification partielle d'un bloc 
(comme les bandes en RAID 5) ?



Pour ma part, j'avais effectué des tests avant de le chiffrer et j'avais
constaté les performances suivantes :

Écriture :  800 Mo/s
Lecture :   960 Mo/s

Je viens à l'instant d'effectuer un test rapide sur mon disque chiffré
et j'ai obtenu les résultats suivants :

Écriture :   92 Mo/s
Lecture :   805 Mo/s


Effectivement ça fait très mal...
Une telle chute en écriture me surprend quand même beaucoup. Comment 
as-tu testé pour obtenir ces valeurs ?
Je n'ai pas ce genre de matériel mais j'ai fait un test sur ma petite 
machine avec un volume chiffré LUKS et j'obtiens une chute de 15% en 
lecture et 30% en écriture séquentielle brute avec dd.

Re: IPv6 static config in /etc/network/interfaces ignored

[tomas]

On Sun, Jan 27, 2019 at 12:07:37PM -0500, Ric Moore wrote:
> On 1/27/19 10:29 AM, Claudio M wrote:
> >Hi,
> >
> >I'm wondering if anyone came across this recently, because I can't
> >find any bug reports or posts referring to it online (so I'm
> >wondering if I messed something up inadvertenenp8s0tly).
> >
> 
> Is /dev/eth0 still viable? Mine is now /dev/enp8s0
> ifconfig should tell you. Ric

Minor nit: it's not /dev/eth0 (nor it is /dev/enp8s0), it's just "eth0"
or "enp8s0". Interface names are not file system nodes (I guess it was
a typo on your part, but it may be confusing to some).

Cheers
-- t


signature.asc
Description: Digital signature

Re: Disque Dur SSD

[Pascal Hambourg]

Le 27/01/2019 à 16:17, Basile Starynkevitch a écrit :

On 1/27/19 3:09 PM, Pascal Hambourg wrote:

Le 27/01/2019 à 12:45, Basile Starynkevitch a écrit :


Je trouve la terminologie "garbage collector" dans les SSD  trop 
ambitieuse.


Pourquoi, et quelle appellation te semblerait plus appropriée ?


Personnellement, j'appellerais ça plutôt : compaction, ou 
reorganisation, ou reordonnancement. Un GC est un parcours de graphe, et 
les secteurs d'un SSD ne sont pas organisés en graphe profond (tout au 
plus en arbre ou en peigne).


Tu me parles chinois...

Note qu'on peut mettre un "petit" swap rapide de haute priorité sur le 
SSD pour ne pas effondrer les performances en cas de pic transitoire 
de consommation mémoire et un gros swap lent de plus basse priorité 
sur le disque dur pour l'hibernation occasionnelle.


A mon avis (mais le tien m'interesse et tu es plus compétent), en 


Pas vraiment, non. Ce qui précède le démontre

pratique, pour ce genre de cas, swapper sur un fichier pourrait suffire. 


Si tu veux vraiment mon avis sur l'utilisation d'un fichier comme swap, 
il est bien tranché : c'est un sale hack à éviter car ça repose sur la 
condition que le noyau puisse établir un mapping statique entre le 
fichier de swap et les blocs du périphérique sous-jacent afin d'accéder 
directement à ceux-ci en court-circuitant la couche du système de 
fichiers. Or certains types de systèmes de fichiers (btrfs, nilfs2, et 
je n'ose même pas penser aux systèmes de fichiers non basés sur un 
périphérique bloc comme ubifs) ou de fichiers (creux, préalloués sur 
ext4 ou xfs) ne permettent pas cela. Cf. les pages de manuel de mkswap 
et swapon.


La seule façon propre d'utiliser un fichier comme swap, c'est de 
l'associer à un périphérique loop et d'utiliser ce dernier comme swap, 
pour forcer le noyau à passer par le système de fichiers. Evidemment, ça 
dégrade un peu les performances.

Re: IPv6 static config in /etc/network/interfaces ignored

[Ric Moore]

On 1/27/19 10:29 AM, Claudio M wrote:

Hi,

I'm wondering if anyone came across this recently, because I can't find 
any bug reports or posts referring to it online (so I'm wondering if I 
messed something up inadvertenenp8s0tly).




Is /dev/eth0 still viable? Mine is now /dev/enp8s0
ifconfig should tell you. Ric

Re: IPv6 static config in /etc/network/interfaces ignored

[Pascal Hambourg]

Le 27/01/2019 à 16:29, Claudio M a écrit :


auto eth0
iface eth0 inet static
   address a.b.c.d
   netmask 255.255.255.224
   gateway c.d.e.f
   up route add -net a.b.c.x netmask 255.255.255.224 gw c.d.e.x dev eth0


AFAIK, the gateway must be directly reachable on eth0. But c.d.e.x does 
not seem to belong to a.b.c.d/255.255.255.224.



   up ip addr add f.g.h.i/32 dev eth0
   down ip addr del  f.g.h.i /32 dev eth0


Hackish. You can create a second inet stanza instead.


   post-up /etc/network/iptables.sh


Is there a chance that this script returns a non zero value ?


iface eth0 inet6 static
   address 2a01:::::2
   netmask 64
   gateway fe80::1


What is the output of

 service networking status

Did you try to swap the inet and inet6 stanzas ?

Re: Dependências desencontradas ao instalar o openscad

[Hiure Queiroz]

Olá Antonio, a saida é a seguinte
Lendo listas de pacotes... Pronto
Construindo árvore de dependências
Lendo informação de estado... Pronto
Starting pkgProblemResolver with broken count: 2
Starting 2 pkgProblemResolver with broken count: 2
Investigating (0) openscad:amd64 < none -> 2015.03-2+dfsg-2 @un puN Ib >
Broken openscad:amd64 Depende on libqt5opengl5:amd64 < none |
5.7.1+dfsg-3+b1 @un uH > (>= 5.4)
  Considering libqt5opengl5:amd64 0 as a solution to openscad:amd64 1
Reinst Failed early because of qtbase-abi-5-7-1:amd64
Broken openscad:amd64 Depende on libqt5printsupport5:amd64 < none |
5.7.1+dfsg-3+b1 @un uH > (>= 5.4)
  Considering libqt5printsupport5:amd64 0 as a solution to openscad:amd64
1
Reinst Failed early because of qtbase-abi-5-7-1:amd64
Investigating (0) libqt5scintilla2-12v5:amd64 < none -> 2.9.3+dfsg-4 @un uN
Ib >
Broken libqt5scintilla2-12v5:amd64 Depende on libqt5printsupport5:amd64 <
none | 5.7.1+dfsg-3+b1 @un uH > (>= 5.0.2)
  Considering libqt5printsupport5:amd64 0 as a solution to
libqt5scintilla2-12v5:amd64 0
  Holding Back libqt5scintilla2-12v5:amd64 rather than change
libqt5printsupport5:amd64
Investigating (1) openscad:amd64 < none -> 2015.03-2+dfsg-2 @un puN Ib >
Broken openscad:amd64 Depende on libqt5opengl5:amd64 < none |
5.7.1+dfsg-3+b1 @un uH > (>= 5.4)
  Considering libqt5opengl5:amd64 0 as a solution to openscad:amd64 1
Broken openscad:amd64 Depende on libqt5printsupport5:amd64 < none |
5.7.1+dfsg-3+b1 @un uH > (>= 5.4)
  Considering libqt5printsupport5:amd64 0 as a solution to openscad:amd64
1
Broken openscad:amd64 Depende on libqt5scintilla2-12v5:amd64 < none |
2.9.3+dfsg-4 @un uH > (>= 2.8.4)
  Considering libqt5scintilla2-12v5:amd64 0 as a solution to openscad:amd64
1
Reinst Failed because of libqt5printsupport5:amd64
Done
Alguns pacotes não puderam ser instalados. Isto pode significar que
você solicitou uma situação impossível ou, se você está usando a
distribuição instável, que alguns pacotes requeridos não foram
criados ainda ou foram retirados da "Incoming".
A informação a seguir pode ajudar a resolver a situação:

Os pacotes a seguir têm dependências desencontradas:
 openscad : Depende: libqt5opengl5 (>= 5.4) mas não será instalado
Depende: libqt5printsupport5 (>= 5.4) mas não será instalado
Depende: libqt5scintilla2-12v5 (>= 2.8.4) mas não será instalado
E: Impossível corrigir problemas, você manteve (hold) pacotes quebrados.


Em dom, 27 de jan de 2019 às 14:12, Antonio Terceiro 
escreveu:

> On Sun, Jan 27, 2019 at 08:45:09AM -0200, Hiure Queiroz wrote:
> > Olá pessoal tudo bem?
> >
> > Estou com um problema aqui e gostaria de anunciar aqui para ver se
> encontro
> > alguma ajuda.
> >
> > É uma duvida meio básica, mas pelo tradicional método de busca que
> > normalmente eu tento, não estou conseguindo (que seria buscar na internet
> > por ai o mesmo erro)
> >
> > Estou tentando instalar o openscad do repositório com sudo apt install
> > openscad , mas a seguinte saida:
> >
> > Os pacotes a seguir têm dependências desencontradas:
> >   openscad : Depende: libqt5opengl5 (>= 5.4) mas não será
> > instalado
> >Depende: libqt5printsupport5 (>= 5.4) mas não será
> instalado
> >Depende: libqt5scintilla2-12v5 (>= 2.8.4) mas não será
> > instalado
> >E: Impossível corrigir problemas, você manteve (hold) pacotes
> > quebrados.
> >
> > Então eu tentei instalar o primeiro pacote que se reclama
> >
> > sudo apt install libqt5opengl5
> >  Os pacotes a seguir têm dependências desencontradas:
> >   libqt5opengl5 : Depende: qtbase-abi-5-7-1
> >
> > então tentei sudo apt install qtbase-abi-5-7-1
> > O pacote qtbase-abi-5-7-1 é um pacote virtual fornecido por:
> >   libqt5core5a 5.7.1+dfsg-3+b1[Não é versão candidata]
> >
> > e fiz sudo apt install libqt5core5a
> > libqt5core5a is already the newest version (5.11.1+dfsg-9).
> > libqt5core5a configurado para instalar manualmente.
> >
> >
> > Eu cai nesse looping que não ser como resolver.
>
> mande por favor a saída de
>
> sudo apt -o debug::pkgProblemResolver=yes install openscad
>


-- 
_
Hiure Queiroz
hiurequeiroz.github.io 
Sítio do Astronauta 
  
__

Re: Debian em Blade com storage

[Alex Porto dos Santos]

Uso Proxmox na empresa onde trabalho e indico aos clientes, temos poucos
problemas com ele.
Ele proporciona uso de vms com o kvm e conteiners com LXC, gerencia de
discos e etc tudo em interface gráfica web ou api rest.
Atualmente estudo a implantação de kubernets (docker) como recomendou o
Helio.

Vai depender muito da aplicação que vai rodar nele, e quais os resultados
que espera

Em dom, 27 de jan de 2019 às 10:10, Fábio Rabelo 
escreveu:

>
>
> Em sáb, 26 de jan de 2019 às 08:33, Helio Loureiro 
> escreveu:
>
>> Use openstack ou kubernetes se for possível trabalhar só com containers.
>>
>> OpenStack tem suporte à containers também mas nunca testei.  Dá trabalho
>> instalar e configurar mas o resultado compensa.
>>
>> Promox é coisa pequena.  Serve pra solução caseira.
>>
>
>
> Proxmox ERA "coisa pequena" e "solução caseira" até a versão 2.x ,
>
> Na versão 3.x ele já recebeu uma série de recursos de qualidade
> proficional ( especialmente HA* ! )
>
> Foi este o motivo desta ferramenta ter atraido a atenção do Facebook (
> https://www.facebook.com/notes/protect-the-graph/facebook-capture-the-flag/1466163253623821
> )
>
> A versão 4 ( que já foi desenvolvida após o contato com o facebook )
> tornou a função HA* independente de hardware ( até a versão 3.x só era
> possível HA em hardware de 3 fabricantes : HP. Dell e Supermicro ) .
>
> Na versão 4.x também foram acrescentados os recursos de OVS** e SDI*** (
> OVS** já estava disponível na versão 3.2, mas não tinha uma ferramenta
> confortável para gerenciamento )
>
> Outro recuso adicionado na versão 4.x foram os Containers .
>
> A versão 5.x arredondou. simplificou e melhorou tudo .
>
> Não existe nenhum recurso que possa ser necessário em um Private Cloud que
> o Proxmox não ofereça na própria interface web .
>
>
> Fábio Rabelo
>
>
> * HA = High Availability (
> https://www.linux.com/news/high-availability-allows-business-continuity-says-dietmar-maurer-proxmox-cto
> )
>
> ** OVS = Open vSwitch (
> https://virtualizationreview.com/articles/2014/03/24/proxmox-update-supports-open-vswitch.aspx
> )
>
> *** SDI = Software Defined Infra-structure (
> https://www.hpe.com/br/pt/what-is/software-defined-infrastructure.html )
>
>
>
>
>
>

-- 

*Alex Porto dos Santos *
*Desenvolvedor de integração com equipamentos de rede.*
*skype: alexporto2200*
*telefone e whatsapp: 49 998193546*

Re: Dependências desencontradas ao instalar o openscad

[Antonio Terceiro]

On Sun, Jan 27, 2019 at 08:45:09AM -0200, Hiure Queiroz wrote:
> Olá pessoal tudo bem?
> 
> Estou com um problema aqui e gostaria de anunciar aqui para ver se encontro
> alguma ajuda.
> 
> É uma duvida meio básica, mas pelo tradicional método de busca que
> normalmente eu tento, não estou conseguindo (que seria buscar na internet
> por ai o mesmo erro)
> 
> Estou tentando instalar o openscad do repositório com sudo apt install
> openscad , mas a seguinte saida:
> 
> Os pacotes a seguir têm dependências desencontradas:
>   openscad : Depende: libqt5opengl5 (>= 5.4) mas não será
> instalado
>Depende: libqt5printsupport5 (>= 5.4) mas não será instalado
>Depende: libqt5scintilla2-12v5 (>= 2.8.4) mas não será
> instalado
>E: Impossível corrigir problemas, você manteve (hold) pacotes
> quebrados.
> 
> Então eu tentei instalar o primeiro pacote que se reclama
> 
> sudo apt install libqt5opengl5
>  Os pacotes a seguir têm dependências desencontradas:
>   libqt5opengl5 : Depende: qtbase-abi-5-7-1
> 
> então tentei sudo apt install qtbase-abi-5-7-1
> O pacote qtbase-abi-5-7-1 é um pacote virtual fornecido por:
>   libqt5core5a 5.7.1+dfsg-3+b1[Não é versão candidata]
> 
> e fiz sudo apt install libqt5core5a
> libqt5core5a is already the newest version (5.11.1+dfsg-9).
> libqt5core5a configurado para instalar manualmente.
> 
> 
> Eu cai nesse looping que não ser como resolver.

mande por favor a saída de

sudo apt -o debug::pkgProblemResolver=yes install openscad


signature.asc
Description: PGP signature

IPv6 static config in /etc/network/interfaces ignored

[Claudio M]

Hi,

I'm wondering if anyone came across this recently, because I can't find any
bug reports or posts referring to it online (so I'm wondering if I messed
something up inadvertently).

I have a root server running Stretch 9.7 (fully upgraded) and up until last
night when I rebooted it (previous uptime 66 days and change) everything
was working fine; now, the inet6 interface is not coming up at boot.

Here's my (redacted) /etc/network/interfaces:

> auto lo
> iface lo inet loopback
> iface lo inet6 loopback
>
> auto eth0
> iface eth0 inet static
>   address a.b.c.d
>   netmask 255.255.255.224
>   gateway c.d.e.f
>   up route add -net a.b.c.x netmask 255.255.255.224 gw c.d.e.x dev eth0
>   up ip addr add f.g.h.i/32 dev eth0
>   down ip addr del  f.g.h.i /32 dev eth0
>   post-up /etc/network/iptables.sh
>
> iface eth0 inet6 static
>   address 2a01:::::2
>   netmask 64
>   gateway fe80::1
>

The server boots fine (with the exception of a plethora of error messages
and stuff that doesn't start because configured to use inet6, like
apache2), and my IPv4 comes up normally, but only the link-local IPv6
address is there, not the one in interfaces
If I use "ip -6 a a  dev eth0" and "ip -6 ro a default via
fe80::1" then it works fine.

Checking /var/log/{messages,dmesg,kern.log,daemon.log} for ipv6 only tells
me (repeatedly, at every boot):

> Jan 27 15:25:11 my-server kernel: [0.725253] mip6: Mobile IPv6
> Jan 27 15:25:11 my-server kernel: [9.689430] IPv6:
> ADDRCONF(NETDEV_UP): eth0: link is not ready
> Jan 27 15:25:12 my-server kernel: [   12.739782] IPv6:
> ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
> Jan 27 15:30:20 my-server kernel: [  320.989615] IPv6:
> ADDRCONF(NETDEV_UP): docker0: link is not ready
>

After boot, systemctl restart networking doesn't do anything.
Someone somewhere said "try adding a pre-up sleep 10" directive for a
similar problem: it seemed to work with systemctl restart networking but
didn't do anything at boot time.

The card is
04:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd.
RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 06)

and up until today was using the r8169 driver; I switched to r8168-dkms but
nothing changed.

The system does not have Network Manager.

Does anyone have any suggestions?

Thanks!

Claudio

Re: Disque Dur SSD

[Basile Starynkevitch]

On 1/27/19 3:09 PM, Pascal Hambourg wrote:

Le 27/01/2019 à 12:45, Basile Starynkevitch a écrit :


Je trouve la terminologie "garbage collector" dans les SSD  trop 
ambitieuse.


Pourquoi, et quelle appellation te semblerait plus appropriée ?



Personnellement, j'appellerais ça plutôt : compaction, ou 
reorganisation, ou reordonnancement. Un GC est un parcours de graphe, et 
les secteurs d'un SSD ne sont pas organisés en graphe profond (tout au 
plus en arbre ou en peigne).




La plupart du temps, le swap n'est pas utilisé. La commande free 
indique une utilisation nulle de swap. Comme j'ai du disque et du 
SSD, et comme le SSD est plus rapide que le disque, je prefère le 
consacrer à autre chose.


D'accord. C'est le type d'argument valable et réfléchi auquel je 
m'attendais.



Du coup, je mets la partition de swap sur le SSD.


Sur le disque dur, tu veux dire.



Oui, ma langue a fourché.



Quand j'aurais effectivement un processus bismon qui dépasse les 
100Go (peut-être dans un an) je songerais éventuellement à swapper 
sur un fichier du SSD. Pour l'instant, le swap ne sert quasiment 
jamais, et je ne veux pas gaspiller du SSD pour un gros truc inutile 
comme une grosse zone de swap, alors que j'ai besoin d'accéder à des 
fichiers -parfois un peu volumineux- qui méritent d'être sur le SSD.


Je tiens toutefois à une zone de swap comparable à la taille de ma 
RAM, pour éventuellement pouvoir hiberner mon système (ce que je fais 
en pratique très rarement


Note qu'on peut mettre un "petit" swap rapide de haute priorité sur le 
SSD pour ne pas effondrer les performances en cas de pic transitoire 
de consommation mémoire et un gros swap lent de plus basse priorité 
sur le disque dur pour l'hibernation occasionnelle.


A mon avis (mais le tien m'interesse et tu es plus compétent), en 
pratique, pour ce genre de cas, swapper sur un fichier pourrait suffire. 
Et je mettrais alors ce fichier en priorité haute pour le swap. Pas 
besoin, me semble-t-il, de reserver une partition de swap sur le SSD.


A priori j'ai largement assez de RAM.


Cordialement.

--
Basile STARYNKEVITCH   == http://starynkevitch.net/Basile
opinions are mine only - les opinions sont seulement miennes
Bourg La Reine, France

Re: Can a recipients rights under GNU GPL be revoked?

[Thomas Schmitt]

Hi,

Ben Finney wrote:
> > In other words: Any copyright holder can *say* they wish to
> > retroactively revoke the GNU GPL to some party.

Well, everybody is free to express wishes. But a granted license with no
applicable revocation clause is irrevocable.

The copyright holders alltogether are entitled to grant any license
they can agree on (and that is not illegal or legally void).
E.g. they can grant non-GPL licenses for their GPLed software.

What they cannot do is to revoke granted GPL on published versions.


rhkra...@gmail.com wrote:
> I believe that the 
> original author of a package could do something like create further 
> modifications to the code and create a non-free version of the code.

An example is the cdrecord-wodim fork. The copyright holders did not
release newer versions of cdrecord under GPL. So some concerned Debian
developers used an earlier GPLed version as base of their fork named
wodim.

> Assuming that is correct, people using (or basing modifications) on the 
> (presumably) older free version could continue to use and develop based on 
> that, but would not have rights to that new non-free version.

I agree and practical examples show that we are not alone. 

The copyright of the original authors remains, so that the forkers cannot
change the old license until they replaced all copyrightable imaterial of
the original authors.


Have a nice day :)

Thomas

Re: Can a recipients rights under GNU GPL be revoked?

[John Hasler]

Note that under USA law the right to *use* (including the right to make
such transient and temporary copies as might be required for effective
use) a copy of a program of which one is a legitimate owner is automatic
and requires no license.  Thus if someone gives (or sells) a copy of a
GPL licensed program to you, you are not affected by the GPL until you
make a copy of your copy and then give or sell it to someone else (or
retain a copy after passing the one you received on).

-- 
John Hasler 
jhas...@newsguy.com
Elmwood, WI USA

Re: Disque Dur SSD

[Pascal Hambourg]

Le 27/01/2019 à 12:45, Basile Starynkevitch a écrit :


Je trouve la terminologie "garbage 
collector" dans les SSD  trop ambitieuse.


Pourquoi, et quelle appellation te semblerait plus appropriée ?

La plupart du temps, le swap n'est pas utilisé. La commande free indique 
une utilisation nulle de swap. Comme j'ai du disque et du SSD, et comme 
le SSD est plus rapide que le disque, je prefère le consacrer à autre 
chose.


D'accord. C'est le type d'argument valable et réfléchi auquel je 
m'attendais.



Du coup, je mets la partition de swap sur le SSD.


Sur le disque dur, tu veux dire.

Quand j'aurais effectivement un processus bismon qui dépasse les 100Go 
(peut-être dans un an) je songerais éventuellement à swapper sur un 
fichier du SSD. Pour l'instant, le swap ne sert quasiment jamais, et je 
ne veux pas gaspiller du SSD pour un gros truc inutile comme une grosse 
zone de swap, alors que j'ai besoin d'accéder à des fichiers -parfois un 
peu volumineux- qui méritent d'être sur le SSD.


Je tiens toutefois à une zone de swap comparable à la taille de ma RAM, 
pour éventuellement pouvoir hiberner mon système (ce que je fais en 
pratique très rarement


Note qu'on peut mettre un "petit" swap rapide de haute priorité sur le 
SSD pour ne pas effondrer les performances en cas de pic transitoire de 
consommation mémoire et un gros swap lent de plus basse priorité sur le 
disque dur pour l'hibernation occasionnelle.

Re: Can a recipients rights under GNU GPL be revoked?

[Matthew Crews]

On 1/27/19 6:23 AM, rhkra...@gmail.com wrote:
> I don't disagree with what is stated here (but I have a headache and didn't 
> read it carefully), but, even without reading carefully, I believe that the 
> original author of a package could do something like create further 
> modifications to the code and create a non-free version of the code.
> 
> Assuming that is correct, people using (or basing modifications) on the 
> (presumably) older free version could continue to use and develop based on 
> that, but would not have rights to that new non-free version.

This is correct, and happens all the time.

-Matt

Re: Can a recipients rights under GNU GPL be revoked?

[rhkramer]

On Sunday, January 27, 2019 06:47:05 AM Ben Finney wrote:
> Howdy all,
> 
> Recently in this forum, some concerns have been raised about works
> covered by GNU GPL. In particular, whether a recipient of a work,
> received under conditions of the GNU GPL, can have the freedoms of the
> GNU GPL later withdrawn in that same work.
> 
> To reassure those who might worry whether they can reply on the freedom
> granted in a work, it is worth reading the GNU FAQ document for the GNU
> GPL at the Free Software Foundation:
> 
> [For any GNU GPL-licensed work,] the public already has the right to
> use the program under the GPL, and this right cannot be withdrawn.
> 
>
> https://www.gnu.org/licenses/gpl-faq.en.html#CanDeveloperThirdParty>
> 
> The same answer is in the FAQ specifically for the GNU GPL version 2.0
> https://www.gnu.org/licenses/old-licenses/gpl-2.0-faq.html#CanDevelope
> rThirdParty>.
> 
> You can read more in the Software Freedom Conservancy's document
> _Copyleft and the GNU General Public License: A Comprehensive Tutorial
> and Guide_, specifically in §7.4 “GPLv2 Irrevocability”. That concludes:
> 
> Whether as a matter of a straightforward contractual obligation, or
> as a matter of promissory estoppel, a contributor’s attempt to
> revoke a copyright license grant and then enforce their copyright
> against a user is highly unlikely to succeed.
> 
>
> https://copyleft.org/guide/comprehensive-gpl-guidech8.html#x11-540007
> .4>
> 
> In other words: Any copyright holder can *say* they wish to
> retroactively revoke the GNU GPL to some party. However, unless that
> party has violated the conditions of the GNU GPL grant they originally
> received, there does not appear to be any enforcible threat of
> revocation that would succeed.
> 
> I hope these, along with the many court cases world-wide that have
> tested the GNU GPL and found it to be enforcible, can reassure those
> considering whether a particular copyright holder's whim can revoke the
> freedoms guaranteed in a GNU GPL-covered work. I'd say there's nothing
> to worry about from those threats.

I don't disagree with what is stated here (but I have a headache and didn't 
read it carefully), but, even without reading carefully, I believe that the 
original author of a package could do something like create further 
modifications to the code and create a non-free version of the code.

Assuming that is correct, people using (or basing modifications) on the 
(presumably) older free version could continue to use and develop based on 
that, but would not have rights to that new non-free version.

Re: Debian em Blade com storage

[Fábio Rabelo]

Em sáb, 26 de jan de 2019 às 08:33, Helio Loureiro 
escreveu:

> Use openstack ou kubernetes se for possível trabalhar só com containers.
>
> OpenStack tem suporte à containers também mas nunca testei.  Dá trabalho
> instalar e configurar mas o resultado compensa.
>
> Promox é coisa pequena.  Serve pra solução caseira.
>


Proxmox ERA "coisa pequena" e "solução caseira" até a versão 2.x ,

Na versão 3.x ele já recebeu uma série de recursos de qualidade proficional
( especialmente HA* ! )

Foi este o motivo desta ferramenta ter atraido a atenção do Facebook (
https://www.facebook.com/notes/protect-the-graph/facebook-capture-the-flag/1466163253623821
)

A versão 4 ( que já foi desenvolvida após o contato com o facebook ) tornou
a função HA* independente de hardware ( até a versão 3.x só era possível HA
em hardware de 3 fabricantes : HP. Dell e Supermicro ) .

Na versão 4.x também foram acrescentados os recursos de OVS** e SDI*** (
OVS** já estava disponível na versão 3.2, mas não tinha uma ferramenta
confortável para gerenciamento )

Outro recuso adicionado na versão 4.x foram os Containers .

A versão 5.x arredondou. simplificou e melhorou tudo .

Não existe nenhum recurso que possa ser necessário em um Private Cloud que
o Proxmox não ofereça na própria interface web .


Fábio Rabelo


* HA = High Availability (
https://www.linux.com/news/high-availability-allows-business-continuity-says-dietmar-maurer-proxmox-cto
)

** OVS = Open vSwitch (
https://virtualizationreview.com/articles/2014/03/24/proxmox-update-supports-open-vswitch.aspx
)

*** SDI = Software Defined Infra-structure (
https://www.hpe.com/br/pt/what-is/software-defined-infrastructure.html )

Re: Disque Dur SSD

[Basile Starynkevitch]

On 1/27/19 9:30 AM, Pascal Hambourg wrote:

Le 27/01/2019 à 07:52, Basile Starynkevitch a écrit :


Je crois savoir qu'il est important de monter le SSD avec l'option 
discard qui n'est pas active par défaut.


Apparemment l'option discard n'est pas le moyen recommandé pour 
effectuer le TRIM. Il serait préférable de lancer un "batch TRIM" 
périodique avec la commande fstrim du paquet util-linux. Le paquet 
util-linux inclut dans ses exemples un service systemd qu'on peut 
utiliser à cet effet.

Un grand merci pour le conseil.


Les arguments avancés sont qu'avec certains SSD, l'option discard est 
susceptible de pénaliser les performances (en bloquant les autres 
opérations de lecture/écriture ou en déclenchant un garbage collector 
immédiat) et/ou d'accélérer l'usure du SSD (en déclenchant le garbage 
collector trop souvent).
Je connais bien les GCs au sens logiciel du mot (voir 
http://gchandbook.org/ pour plus de détails), et j'en ai implémenté 
quelques uns (dans Qish, dans GCC MELT, dans bismon) - tous en logiciels 
libres (et il y a plus de 10 ans dans le projet TWO qui est FP5 ou FP6 
européen mais propriétaire). Je trouve la terminologie "garbage 
collector" dans les SSD  trop ambitieuse. Cf 
https://en.wikipedia.org/wiki/Write_amplification#BG-GC



Et ma partition de swap n'est pas sur SSD.



Je précise plusieurs choses:

J'ai des desktops aussi bien au boulot qu'à la maison. Sous Debian/Sid 
dans les deux cas. Avec un SSD et un disque rotatif et deux grands 
écrans dans les deux cas.


Je travaille principalement sur bismon (à temps à peu près plein). C'est 
un logiciel libre (GPLv3+), pas encore publié (not yet released) mais 
dont le code source encore embryonnaire est déjà disponible en 
http://github.com/bstarynk/bismon/ et il évolue constamment. Ce bismon 
est un système persistent et homoiconique. Pour plus de détails, lire 
mon brouillon de rapport (en anglais, futur livrable ou fourniture d'un 
projet européen H2020) 
http://starynkevitch.net/Basile/bismon-chariot-doc.pdf qui évolue encore 
souvent. Ceux qui ont le courage de lire et commenter ma prose sont 
bienvenus (le rapport final est dû en 2020, mais ça m'arrangerait qu'il 
soit acceptable). Bismon est financé par les projets européens H2020 
CHARIOT et DECODER mentionnés dans son README.md


J'ai des machines assez grosses: au boulot, une station Dell 7920 avec 
128Gigaoctets de RAM et un Intel Xeon Silver 4114T à 10 coeurs. A la 
maison, je viens de commander une machine avec 64Gigaoctets de RAM 
(extensible à 128) et un Threadripper 2970WX (et j'attends sa livraison 
avec impatience) à 24 coeurs. En ce moment précis, j'utilise encore un 
vieux i5-4690S avec 32Go de RAM à la maison. Dans deux semaines tout au 
plus, ça va changer.


Je ne fait pas d'hibernation (au sens système du mot), car bismon est 
persistent et donc "hiberne" par lui même. Pour plus de détails, lire 
mon rapport et/ou mon README.


La plupart du temps, le swap n'est pas utilisé. La commande free indique 
une utilisation nulle de swap. Comme j'ai du disque et du SSD, et comme 
le SSD est plus rapide que le disque, je prefère le consacrer à autre 
chose. Du coup, je mais la partition de swap sur le SSD.


Quand j'aurais effectivement un processus bismon qui dépasse les 100Go 
(peut-être dans un an) je songerais éventuellement à swapper sur un 
fichier du SSD. Pour l'instant, le swap ne sert quasiment jamais, et je 
ne veux pas gaspiller du SSD pour un gros truc inutile comme une grosse 
zone de swap, alors que j'ai besoin d'accéder à des fichiers -parfois un 
peu volumineux- qui méritent d'être sur le SSD.


Je tiens toutefois à une zone de swap comparable à la taille de ma RAM, 
pour éventuellement pouvoir hiberner mon système (ce que je fais en 
pratique très rarement: à la maison, mon ordinateur reste allumé 24h/24, 
au boulot je l'allume le matin avant le café; les consignes de sécurité 
incendie déconseillent de laisser un desktop allumé la nuit).


Cordialement, et merci des conseils

--
Basile STARYNKEVITCH   == http://starynkevitch.net/Basile
opinions are mine only - les opinions sont seulement miennes
Bourg La Reine, France

Can a recipients rights under GNU GPL be revoked?

[Ben Finney]

Howdy all,

Recently in this forum, some concerns have been raised about works
covered by GNU GPL. In particular, whether a recipient of a work,
received under conditions of the GNU GPL, can have the freedoms of the
GNU GPL later withdrawn in that same work.

To reassure those who might worry whether they can reply on the freedom
granted in a work, it is worth reading the GNU FAQ document for the GNU
GPL at the Free Software Foundation:

[For any GNU GPL-licensed work,] the public already has the right to
use the program under the GPL, and this right cannot be withdrawn.

https://www.gnu.org/licenses/gpl-faq.en.html#CanDeveloperThirdParty>

The same answer is in the FAQ specifically for the GNU GPL version 2.0
https://www.gnu.org/licenses/old-licenses/gpl-2.0-faq.html#CanDeveloperThirdParty>.

You can read more in the Software Freedom Conservancy's document
_Copyleft and the GNU General Public License: A Comprehensive Tutorial
and Guide_, specifically in §7.4 “GPLv2 Irrevocability”. That concludes:

Whether as a matter of a straightforward contractual obligation, or
as a matter of promissory estoppel, a contributor’s attempt to
revoke a copyright license grant and then enforce their copyright
against a user is highly unlikely to succeed.


https://copyleft.org/guide/comprehensive-gpl-guidech8.html#x11-540007.4>

In other words: Any copyright holder can *say* they wish to
retroactively revoke the GNU GPL to some party. However, unless that
party has violated the conditions of the GNU GPL grant they originally
received, there does not appear to be any enforcible threat of
revocation that would succeed.

I hope these, along with the many court cases world-wide that have
tested the GNU GPL and found it to be enforcible, can reassure those
considering whether a particular copyright holder's whim can revoke the
freedoms guaranteed in a GNU GPL-covered work. I'd say there's nothing
to worry about from those threats.

-- 
 \   “I distrust those people who know so well what God wants them |
  `\to do to their fellows, because it always coincides with their |
_o__)  own desires.” —Susan Brownell Anthony, 1896 |
Ben Finney

Re: Disque Dur SSD

[Sébastien Dinot]

Bonjour,

Pascal Hambourg a écrit :
> Le chiffrement n'est pas symétrique ?

Le chiffrement symétrique est réputé bien plus performant que le
chiffrement assymétrique, mais symétrique ne veut nullement dire que
chiffrement et déchiffrement ont le même cout. Dans la pratique, le
chiffrement se révèle bien plus couteux que le déchiffrement.

Un benchmark que j'ai trouvé sur le net crédite le disque NVME qui
équipe mon portable des performances suivantes :

Écriture :  780 Mo/s
Lecture :  1190 Mo/s

Pour ma part, j'avais effectué des tests avant de le chiffrer et j'avais
constaté les performances suivantes :

Écriture :  800 Mo/s
Lecture :   960 Mo/s

Je viens à l'instant d'effectuer un test rapide sur mon disque chiffré
et j'ai obtenu les résultats suivants :

Écriture :   92 Mo/s
Lecture :   805 Mo/s

Du coup, lorsque j'effectue des compilations intensives d'un gros projet
en C++ comme c'est le cas actuellement, le chiffrement est pénalisant
(le volume du projet - 13 Go de données produites - faisant que tous les
fichiers générés ne peuvent être cachés en mémoire). Mais vu que la
compilation ex-nihilo dudit projet prend 58 minutes, ce ne sont pas les
2 minutes d'écriture des données sur disque qui ralentissent mon
travail...

Et le reste du temps, il serait hypocrite de ma part de dire que je
perçois l'impact du chiffrement du disque.

Sébastien

-- 
Sébastien Dinot, sebastien.di...@free.fr
http://sebastien.dinot.free.fr/
Ne goûtez pas au logiciel libre, vous ne pourriez plus vous en passer !

Dependências desencontradas ao instalar o openscad

[Hiure Queiroz]

Olá pessoal tudo bem?

Estou com um problema aqui e gostaria de anunciar aqui para ver se encontro
alguma ajuda.

É uma duvida meio básica, mas pelo tradicional método de busca que
normalmente eu tento, não estou conseguindo (que seria buscar na internet
por ai o mesmo erro)

Estou tentando instalar o openscad do repositório com sudo apt install
openscad , mas a seguinte saida:

Os pacotes a seguir têm dependências desencontradas:
  openscad : Depende: libqt5opengl5 (>= 5.4) mas não será
instalado
   Depende: libqt5printsupport5 (>= 5.4) mas não será instalado
   Depende: libqt5scintilla2-12v5 (>= 2.8.4) mas não será
instalado
   E: Impossível corrigir problemas, você manteve (hold) pacotes
quebrados.

Então eu tentei instalar o primeiro pacote que se reclama

sudo apt install libqt5opengl5
 Os pacotes a seguir têm dependências desencontradas:
  libqt5opengl5 : Depende: qtbase-abi-5-7-1

então tentei sudo apt install qtbase-abi-5-7-1
O pacote qtbase-abi-5-7-1 é um pacote virtual fornecido por:
  libqt5core5a 5.7.1+dfsg-3+b1[Não é versão candidata]

e fiz sudo apt install libqt5core5a
libqt5core5a is already the newest version (5.11.1+dfsg-9).
libqt5core5a configurado para instalar manualmente.


Eu cai nesse looping que não ser como resolver.

Meu soruce.list é o seguinte

deb http://ftp.br.debian.org/debian stretch main contrib non-free
deb-src http://ftp.br.debian.org/debian stretch main contrib non-free

deb http://ftp.br.debian.org/debian stretch-updates main contrib non-free
deb-src http://deb.debian.org/debian stretch-updates main contrib non-free

deb http://security.debian.org/debian-security/ stretch/updates main
contrib non-free
deb-src http://security.debian.org/debian-security/ stretch/updates main
contrib non-free

 deb http://ftp.br.debian.org/debian/ stretch-backports main contrib
non-free

Obrigado a atenção
Hiure



-- 
_
Hiure Queiroz
hiurequeiroz.github.io 
Sítio do Astronauta 
  
__

Re: Disque Dur SSD

[Pascal Hambourg]

Le 25/01/2019 à 11:27, Sébastien Dinot a écrit :


J'ai une partition de swap sur mon disque NVMe, mais vu le niveau de performance de ce disque (surtout en lecture en ce qui me concerne, vu qu'en écriture, le débit est sérieusement ralenti par la capacité de chiffrement du processeur), 


Le chiffrement n'est pas symétrique ?

Re: Disque Dur SSD

[Pascal Hambourg]

Le 25/01/2019 à 08:56, aishen a écrit :
Pas seulement, au début linux/unix on mettait de la swap pour avoir une 
mémoire disque plus rapide, donc le ssd était presque de la RAM,


Pardon ?
"De la swap pour avoir une mémoire disque plus rapide", qu'est-ce que ça 
veut dire ?

Un SSD aux débuts de Linux/Unix ?

chacun fait ce qu'il veut mais ne devrait pas être traité 
d'absurde ou idiot


Qui a été traité d'absurde ou d'idiot dans cette discussion ?

Re: Disque Dur SSD

[Pascal Hambourg]

Le 25/01/2019 à 08:38, Daniel Caillibaud a écrit :

Le 25/01/19 à 07:35, aishen  a écrit :

on n'a pas besoin de swap sur un ssd linux,



Quel rapport entre le besoin de swap (manque de RAM du système) et la
nature du disque ?


d'après les avis il est
même déconseillé d'en mettre pour éviter des lectures/écritures
constantes qui "usent" la mémoire des ssd


La lecture n'use pas les SSD. Et avoir un swap n'implique pas qu'il va y 
avoir des écritures constantes. Au contraire cela peut diminuer le 
volume d'écritures par une meilleure gestion des caches disque.


La situation où le système lit et écrit constamment dans le swap 
(thrashing) correspond à un manque de mémoire critique pour 
l'utilisation. Sans swap, le système ou des processus planteraient. Avec 
du swap sur disque dur, le système serait tellement lent qu'il en 
deviendrait inutilisable. Avec du swap sur SSD, au moins le système 
reste utilisable dans ces conditions. Mais la vraie réponse, c'est 
d'augmenter la RAM.



Le swap, c'est fait pour que le système ne crash pas quand il manque de
RAM, il utilise alors le disque.


Pas seulement. Le swap sert aussi à décharger de la mémoire les données 
des processus qui sont rarement accédées au profit des caches disque, 
dans le but d'améliorer les performances. On peut envisager qu'avec un 
SSD très rapide (type NVMe), l'écart réduit entre la vitesses de la RAM 
et la vitesse du stockage rend le besoin d'un gros cache disque moins fort.

Re: Disque Dur SSD

[Pascal Hambourg]

Le 27/01/2019 à 07:52, Basile Starynkevitch a écrit :


Je crois savoir qu'il est important de monter 
le SSD avec l'option discard qui n'est pas active par défaut.


Apparemment l'option discard n'est pas le moyen recommandé pour 
effectuer le TRIM. Il serait préférable de lancer un "batch TRIM" 
périodique avec la commande fstrim du paquet util-linux. Le paquet 
util-linux inclut dans ses exemples un service systemd qu'on peut 
utiliser à cet effet.


Les arguments avancés sont qu'avec certains SSD, l'option discard est 
susceptible de pénaliser les performances (en bloquant les autres 
opérations de lecture/écriture ou en déclenchant un garbage collector 
immédiat) et/ou d'accélérer l'usure du SSD (en déclenchant le garbage 
collector trop souvent).



Et ma partition de swap n'est pas sur SSD.


Pourquoi ?