Re: [gentoo-user] Re: *dev-less gentoo

[waltdnes]

On Tue, Jan 19, 2016 at 10:05:52PM +0100, k...@aspodata.se wrote
> Alec McKinnon:
> > On 19/01/2016 18:51, k...@aspodata.se wrote:
> ...
> > > I have had no pain useing an old plain /dev. What's the pain ?
> > take a machine running a desktop. Plug in a usb printer. Where's your node?
> 
> To find that out I'd investigate /sys/bus/usb, either directly or via 
> usb-devices or some other program. I guess "some other program" is
> probably udev or similar for you, it might not be for me.
> 
> If it is a usb disk, I just look at the output of sg_map -x -i, and then
> decide what to do.
> 
> > That's the whole point of a dynamic dev manager, it responds to devices
> > changes that occur on normal modern machines and does TheRightThing(tm)
> > - currently defined as whatever the dev-manager config tells it to do.
> 
> Ok, I don't have any usb printer, all my printers are network connected 
> and do handle postscript and lpd.
> 
> And my "dev-manager" tells the system to do nothing till the owner of 
> the system tells it to do so, which is the right thing for me.
> The right thing might be something else for you.

  I'm the -disturber who started the following wiki entries.  There
have since been many contributions by other users...

https://wiki.gentoo.org/wiki/Mdev
https://wiki.gentoo.org/wiki/Mdev/Automount_USB
https://wiki.gentoo.org/wiki/Mdev/Automount_USB/automount

  A minimal system running busybox probably also has mdev built in.
It's "dynamic" and can automount if desired.  But if you tell it not to
automount, it won't.

-- 
Walter Dnes 
I don't run "desktop environments"; I run useful applications

[gentoo-user] Re: *dev-less gentoo

[James]

  aspodata.se> writes:


> I'm new to gentoo, is there some special semantic to the "bgo #" ?
WELCOME Karl!

You'll find gentoo is full of traditional *nix users and minimalists.
Don't let the progressives disturb your reticent ways... you are in good
company.

https://bugs.gentoo.org/show_bug.cgi?id=107875



> I have had no pain useing an old plain /dev. What's the pain ?

I try to do to many things, and often get confused on what use to work,
what may work and what might work. 



> > For explicit clarity, you've got a "/dev" from using dev-manager on the
> > system previously, and now you desire to switch to a static-dev? (Why ?)
> >  Or did you derive from scratch (or other means) a '/dev' for a specific
> > need you are working on by design, historical example etc?
>
> No, I never used udev et al on my boxes, there has simply been no need.


On workstations, I use openrc and eudev; but I work on many many different
devices that have /dev issues, particularly in an embedded platform.


> > I apologize in advance, but this thread intersects some critical new
> > thinking on systems cluster formation. I have ran into a small group of
> > extraordinary coders that are building a Hi Performance Cluster out of C
> > C Rust and a minimized static-dev.  So I am very curious as to your  
> > specific and detailed motives for this 'static-dev'. If a private note >
> is warranted, feel encourage for that type of response. If this 
> > unbounded curiosity of mine is unwelcome, you have my deepest apologies.

> I never had any compelling reason to let some daemon with mess with
> /dev. And I have had a compelling reason to avoid it, when doing an
> "usual" stable dist-upgrade of Debian lenny to squeze (I think), Debian
> installed udev per default and everything just stopped working. And
> that darn thing wouldn't uninstall and /dev wouldn't unmount to get
> back my /dev-entries. So udev have only giving me pain and no gain.
> The only thing dynamic theese days are usb. Usb disks I can handle
> manually, usb kbd/mouse has always worked. I usually don't use more
> than one keyboard so I don't really need xkb, nor do I need something
> to autodetect keyboard layout, since I change it to something else 
> anyhow. And udev woun't detect my serial mouse anyhow... so much for
> that.


Excellent! I like you quite a lot, Karl. Just so you know, folks are
encouraged to maintain ebuilds here at Gentoo, via the proxy maintainer
project. So if you find an orphaned packaged you like (equery m )
then just drop them an email. [1]  It's a good way to help customize gentoo
in a move-at-your-pace platform.


> That said, if I would like to test some "dev-manager" (except myself)
> than I'd look into something that behaves nicely, like mdev (busybox)
> or vdev (https://github.com/jcnelson/vdev.git).

With gentoo, all things are possible. eudev + openrc might be a pleasant
combo and they are both "gentoo source projects" or close enough. The folks
that work on those are quite busy and would most likely welcome your input
and participate. On the desktop, you might like lxqt or lxde?

If you find codes you like, and there is no ebuild for them (eix -R
), then you can easily create an ebuild, most of the time. Complex
dependency are a bit trickier. Check out the gentoo devmanual .

> Regards,
> /Karl Hammar


Thanks for the info! ;-)

James

[1] https://wiki.gentoo.org/wiki/Project:Proxy_Maintainers

Re: [gentoo-user] Re: *dev-less gentoo

[karl]

Alec McKinnon:
> On 19/01/2016 18:51, k...@aspodata.se wrote:
...
> > I have had no pain useing an old plain /dev. What's the pain ?
> take a machine running a desktop. Plug in a usb printer. Where's your node?

To find that out I'd investigate /sys/bus/usb, either directly or via 
usb-devices or some other program. I guess "some other program" is
probably udev or similar for you, it might not be for me.

If it is a usb disk, I just look at the output of sg_map -x -i, and then
decide what to do.

> That's the whole point of a dynamic dev manager, it responds to devices
> changes that occur on normal modern machines and does TheRightThing(tm)
> - currently defined as whatever the dev-manager config tells it to do.

Ok, I don't have any usb printer, all my printers are network connected 
and do handle postscript and lpd.

And my "dev-manager" tells the system to do nothing till the owner of 
the system tells it to do so, which is the right thing for me.
The right thing might be something else for you.

> I'm having a hard time thinking what kind of machine you have in this
> day and age that can do mail and also does not need a dynamic device
> maanger. Please enlighten us, or are you perhaps using MAKEDEV?

Please be aware of that I'm not impling anything about anyone else than 
me och don't ridicule me.

To do mail, all you have to have is a network connection, a mail 
program and a mail server to relay through. All of that has been done 
for ages without any program like udev. So I don't understand why you 
have any problem understanding how that is done, or why you choose such
an example.

And I don't use MAKEDEV, the dev-nodes are already there, there is no
need to create them again. What's the fuss ?

...
> Sounds like you made one mistake once and that has now become the world
> for you. Almost no-one else here has reported dynamic dev managers make
> "everything just stop working". What you will hear is lots of whinging
> about udev - actually it's whinging about udev's maintainers cleverly
> disguised as whinging about the software - but as a class of software
> they all get the job done and do it well.

No, I did not do the mistake, the upgrade program or the udev 
installer did. And since udev (or something related to it) mounts
something on /dev, which makes it in practice inpossible to unmount.
So if udev do not fill up the new fs correctly, the system is hosed,
yea, unless I value running mknod by hand and from memory.
That very problem I had is probably fixed by now. But I don't see the
need to get exposed to it again. If udev had used e.g. /udev and
populated that dir seperately from /dev, I would not have that special
problem. udev seems to be hardcoded to /dev, but other similar program
are more malleable in this regard, and if need arises I wouldn't 
hesitate to test them.

Regards,
/Karl Hammar

---
Aspö Data
Lilla Aspö 148
S-742 94 Östhammar
Sweden
+46 173 140 57

Re: [gentoo-user] Re: *dev-less gentoo

[Alan McKinnon]

On 19/01/2016 18:51, k...@aspodata.se wrote:
> James:
>>   aspodata.se> writes:
> I found a workaround in the sys-fs/static-dev package.
>>
>> Interesting read :: bgo #107875
> 
> I'm new to gentoo, is there some special semantic to the "bgo #" ?
> 
 Let's be clear: static-dev is NOT a workaround. It is a full proper
 solution for the case when a dynamic device node solution is not 
 desired.
>> Well, I can think of embedded (linux) systems, a lock-down server and
>> machine(s) loaded up with (NFV) Network Function Virtuals, as prime examples
>> where a static dev is very useful; albeit a management pain if one is not
>> careful. This is a very interesting topic for me.
> 
> I have had no pain useing an old plain /dev. What's the pain ?


take a machine running a desktop. Plug in a usb printer. Where's your node?

That's the whole point of a dynamic dev manager, it responds to devices
changes that occur on normal modern machines and does TheRightThing(tm)
- currently defined as whatever the dev-manager config tells it to do.

I'm having a hard time thinking what kind of machine you have in this
day and age that can do mail and also does not need a dynamic device
maanger. Please enlighten us, or are you perhaps using MAKEDEV?

> 
 Of course it means you have to mknod every device you need yourself. But
 you know that going in right?
>>
>>> Yes (though I alreade have a /dev from before).
>>
>> For explicit clarity, you've got a "/dev" from using dev-manager on the
>> system previously, and now you desire to switch to a static-dev? (Why ?)
>>  Or did you derive from scratch (or other means) a '/dev' for a specific
>> need you are working on by design, historical example etc?
> 
> No, I never used udev et al on my boxes, there has simply been no need.
> 
>> I apologize in advance, but this thread intersects some critical new
>> thinking on systems cluster formation. I have ran into a small group of
>> extraordinary coders that are building a Hi Performance Cluster out of C,
>> Rust and a minimized static-dev.  So I am very curious as to your specific
>> and detailed motives for this 'static-dev'. If a private note is warranted,
>> feel encourage for that type of response. If this unbounded curiosity of
>> mine is unwelcome, you have my deepest apologies.
> 
> I never had any compelling reason to let some daemon with mess with
> /dev. And I have had a compelling reason to avoid it, when doing an
> "usual" stable dist-upgrade of Debian lenny to squeze (I think), Debian
> installed udev per default and everything just stopped working. And
> that darn thing wouldn't uninstall and /dev wouldn't unmount to get
> back my /dev-entries. So udev have only giving me pain and no gain.
> The only thing dynamic theese days are usb. Usb disks I can handle
> manually, usb kbd/mouse has always worked. I usually don't use more
> than one keyboard so I don't really need xkb, nor do I need something
> to autodetect keyboard layout, since I change it to something else 
> anyhow. And udev woun't detect my serial mouse anyhow... so much for
> that.
> 
> That said, if I would like to test some "dev-manager" (except myself)
> than I'd look into something that behaves nicely, like mdev (busybox)
> or vdev (https://github.com/jcnelson/vdev.git).

Sounds like you made one mistake once and that has now become the world
for you. Almost no-one else here has reported dynamic dev managers make
"everything just stop working". What you will hear is lots of whinging
about udev - actually it's whinging about udev's maintainers cleverly
disguised as whinging about the software - but as a class of software
they all get the job done and do it well.

-- 
Alan McKinnon
alan.mckin...@gmail.com

Re: [gentoo-user] Re: *dev-less gentoo

[Neil Bothwick]

On Tue, 19 Jan 2016 16:06:26 + (UTC), James wrote:

> > > Let's be clear: static-dev is NOT a workaround. It is a full proper
> > > solution for the case when a dynamic device node solution is not 
> > > desired.  
> 
> Well, I can think of embedded (linux) systems, a lock-down server and
> machine(s) loaded up with (NFV) Network Function Virtuals, as prime
> examples where a static dev is very useful; albeit a management pain if
> one is not careful. This is a very interesting topic for me.

Whatever your setup, you need something to manage your entries in /dev.
That's why there is a dependency on the dev-manager/virtual. What you use
is up to you: udev, eudev, systemd, devfsd, busybox or doing it manually,
is up to you. That's why any of those satisfy the dev-manager virtual.
That's why Alan said that static-dev is not a work around, it is a valid
choice that sets up a limited number of static nodes that you then manage
yourself. You are the dev-manager.


-- 
Neil Bothwick

Don't judge a book by its movie.


pgpsuCuIO7vT_.pgp
Description: OpenPGP digital signature

[gentoo-user] Re: *dev-less gentoo

[James]

  aspodata.se> writes:


> > > I found a workaround in the sys-fs/static-dev package.

Interesting read :: bgo #107875

> > Let's be clear: static-dev is NOT a workaround. It is a full proper
> > solution for the case when a dynamic device node solution is not 
> > desired.

Well, I can think of embedded (linux) systems, a lock-down server and
machine(s) loaded up with (NFV) Network Function Virtuals, as prime examples
where a static dev is very useful; albeit a management pain if one is not
careful. This is a very interesting topic for me.


> > Of course it means you have to mknod every device you need yourself. But
> > you know that going in right?

> Yes (though I alreade have a /dev from before).

For explicit clarity, you've got a "/dev" from using dev-manager on the
system previously, and now you desire to switch to a static-dev? (Why ?)
 Or did you derive from scratch (or other means) a '/dev' for a specific
need you are working on by design, historical example etc?


I apologize in advance, but this thread intersects some critical new
thinking on systems cluster formation. I have ran into a small group of
extraordinary coders that are building a Hi Performance Cluster out of C,
Rust and a minimized static-dev.  So I am very curious as to your specific
and detailed motives for this 'static-dev'. If a private note is warranted,
feel encourage for that type of response. If this unbounded curiosity of
mine is unwelcome, you have my deepest apologies.


curiously,
James

Re: [gentoo-user] Re: *dev-less gentoo

[Alec Ten Harmsel]

On Tue, Jan 19, 2016 at 05:51:11PM +0100, k...@aspodata.se wrote:
> James:
> >   aspodata.se> writes:
> > > > > I found a workaround in the sys-fs/static-dev package.
> > 
> > Interesting read :: bgo #107875
> 
> I'm new to gentoo, is there some special semantic to the "bgo #" ?

bgo == https://bugs.gentoo.org

Alec

Re: [gentoo-user] Re: *dev-less gentoo

[karl]

James:
>   aspodata.se> writes:
> > > > I found a workaround in the sys-fs/static-dev package.
> 
> Interesting read :: bgo #107875

I'm new to gentoo, is there some special semantic to the "bgo #" ?

> > > Let's be clear: static-dev is NOT a workaround. It is a full proper
> > > solution for the case when a dynamic device node solution is not 
> > > desired.
> Well, I can think of embedded (linux) systems, a lock-down server and
> machine(s) loaded up with (NFV) Network Function Virtuals, as prime examples
> where a static dev is very useful; albeit a management pain if one is not
> careful. This is a very interesting topic for me.

I have had no pain useing an old plain /dev. What's the pain ?

> > > Of course it means you have to mknod every device you need yourself. But
> > > you know that going in right?
> 
> > Yes (though I alreade have a /dev from before).
> 
> For explicit clarity, you've got a "/dev" from using dev-manager on the
> system previously, and now you desire to switch to a static-dev? (Why ?)
>  Or did you derive from scratch (or other means) a '/dev' for a specific
> need you are working on by design, historical example etc?

No, I never used udev et al on my boxes, there has simply been no need.

> I apologize in advance, but this thread intersects some critical new
> thinking on systems cluster formation. I have ran into a small group of
> extraordinary coders that are building a Hi Performance Cluster out of C,
> Rust and a minimized static-dev.  So I am very curious as to your specific
> and detailed motives for this 'static-dev'. If a private note is warranted,
> feel encourage for that type of response. If this unbounded curiosity of
> mine is unwelcome, you have my deepest apologies.

I never had any compelling reason to let some daemon with mess with
/dev. And I have had a compelling reason to avoid it, when doing an
"usual" stable dist-upgrade of Debian lenny to squeze (I think), Debian
installed udev per default and everything just stopped working. And
that darn thing wouldn't uninstall and /dev wouldn't unmount to get
back my /dev-entries. So udev have only giving me pain and no gain.
The only thing dynamic theese days are usb. Usb disks I can handle
manually, usb kbd/mouse has always worked. I usually don't use more
than one keyboard so I don't really need xkb, nor do I need something
to autodetect keyboard layout, since I change it to something else 
anyhow. And udev woun't detect my serial mouse anyhow... so much for
that.

That said, if I would like to test some "dev-manager" (except myself)
than I'd look into something that behaves nicely, like mdev (busybox)
or vdev (https://github.com/jcnelson/vdev.git).

Regards,
/Karl Hammar

---
Aspö Data
Lilla Aspö 148
S-742 94 Östhammar
Sweden
+46 173 140 57

Re: [gentoo-user] Re: *dev-less gentoo

[Alan McKinnon]

On 18/01/2016 23:05, k...@aspodata.se wrote:
> boxc...@gmx.net:
>> On Mon, 18 Jan 2016 19:48:58 +0100 (CET)
>> k...@aspodata.se wrote:
> ...
>>> What info is there on @system ?
>>> I can change what's in @world, it seems to be the content of
>>> /var/lib/portage/world. Is there a similar file for @system ?
>>
>> It's in /usr/portage/profiles/base/packages -- I think that will be
>> overwritten when profiles are updated, so I don't think it helps you.
> 
> Great, thanks. Would it work to have that in an overlay, ehh, something ?
> 
>> Would putting virtual/dev-manager into packages.provided work to solve
>> your problem?  (I phrase it as a question because I've never used
>> packages.provided.)
> 
> No, then would packages that actually needs it be fooled.
> I found a workaround in the sys-fs/static-dev package.


Let's be clear: static-dev is NOT a workaround. It is a full proper
solution for the case when a dynamic device node solution is not desired.

Of course it means you have to mknod every device you need yourself. But
you know that going in right?


-- 
Alan McKinnon
alan.mckin...@gmail.com

Re: [gentoo-user] Re: *dev-less gentoo

[karl]

Alan McKinnon:
> On 18/01/2016 23:05, k...@aspodata.se wrote:
...
> > I found a workaround in the sys-fs/static-dev package.
> Let's be clear: static-dev is NOT a workaround. It is a full proper
> solution for the case when a dynamic device node solution is not desired.

Ok, fine with me (the wording "dev-manager" go me off track).

> Of course it means you have to mknod every device you need yourself. But
> you know that going in right?

Yes (though I alreade have a /dev from before).

Regards,
/Karl Hammar

---
Aspö Data
Lilla Aspö 148
S-742 94 Östhammar
Sweden
+46 173 140 57

Re: [gentoo-user] Re: *dev-less gentoo

[karl]

boxc...@gmx.net:
> On Mon, 18 Jan 2016 19:48:58 +0100 (CET)
> k...@aspodata.se wrote:
...
> > What info is there on @system ?
> > I can change what's in @world, it seems to be the content of
> > /var/lib/portage/world. Is there a similar file for @system ?
> 
> It's in /usr/portage/profiles/base/packages -- I think that will be
> overwritten when profiles are updated, so I don't think it helps you.

Great, thanks. Would it work to have that in an overlay, ehh, something ?

> Would putting virtual/dev-manager into packages.provided work to solve
> your problem?  (I phrase it as a question because I've never used
> packages.provided.)

No, then would packages that actually needs it be fooled.
I found a workaround in the sys-fs/static-dev package.

Regards,
/Karl Hammar

---
Aspö Data
Lilla Aspö 148
S-742 94 Östhammar
Sweden
+46 173 140 57

[gentoo-user] Re: *dev-less gentoo

[»Q«]

On Mon, 18 Jan 2016 19:48:58 +0100 (CET)
k...@aspodata.se wrote:

> Francisco Ares:
> > 2016-01-18 15:15 GMT-02:00 :
> >   
> > > # emerge -auDN @system
> > > ...
> > > [ebuild  N ] virtual/dev-manager-0
> > >
> > > How can I get rid of dev-manager-0 from @system ?  
> ...
> > Try updating to a new kernel.
> > 
> > I'm saying this because of the output of equery d
> > virtual/dev-manager on my system:
> > 
> >  ~ $ equery d virtual/dev-manager
> >  * These packages depend on virtual/dev-manager:
> > sys-kernel/gentoo-sources-3.18.9 (virtual/dev-manager)
> > sys-kernel/gentoo-sources-3.18.12 (virtual/dev-manager)  
> 
> Not so here:
> 
> # equery d virtual/dev-manager
>  * These packages depend on virtual/dev-manager:
> #
> 
> ///
> 
> What info is there on @system ?
> I can change what's in @world, it seems to be the content of
> /var/lib/portage/world. Is there a similar file for @system ?

It's in /usr/portage/profiles/base/packages -- I think that will be
overwritten when profiles are updated, so I don't think it helps you.

Would putting virtual/dev-manager into packages.provided work to solve
your problem?  (I phrase it as a question because I've never used
packages.provided.)