[gentoo-dev] Global Systemd USE Flag

[gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

Hi,

Sorry if this has been discussed to death, but I couldn't find any
definitive decisions on it, so I thought I'd mention things in a
fairly simple manner:

Step 1: I use OpenRC/sysvinit.

Dell ~ # readlink -f /proc/1/exe
/sbin/init
Dell ~ # equery b /sbin/init
 * Searching for /sbin/init ...
sys-apps/sysvinit-2.88-r3 (/sbin/init)


Step 2: There are lots of systemd service files installed.

Dell ~ # ls /usr/lib/systemd/system/*.service|wc -l
21


Step 3: What on earth is installing them?

Dell ~ # equery b /usr/lib/systemd/system/*.service
media-libs/libcanberra-0.29
(/usr/lib/systemd/system/canberra-system-shutdown-reboot.service)
media-libs/libcanberra-0.29
(/usr/lib/systemd/system/canberra-system-bootup.service)
media-libs/libcanberra-0.29
(/usr/lib/systemd/system/canberra-system-shutdown.service)
media-sound/alsa-utils-1.0.25-r2 (/usr/lib/systemd/system/alsa-restore.service)
media-sound/alsa-utils-1.0.25-r2 (/usr/lib/systemd/system/alsa-store.service)
net-misc/dhcpcd-5.5.6 (/usr/lib/systemd/system/dhcpcd.service)
net-misc/openssh-6.0_p1-r1 (/usr/lib/systemd/system/sshd@.service)
net-misc/openssh-6.0_p1-r1 (/usr/lib/systemd/system/sshd.service)
net-wireless/bluez-4.101-r1 (/usr/lib/systemd/system/bluetooth.service)
net-wireless/wpa_supplicant-1.0 (/usr/lib/systemd/system/wpa_supplicant.service)
net-wireless/wpa_supplicant-1.0
(/usr/lib/systemd/system/wpa_supplicant@.service)
sys-apps/dbus-1.6.4 (/usr/lib/systemd/system/dbus.service)
sys-auth/consolekit-0.4.5_p20120320
(/usr/lib/systemd/system/console-kit-log-system-start.service)
sys-auth/consolekit-0.4.5_p20120320
(/usr/lib/systemd/system/console-kit-daemon.service)
sys-auth/consolekit-0.4.5_p20120320
(/usr/lib/systemd/system/console-kit-log-system-restart.service)
sys-auth/consolekit-0.4.5_p20120320
(/usr/lib/systemd/system/console-kit-log-system-stop.service)
sys-auth/polkit-0.107 (/usr/lib/systemd/system/polkit.service)
sys-fs/udev-187-r1 (/usr/lib/systemd/system/systemd-udev-trigger.service)
sys-fs/udev-187-r1 (/usr/lib/systemd/system/systemd-udev-settle.service)
sys-fs/udev-187-r1 (/usr/lib/systemd/system/systemd-udevd.service)
sys-power/upower-0.9.17-r1 (/usr/lib/systemd/system/upower.service)


Yowza! All the packages that provide systemd unit files are installing
them?! But I don't even use systemd. I don't want this cruft on my
system.

Proposal: global USE flag for systemd, just like there's one for openrc.


Thanks,
Jason

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 2829 bytes --]

On Wed, 8 Aug 2012 15:11:42 +0200
"Jason A. Donenfeld" <Jason@zx2c4.com> wrote:

> Sorry if this has been discussed to death, but I couldn't find any
> definitive decisions on it, so I thought I'd mention things in a
> fairly simple manner:
> 
> Step 1: I use OpenRC/sysvinit.
> 
> Dell ~ # readlink -f /proc/1/exe
> /sbin/init
> Dell ~ # equery b /sbin/init
>  * Searching for /sbin/init ...
> sys-apps/sysvinit-2.88-r3 (/sbin/init)
> 
> 
> Step 2: There are lots of systemd service files installed.
> 
> Dell ~ # ls /usr/lib/systemd/system/*.service|wc -l
> 21
> 
> 
> Step 3: What on earth is installing them?
> 
> Dell ~ # equery b /usr/lib/systemd/system/*.service
> media-libs/libcanberra-0.29
> (/usr/lib/systemd/system/canberra-system-shutdown-reboot.service)
> media-libs/libcanberra-0.29
> (/usr/lib/systemd/system/canberra-system-bootup.service)
> media-libs/libcanberra-0.29
> (/usr/lib/systemd/system/canberra-system-shutdown.service)
> media-sound/alsa-utils-1.0.25-r2
> (/usr/lib/systemd/system/alsa-restore.service)
> media-sound/alsa-utils-1.0.25-r2
> (/usr/lib/systemd/system/alsa-store.service) net-misc/dhcpcd-5.5.6
> (/usr/lib/systemd/system/dhcpcd.service) net-misc/openssh-6.0_p1-r1
> (/usr/lib/systemd/system/sshd@.service) net-misc/openssh-6.0_p1-r1
> (/usr/lib/systemd/system/sshd.service) net-wireless/bluez-4.101-r1
> (/usr/lib/systemd/system/bluetooth.service)
> net-wireless/wpa_supplicant-1.0
> (/usr/lib/systemd/system/wpa_supplicant.service)
> net-wireless/wpa_supplicant-1.0
> (/usr/lib/systemd/system/wpa_supplicant@.service) sys-apps/dbus-1.6.4
> (/usr/lib/systemd/system/dbus.service)
> sys-auth/consolekit-0.4.5_p20120320
> (/usr/lib/systemd/system/console-kit-log-system-start.service)
> sys-auth/consolekit-0.4.5_p20120320
> (/usr/lib/systemd/system/console-kit-daemon.service)
> sys-auth/consolekit-0.4.5_p20120320
> (/usr/lib/systemd/system/console-kit-log-system-restart.service)
> sys-auth/consolekit-0.4.5_p20120320
> (/usr/lib/systemd/system/console-kit-log-system-stop.service)
> sys-auth/polkit-0.107 (/usr/lib/systemd/system/polkit.service)
> sys-fs/udev-187-r1
> (/usr/lib/systemd/system/systemd-udev-trigger.service)
> sys-fs/udev-187-r1
> (/usr/lib/systemd/system/systemd-udev-settle.service)
> sys-fs/udev-187-r1 (/usr/lib/systemd/system/systemd-udevd.service)
> sys-power/upower-0.9.17-r1 (/usr/lib/systemd/system/upower.service)
> 
> 
> Yowza! All the packages that provide systemd unit files are installing
> them?! But I don't even use systemd. I don't want this cruft on my
> system.
> 
> Proposal: global USE flag for systemd, just like there's one for
> openrc.

INSTALL_MASK=/usr/lib/systemd

And live happy to the day you notice your system no longer boots.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

On Wed, Aug 8, 2012 at 4:15 PM, Michał Górny <mgorny@gentoo.org> wrote:
> INSTALL_MASK=/usr/lib/systemd
>
> And live happy to the day you notice your system no longer boots.

This is a nice bandaid, and sure, it "solves" the immediate issue...
but it doesn't actually solve the actual issue: when packages
optionally install unwanted bloat, we make them an option via a USE
flag. In this case, especially, since systemd isn't even the default
(nor officially supported, whatever that amounts to), users certainly
should not have to manually add an install mask to make portage do
what it already should do.

Besides, as systemd gains momentum, we can probably expect that
various pieces of software will have options to enable a systemd mode
or a systemd build, or what have you, and then in this case, a global
USE flag becomes even more imperative.

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 1418 bytes --]

On Wed, 8 Aug 2012 16:20:55 +0200
"Jason A. Donenfeld" <Jason@zx2c4.com> wrote:

> On Wed, Aug 8, 2012 at 4:15 PM, Michał Górny <mgorny@gentoo.org>
> wrote:
> > INSTALL_MASK=/usr/lib/systemd
> >
> > And live happy to the day you notice your system no longer boots.
> 
> This is a nice bandaid, and sure, it "solves" the immediate issue...
> but it doesn't actually solve the actual issue: when packages
> optionally install unwanted bloat, we make them an option via a USE
> flag. In this case, especially, since systemd isn't even the default
> (nor officially supported, whatever that amounts to), users certainly
> should not have to manually add an install mask to make portage do
> what it already should do.
> 
> Besides, as systemd gains momentum, we can probably expect that
> various pieces of software will have options to enable a systemd mode
> or a systemd build, or what have you, and then in this case, a global
> USE flag becomes even more imperative.

The flag is there already, and it is used whenever it involves
additional dependencies or in any other way makes the package
incompatible with non-systemd systems.

We aren't going to add USE flags which don't do anything. That topic
was discussed a thousand times, and rising it once more won't change
our decision.

Similarly, bash-completion flag will be gone at some point.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

On Wed, Aug 8, 2012 at 4:36 PM, Michał Górny <mgorny@gentoo.org> wrote:
> We aren't going to add USE flags which don't do anything. That topic
> was discussed a thousand times, and rising it once more won't change
> our decision.
>
> Similarly, bash-completion flag will be gone at some point.

Everyone has bash. Not everyone has systemd.

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 588 bytes --]

On Wed, 8 Aug 2012 16:38:07 +0200
"Jason A. Donenfeld" <Jason@zx2c4.com> wrote:

> On Wed, Aug 8, 2012 at 4:36 PM, Michał Górny <mgorny@gentoo.org>
> wrote:
> > We aren't going to add USE flags which don't do anything. That topic
> > was discussed a thousand times, and rising it once more won't change
> > our decision.
> >
> > Similarly, bash-completion flag will be gone at some point.
> 
> Everyone has bash. Not everyone has systemd.

Not everyone uses bash. Not everyone cares at all about
bash-completion. What is your point?

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

On Wed, Aug 8, 2012 at 4:45 PM, Michał Górny <mgorny@gentoo.org> wrote:
> Not everyone uses bash. Not everyone cares at all about
> bash-completion. What is your point?

I'm not saying I agree with the removal of bash-completion flag (that
discussion is for elsewhere), but just that your analogy doesn't hold.

zx2c4@Dell /usr/lib/systemd $ equery d bash|grep portage
sys-apps/portage-2.2.0_alpha120

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 682 bytes --]

On Wed, 8 Aug 2012 16:48:20 +0200
"Jason A. Donenfeld" <Jason@zx2c4.com> wrote:

> On Wed, Aug 8, 2012 at 4:45 PM, Michał Górny <mgorny@gentoo.org>
> wrote:
> > Not everyone uses bash. Not everyone cares at all about
> > bash-completion. What is your point?
> 
> I'm not saying I agree with the removal of bash-completion flag (that
> discussion is for elsewhere), but just that your analogy doesn't hold.
> 
> zx2c4@Dell /usr/lib/systemd $ equery d bash|grep portage
> sys-apps/portage-2.2.0_alpha120

As you may or may not know, portage does not use bash completion. Bash
completion is a strictly interactive-shell feature.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Michael Mol]

On Wed, Aug 8, 2012 at 11:14 AM, Michał Górny <mgorny@gentoo.org> wrote:
> On Wed, 8 Aug 2012 16:48:20 +0200
> "Jason A. Donenfeld" <Jason@zx2c4.com> wrote:
>
>> On Wed, Aug 8, 2012 at 4:45 PM, Michał Górny <mgorny@gentoo.org>
>> wrote:
>> > Not everyone uses bash. Not everyone cares at all about
>> > bash-completion. What is your point?
>>
>> I'm not saying I agree with the removal of bash-completion flag (that
>> discussion is for elsewhere), but just that your analogy doesn't hold.
>>
>> zx2c4@Dell /usr/lib/systemd $ equery d bash|grep portage
>> sys-apps/portage-2.2.0_alpha120
>
> As you may or may not know, portage does not use bash completion. Bash
> completion is a strictly interactive-shell feature.

Pretty sure he was responding to your statement 'not everybody uses
bash', not to your statement about bash completion.

-- 
:wq

Re: [gentoo-dev] Global Systemd USE Flag

[Rich Freeman]

On Wed, Aug 8, 2012 at 10:38 AM, Jason A. Donenfeld <Jason@zx2c4.com> wrote:
> On Wed, Aug 8, 2012 at 4:36 PM, Michał Górny <mgorny@gentoo.org> wrote:
>> We aren't going to add USE flags which don't do anything. That topic
>> was discussed a thousand times, and rising it once more won't change
>> our decision.
>>
>> Similarly, bash-completion flag will be gone at some point.
>
> Everyone has bash. Not everyone has systemd.
>

Then why does OpenRC go out of its way to avoid depending on bash?
The answer is that not everybody has bash.

In any case, this has been discussed on -dev before.  The reason that
the unit files were not made a use flag was that they're just simple
text files that don't take much space and don't do anything unless you
use systemd.  Having a USE flag to trigger their install doesn't have
much of a point, and it also means that to switch to systemd you'd
have to re-emerge anything that installs a unit file.

If we went this route we'd end up adding an openrc use flag for
anything that sticks files in /etc/init.d, bash-completion, and
probably a bunch of other stuff as well.

If a package is pulling in dependencies that is a different story, but
if we're just talking about a text file I think a USE flag is
overkill.

Rich

Re: [gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

On Wed, Aug 8, 2012 at 4:15 PM, Michał Górny <mgorny@gentoo.org> wrote:
> INSTALL_MASK=/usr/lib/systemd

As an unrelated side note, in case any one on the internet finds this
thread trying to "solve" this issue, it's worth pointing out that
since udev now installs that directory, the INSTALL_MASK should
actually be /usr/lib/systemd/system.

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 674 bytes --]

On Wed, 8 Aug 2012 16:22:47 +0200
"Jason A. Donenfeld" <Jason@zx2c4.com> wrote:

> On Wed, Aug 8, 2012 at 4:15 PM, Michał Górny <mgorny@gentoo.org>
> wrote:
> > INSTALL_MASK=/usr/lib/systemd
> 
> As an unrelated side note, in case any one on the internet finds this
> thread trying to "solve" this issue, it's worth pointing out that
> since udev now installs that directory, the INSTALL_MASK should
> actually be /usr/lib/systemd/system.

You are right. In case users really intend to use that, they may be
better using app-portage/install-mask, and:

$ install-mask -a systemd

which will add just the right path.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

On Wed, Aug 8, 2012 at 4:33 PM, Michał Górny <mgorny@gentoo.org> wrote:
> You are right. In case users really intend to use that, they may be
> better using app-portage/install-mask, and:
>
> $ install-mask -a systemd
>
> which will add just the right path.

Still misses the point. USE flags were invented to deal with these
options. On a default install, which uses OpenRC, users shouldn't have
to then emerge an additional program to add more configuration in
order to have a clean system.

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 677 bytes --]

On Wed, 8 Aug 2012 16:35:22 +0200
"Jason A. Donenfeld" <Jason@zx2c4.com> wrote:

> On Wed, Aug 8, 2012 at 4:33 PM, Michał Górny <mgorny@gentoo.org>
> wrote:
> > You are right. In case users really intend to use that, they may be
> > better using app-portage/install-mask, and:
> >
> > $ install-mask -a systemd
> >
> > which will add just the right path.
> 
> Still misses the point. USE flags were invented to deal with these
> options. On a default install, which uses OpenRC, users shouldn't have
> to then emerge an additional program to add more configuration in
> order to have a clean system.

No, they weren't.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Walter Dnes]

On Wed, Aug 08, 2012 at 04:43:33PM +0200, Micha?? G??rny wrote
> On Wed, 8 Aug 2012 16:35:22 +0200
> "Jason A. Donenfeld" <Jason@zx2c4.com> wrote:
> > 
> > Still misses the point. USE flags were invented to deal with these
> > options. On a default install, which uses OpenRC, users shouldn't have
> > to then emerge an additional program to add more configuration in
> > order to have a clean system.
> 
> No, they weren't.

  This looks very similar in principle to USE="-docs".

-- 
Walter Dnes <waltdnes@waltdnes.org>

Re: [gentoo-dev] Global Systemd USE Flag

[Mart Raudsepp]

On N, 1970-01-01 at 00:00 +0000, Jason A. Donenfeld wrote:
> On Wed, Aug 8, 2012 at 4:33 PM, Michał Górny <mgorny@gentoo.org> wrote:
> > You are right. In case users really intend to use that, they may be
> > better using app-portage/install-mask, and:
> >
> > $ install-mask -a systemd
> >
> > which will add just the right path.
> 
> Still misses the point. USE flags were invented to deal with these
> options. On a default install, which uses OpenRC, users shouldn't have
> to then emerge an additional program to add more configuration in
> order to have a clean system.

USE flags are not meant for controlling every little thing, such as
conditional installing a 400 byte file that does no harm when not used,
other than taking 1 block of filesystem space (4kB or so), which can be
workarounded by INSTALL_MASK if you are building an embedded system. I
seriously doubt they were invented for such a purpose, but rather to
control ./configure arguments and external dependencies.

No, wanting to get rid of those on a desktop/server via a USE flag (as
opposed to an INSTALL_MASK) is not a consideration, as that's users
completely unneeded desire for no technical reason. If taking 500kB
total for systemd service files is an issue, then the issue really is
that you are using a 1GB /usr partition or something.

This all is similar to how we in GNOME unconditionally install user and
developer documentation, as long as it does not impose any extra build
time or downloads.
(no, this is not really negotiable for change, and we are talking about
more than 400 byte files here; we'd be happy however if portage binary
packages supported splitting of the source packages files to separate
packages, so that binary distribution derivatives could work in a
similar model as purely binary distributions)

USE flags typically control the functionality of compiled binaries,
usually involving external dependencies to achieve such extra
functionality.

http://www.gentoo.org/doc/en/handbook/handbook-x86.xml?part=2&chap=2#doc_chap1_sect2


Best Regards,
Mart Raudsepp

Re: [gentoo-dev] Global Systemd USE Flag

[Patrick Lauer]

On 08/08/12 22:35, Jason A. Donenfeld wrote:
> On Wed, Aug 8, 2012 at 4:33 PM, Michał Górny <mgorny@gentoo.org> wrote:
>> You are right. In case users really intend to use that, they may be
>> better using app-portage/install-mask, and:
>>
>> $ install-mask -a systemd
>>
>> which will add just the right path.
> Still misses the point. USE flags were invented to deal with these
> options. On a default install, which uses OpenRC, users shouldn't have
> to then emerge an additional program to add more configuration in
> order to have a clean system.
>
And while we're at it -

can we *please* use the openrc useflag to have correct paths and binary
names again?
Just because upstream says we should be fedora doesn't mean we have to
do it.

Right now it's really frustrating to have systemd artifacts all over my
system even when I explicitly ask for it to not be near it.
Very rude. Very not Gentoo.

Re: [gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

On Wed, Aug 8, 2012 at 4:48 PM, Patrick Lauer <patrick@gentoo.org> wrote:
> can we *please* use the openrc useflag to have correct paths and binary
> names again?
> Just because upstream says we should be fedora doesn't mean we have to
> do it.
>
> Right now it's really frustrating to have systemd artifacts all over my
> system even when I explicitly ask for it to not be near it.
> Very rude. Very not Gentoo.

+100

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 1180 bytes --]

On Wed, 08 Aug 2012 22:48:58 +0800
Patrick Lauer <patrick@gentoo.org> wrote:

> On 08/08/12 22:35, Jason A. Donenfeld wrote:
> > On Wed, Aug 8, 2012 at 4:33 PM, Michał Górny <mgorny@gentoo.org>
> > wrote:
> >> You are right. In case users really intend to use that, they may be
> >> better using app-portage/install-mask, and:
> >>
> >> $ install-mask -a systemd
> >>
> >> which will add just the right path.
> > Still misses the point. USE flags were invented to deal with these
> > options. On a default install, which uses OpenRC, users shouldn't
> > have to then emerge an additional program to add more configuration
> > in order to have a clean system.
> >
> And while we're at it -
> 
> can we *please* use the openrc useflag to have correct paths and
> binary names again?
> Just because upstream says we should be fedora doesn't mean we have to
> do it.
> 
> Right now it's really frustrating to have systemd artifacts all over
> my system even when I explicitly ask for it to not be near it.
> Very rude. Very not Gentoo.

Yes, and please remove all the occurrences of 'GNU' because I don't
like it.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Luca Barbato]

On 08/08/2012 04:53 PM, Michał Górny wrote:
> Yes, and please remove all the occurrences of 'GNU' because I don't
> like it.

We have people working on a clang/freebsd gentoo, you might help them
and use that. It sort of works fine.

For a project Flameeyes replaced most of system using smaller
alternatives to most of the gnu runtime.

I'm helping getting musl as a first class libc in Gentoo, if uclibc
feels too GNU-ish for you.

As strange as it might feel for you we have people working on providing
alternatives that might be useful for specific purposes.

lu

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 884 bytes --]

On Thu, 09 Aug 2012 10:48:38 +0200
Luca Barbato <lu_zero@gentoo.org> wrote:

> On 08/08/2012 04:53 PM, Michał Górny wrote:
> > Yes, and please remove all the occurrences of 'GNU' because I don't
> > like it.
> 
> We have people working on a clang/freebsd gentoo, you might help them
> and use that. It sort of works fine.
> 
> For a project Flameeyes replaced most of system using smaller
> alternatives to most of the gnu runtime.
> 
> I'm helping getting musl as a first class libc in Gentoo, if uclibc
> feels too GNU-ish for you.
> 
> As strange as it might feel for you we have people working on
> providing alternatives that might be useful for specific purposes.

No. I meant to have 'GNU' tools with 'GNU' stripped. Isn't that what
the whole discussion is about? Changing names of tools just for
someone's liking?

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Luca Barbato]

On 08/09/2012 10:57 AM, Michał Górny wrote:
> No. I meant to have 'GNU' tools with 'GNU' stripped. Isn't that what
> the whole discussion is about? Changing names of tools just for
> someone's liking?

No, we are discussing about an upstream merging two unrelated projects
assuring users that nothing would change for them.

In a week they claimed that it was unsupported, then backpedaled, then
they changed its paths.

Forking udev hadn't been considered mostly just on that premise.

lu

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 796 bytes --]

On Thu, 09 Aug 2012 11:20:52 +0200
Luca Barbato <lu_zero@gentoo.org> wrote:

> On 08/09/2012 10:57 AM, Michał Górny wrote:
> > No. I meant to have 'GNU' tools with 'GNU' stripped. Isn't that what
> > the whole discussion is about? Changing names of tools just for
> > someone's liking?
> 
> No, we are discussing about an upstream merging two unrelated projects
> assuring users that nothing would change for them.
> 
> In a week they claimed that it was unsupported, then backpedaled, then
> they changed its paths.
> 
> Forking udev hadn't been considered mostly just on that premise.

So someone should just *finally* fork it, rather than talking about it
all the time.

And I believe that renaming executables won't undo the merge.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Rich Freeman]

On Thu, Aug 9, 2012 at 6:01 AM, Michał Górny <mgorny@gentoo.org> wrote:
> On Thu, 09 Aug 2012 11:20:52 +0200
> Luca Barbato <lu_zero@gentoo.org> wrote:
>> Forking udev hadn't been considered mostly just on that premise.
>
> So someone should just *finally* fork it, rather than talking about it
> all the time.
>

++

If the sky actually starts falling there are things we can do about
it.  Until then, running around yelling about it isn't terribly
productive.  There is this great fear about what udev MIGHT do.  Well,
if they do it then lots of of people will fork it if systemd isn't
universally embraced as the ultimate init replacement.

However, right now everybody is worried about a future that may or may
not happen, or which may or may not be welcomed with open arms by the
time it does.

In the meantime I imagine most Gentoo packages will at least ship
openrc init scripts, if they bother to ship init scripts at all.
There isn't any requirement that packages have them, afaik, so if
somebody ships a systemd unit but not an openrc script let's be
careful about pulling out the pitchforks.  :)

Rich

Re: [gentoo-dev] Global Systemd USE Flag

[Luca Barbato]

On 08/09/2012 12:01 PM, Michał Górny wrote:
> On Thu, 09 Aug 2012 11:20:52 +0200
> Luca Barbato <lu_zero@gentoo.org> wrote:
> 
>> On 08/09/2012 10:57 AM, Michał Górny wrote:
>>> No. I meant to have 'GNU' tools with 'GNU' stripped. Isn't that what
>>> the whole discussion is about? Changing names of tools just for
>>> someone's liking?
>>
>> No, we are discussing about an upstream merging two unrelated projects
>> assuring users that nothing would change for them.
>>
>> In a week they claimed that it was unsupported, then backpedaled, then
>> they changed its paths.
>>
>> Forking udev hadn't been considered mostly just on that premise.
> 
> So someone should just *finally* fork it, rather than talking about it
> all the time.

I had that[1] since ages, mostly because I was curious about how complex
udev is internally.

If enough people want to play this game welcome. I expect the same
people stomping on mdev complaining about the next little experiment.

https://github.com/lu-zero/udev

lu

Re: [gentoo-dev] Global Systemd USE Flag

[Samuli Suominen]

On 08/09/2012 12:20 PM, Luca Barbato wrote:
> On 08/09/2012 10:57 AM, Michał Górny wrote:
>> No. I meant to have 'GNU' tools with 'GNU' stripped. Isn't that what
>> the whole discussion is about? Changing names of tools just for
>> someone's liking?
>
> No, we are discussing about an upstream merging two unrelated projects
> assuring users that nothing would change for them.
>
> In a week they claimed that it was unsupported, then backpedaled, then
> they changed its paths.

(picked near random mail from the thread)

sep. /usr was not supported prior to renaming and $udevdir has been 
dynamic from the udev.pc pkg-config file for long as I can remember
therefore the renaming and path change is merely cosmetics to which 
*users* don't need to pay attention to

so yeah, this whole thread is just that, trying to introduce regression 
due to personal preference

187-r3 does it right and after fixing the few hardcoded paths in tree, 
we can drop the 2 patches and backwards compability
i'm nearly done with pushing fixes to drop the few known hardcoded paths 
we have in tree, and about to ask for a tinderbox run to catch the rest 
(which can take it's time, the patches are there, as mentioned)

so help is welcome with the migration[1], see eg. usb_modeswitch, 
libmtp, udisks:0, udisks:2 (configure.ac), etc. for example

[1] this is about using non-hardcoded paths and respecting udev.pc set 
udevdir=, not really about migrating, so even if the move wasn't 
happening, these have always been "bugs" of sort

- Samuli

Re: [gentoo-dev] Global Systemd USE Flag

[Rich Freeman]

On Wed, Aug 8, 2012 at 10:48 AM, Patrick Lauer <patrick@gentoo.org> wrote:
> can we *please* use the openrc useflag to have correct paths and binary
> names again?
> Just because upstream says we should be fedora doesn't mean we have to
> do it.

I think that having binaries going in different places based on a USE
flag is going to lead to a big mess - especially if we're talking
about system packages.

If we want to argue about where we put something by all means hash it
out or escalate to council.  If we want to debate whether to install
compatibility symlinks I think that is also more reasonable.

However, I don't want the path to bash or glibc or whatever to depend
on whether a particular package maintainer believes in the /usr move
or even moreso whether some USE flag is set.

Rich

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 1132 bytes --]

On Wed, 8 Aug 2012 11:03:25 -0400
Rich Freeman <rich0@gentoo.org> wrote:

> On Wed, Aug 8, 2012 at 10:48 AM, Patrick Lauer <patrick@gentoo.org>
> wrote:
> > can we *please* use the openrc useflag to have correct paths and
> > binary names again?
> > Just because upstream says we should be fedora doesn't mean we have
> > to do it.
> 
> I think that having binaries going in different places based on a USE
> flag is going to lead to a big mess - especially if we're talking
> about system packages.
> 
> If we want to argue about where we put something by all means hash it
> out or escalate to council.  If we want to debate whether to install
> compatibility symlinks I think that is also more reasonable.
> 
> However, I don't want the path to bash or glibc or whatever to depend
> on whether a particular package maintainer believes in the /usr move
> or even moreso whether some USE flag is set.

Path to bash can't change because it will break most of scripts
in the world.

Path to libc can't change because it will break all of the executables
in the world.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Rich Freeman]

On Wed, Aug 8, 2012 at 11:13 AM, Michał Górny <mgorny@gentoo.org> wrote:
> Path to bash can't change because it will break most of scripts
> in the world.
>
> Path to libc can't change because it will break all of the executables
> in the world.

My point was illustrative.  Basically if we're going to move
something, then MOVE it, and don't make that conditional on some flag
(unless you're talking about something more nuanced like library
location on a multilib system, which is also to be carefully planned).

You can say that something belongs in A, and I might say it belongs in
B, but I think we should all agree that it should only be in one place
or the other, perhaps with compatibility symlinks where necessary (and
the symlink behavior might be more conditional whether via USE or
eselect or whatever).

I don't want a USE=traditional/openrc/gentoo/my-preference flag that
randomly breaks things when it is or isn't present - depending on
whether various package maintainers worship at one altar or another.

Rich

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 1311 bytes --]

On Wed, 8 Aug 2012 17:13:26 +0200
Michał Górny <mgorny@gentoo.org> wrote:

> On Wed, 8 Aug 2012 11:03:25 -0400
> Rich Freeman <rich0@gentoo.org> wrote:
> 
> > On Wed, Aug 8, 2012 at 10:48 AM, Patrick Lauer <patrick@gentoo.org>
> > wrote:
> > > can we *please* use the openrc useflag to have correct paths and
> > > binary names again?
> > > Just because upstream says we should be fedora doesn't mean we
> > > have to do it.
> > 
> > I think that having binaries going in different places based on a
> > USE flag is going to lead to a big mess - especially if we're
> > talking about system packages.
> > 
> > If we want to argue about where we put something by all means hash
> > it out or escalate to council.  If we want to debate whether to
> > install compatibility symlinks I think that is also more reasonable.
> > 
> > However, I don't want the path to bash or glibc or whatever to
> > depend on whether a particular package maintainer believes in
> > the /usr move or even moreso whether some USE flag is set.
> 
> Path to bash can't change because it will break most of scripts
> in the world.
> 
> Path to libc can't change because it will break all of the executables
> in the world.

Ah, sorry. Let me clarify: path to ld.so.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Patrick Lauer]

On 08/08/12 22:15, Michał Górny wrote:
> On Wed, 8 Aug 2012 15:11:42 +0200
> "Jason A. Donenfeld" <Jason@zx2c4.com> wrote:
[snip]
>
> Yowza! All the packages that provide systemd unit files are installing
> them?! But I don't even use systemd. I don't want this cruft on my
> system.
>
> Proposal: global USE flag for systemd, just like there's one for
> openrc.
> INSTALL_MASK=/usr/lib/systemd
>
> And live happy to the day you notice your system no longer boots.
>
That doesn't work anymore - "improvement" in udev-186:

equery f udev | grep udevd

/usr/lib/systemd/systemd-udevd


And as long as our maintainers refuse to use the proper paths this is
just one of the little things that makes life more exciting for us.

Can we please add some sanity back?

Re: [gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

On Wed, Aug 8, 2012 at 4:31 PM, Patrick Lauer <patrick@gentoo.org> wrote:
> And as long as our maintainers refuse to use the proper paths this is
> just one of the little things that makes life more exciting for us.
>
> Can we please add some sanity back?

Exactly. Right now, with no USE flag, and no differentiation,
maintainers are kind of just stepping on each others toes. There's
simply no protocol for dealing with the increasingly aggressive
upstream systemdification. With a global USE flags, maintainers can
then think about things in terms of "okay, the systemd world likes it
this way; the rest of the world likes it this way... therefore: use
systemd && kitten_killer".

Re: [gentoo-dev] Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 841 bytes --]

On Wed, 08 Aug 2012 22:31:40 +0800
Patrick Lauer <patrick@gentoo.org> wrote:

> On 08/08/12 22:15, Michał Górny wrote:
> > On Wed, 8 Aug 2012 15:11:42 +0200
> > "Jason A. Donenfeld" <Jason@zx2c4.com> wrote:
> [snip]
> >
> > Yowza! All the packages that provide systemd unit files are
> > installing them?! But I don't even use systemd. I don't want this
> > cruft on my system.
> >
> > Proposal: global USE flag for systemd, just like there's one for
> > openrc.
> > INSTALL_MASK=/usr/lib/systemd
> >
> > And live happy to the day you notice your system no longer boots.
> >
> That doesn't work anymore - "improvement" in udev-186:
> 
> equery f udev | grep udevd
> 
> /usr/lib/systemd/systemd-udevd

Yes, sorry, I was lazy and didn't add '/system' there. Forgot about
udev.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Richard Yao]

[-- Attachment #1: Type: text/plain, Size: 907 bytes --]

On 08/08/2012 10:31 AM, Patrick Lauer wrote:
> On 08/08/12 22:15, Michał Górny wrote:
>> On Wed, 8 Aug 2012 15:11:42 +0200
>> "Jason A. Donenfeld" <Jason@zx2c4.com> wrote:
> [snip]
>>
>> Yowza! All the packages that provide systemd unit files are installing
>> them?! But I don't even use systemd. I don't want this cruft on my
>> system.
>>
>> Proposal: global USE flag for systemd, just like there's one for
>> openrc.
>> INSTALL_MASK=/usr/lib/systemd
>>
>> And live happy to the day you notice your system no longer boots.
>>
> That doesn't work anymore - "improvement" in udev-186:
> 
> equery f udev | grep udevd
> 
> /usr/lib/systemd/systemd-udevd
> 
> 
> And as long as our maintainers refuse to use the proper paths this is
> just one of the little things that makes life more exciting for us.
> 
> Can we please add some sanity back?
> 

I second this suggestion.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 900 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[William Hubbs]

[-- Attachment #1: Type: text/plain, Size: 1729 bytes --]

On Wed, Aug 08, 2012 at 10:37:42AM -0400, Richard Yao wrote:
> > That doesn't work anymore - "improvement" in udev-186:
> > 
> > equery f udev | grep udevd
> > 
> > /usr/lib/systemd/systemd-udevd
> > 
> > 
> > And as long as our maintainers refuse to use the proper paths this is
> > just one of the little things that makes life more exciting for us.
> > 
> > Can we please add some sanity back?
> > 
> 
> I second this suggestion.
> 

Folks, I am going to point out a couple of things.

First, using /usr/lib/dirname/* for binaries does not break any linux or
unix standards. There are many packages that do this. Some use libexec,
but this is being changed to lib as I understand it.

Second, upstream renaming a binary doesn't constitute breaking any
standards. There is no rule or law that says, for example, that upstream
udev must call their daemon udevd. What if they decide to change it to
device-manager-daemon-for-linux? They can do exactly this if they want,
and it is up to us, the packagers, to make sure that things don't break
for our distributions.

Third, putting daemons outside the path doesn't break any standards. Udev
isn't the only package doing this. I believe, postfix, for one, doesn't
install its daemons in a directory on the path, but I don't see anyone
complaining about this.

I don't see anything wrong with moving a deamon out of the path, because
afaik in day-to-day operations, you don't run a daemon directly from the
command line. it is started or stopped by your init system.

So, I ask again. You keep complaining about "insanity". What's the
insanity and why should we go to all of the extra effort you want us to
go to to avoid it?

William


[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

On Thu, Aug 9, 2012 at 12:19 AM, William Hubbs <williamh@gentoo.org> wrote:
> So, I ask again. You keep complaining about "insanity". What's the
> insanity and why should we go to all of the extra effort you want us to
> go to to avoid it?

I think it's more of a knee-jerk reaction to this: Redhat is pushing
systemd very hard, and its technical merits have gained a lot of sway
in many places. As such, it seems like lots of everything are joining
a fast-paced systemd stampede. Seeing that udevd has been renamed to
something with "systemd" in the name raises a gut fear "oh no, now
something as fundamental as udev is becoming systemdified. It's just a
matter of time before init.d disappears forever." To the
non-systemd'ers, I think the general perception is that without any
set of policies to manage the stampede, systemd will eventually take
over.

Maybe this fear is warranted. Maybe it's silly. I don't know. I am
glad, though, that Gentoo is sticking with OpenRC, and I hope that the
consequences of that decision are respected by ebuild maintainers.

[gentoo-dev] Re: Global Systemd USE Flag

[Duncan]

Jason A. Donenfeld posted on Thu, 09 Aug 2012 06:33:02 +0200 as excerpted:

> Redhat is pushing systemd very hard [and] it seems like lots of
> everything are joining a fast-paced systemd stampede. [I] think the
> general perception is that without any set of policies to manage the
> stampede, systemd will eventually take over.
> 
> Maybe this fear is warranted. Maybe it's silly. I don't know. I am glad,
> though, that Gentoo is sticking with OpenRC, and I hope that the
> consequences of that decision are respected by ebuild maintainers.

... For now.  Gentoo's sticking with openrc by default, for now.

(TL;DR: While prediction is risky, I believe openrc's a good bet out say 
two years, but that people will be switching to systemd by then, and by 
five years out, systemd will be assumed.  But given that some gentoo devs 
run larger gentoo installations and won't want to switch to systemd on 
servers, I expect that openrc will still be supported five years out, but 
a decade out's just too much unforeseen change away to predict.)

Consider that say five years is a long time in Linux, which evolves "at 
internet speed."  While I personally plan on continuing to use openrc for 
the time being, I also expect that in five years, I and most everyone 
else on "desktop Linux" will have been "assimilated" into the systemd 
borg.

Consider... five years ago was 2007. Android hadn't been released yet at 
this point in 2007 (November, according to the LWN 2007 timeline I'm 
looking at [1]).  Nokia releases the N800.  KDE4 was yet to come out 
(January, 2008) altho the PR machine had been in full swing for awhile.  
Gnome3 was still a ways out.  GPLv3 is released and Bruce Perens among 
others predicts Linus and Linux will switch after kicking and screaming 
for a couple years.  Kernel 2.6.20-2.6.23.  Con Colivas quits the kernel, 
but eventually releases the brainfuck scheduler without attempting a 
mainline merge.  The MS/Novell agreement and controversy.  AMD announces 
that it is opening up ATI graphics documentation.  SCO files for chapter 
11 (which just now it's trying to convert to chapter 7) bankruptcy.  One-
laptop-per-child XO goes into mass production.  SAMBA gets access to the 
MS protocol docs...

For gentoo in 2007, Daniel Robbins returned for a very short period, got 
in a fight on this list and left in a bit of a huff after his demands to 
list-ban the other party weren't just magically followed as they had been 
years before.

Especially the Android thing is interesting.  Who would have predicted 
that it would have the market share it does today, and that Nokia and 
Blackberry would be where they are?


The point being, five years is a LONG time in Linux/FLOSS.  Five years 
from now, many of the gentoo's current developers will have moved on, 
hopefully to have been replaced.  As such, even if no gentoo devs change 
their minds, it's quite likely the new blood will be changing gentoo's 
direction.

Five years from now, I expect xorg will be fading and the big desktops 
(will xfce replace gnome due tot he gnome3 fiasco?  mate?  cinnamon? kde5/
kde-frameworks?) will be focusing on wayland.

And five years from now I expect the big desktops will require systemd 
and whatever it has engulfed by then, for many of their big features.  
Hopefully there will be USE flags to toggle that, much as there are USE 
flags toggling udev assuming features today, but that remains to be 
seen.  Of course, five years ago hal was still big too, so five years 
from now systemd might be fading as well, for all I know.

But it's worth noting that at least for me, and I'd assume at least some 
others, the problem with systemd isn't that people are entirely opposed 
to it, but rather, the speed with which it is happening, and the 
perceived immaturity and continued heavy development of the systemd 
solution as it exists today.  In another two years, I really expect it to 
be showing signs of stabilization and maturity.  In 2-3 years, I expect 
it to be a rather more reasonable and stable solution than it is today, 
something that most people won't have such a problem with.  And by five 
years out, I think that most will have already switched.


I do hope/expect that five years from now openrc will at least still be 
supported as an alternate, for use with legacy xorg and for no-GUI server 
installations, etc, but I really expect that systemd will be the assumed 
default by then, much as udev is the assumed default today, even if 
static dev or the simple kernel devfs are still supported, but as 
exception-case.

And even if it's not mainline gentoo supported, there's the kde-sunset 
overlay precedent, with user support but cooperation from mainline gentoo/
kde to try to keep conflicts to a minimum.  I expect openrc will at least 
get /that/ level of support.  And actually, given that a number of gentoo 
devs support larger installations of gentoo and aren't likely to be 
wanting to switch servers, etc, to systemd just because it's there, I 
expect there will still be active gentoo developer support for openrc, 
the key missing resource in the kde3 case, so openrc will continue with 
mainline gentoo support likely out seven years or so.  Beyond that, 
pretty much /everything/ is in too much flux to try to predict, so I 
won't even attempt to /guess/ whether openrc, or for that matter, pretty 
much /any/ important today gentoo component, will be around at least "as 
we know it" in a decade.


So I really expect people to be switching to systemd 2-3 years from now, 
and that it'll be the gentoo default in 3-5 years, tho openrc will almost 
certainly be supported in /some/ form, at least comparable to the kde-
sunset overlay and probably officially, at least five years out.  But a 
decade out, all bets are off!

----
[1] LWN 2007 timeline, one big page version:

http://lwn.net/Articles/307167/

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

Re: [gentoo-dev] Re: Global Systemd USE Flag

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 569 bytes --]

On Thu, 9 Aug 2012 06:57:17 +0000 (UTC)
Duncan <1i5t5.duncan@cox.net> wrote:

> So I really expect people to be switching to systemd 2-3 years from
> now, and that it'll be the gentoo default in 3-5 years, tho openrc
> will almost certainly be supported in /some/ form, at least
> comparable to the kde- sunset overlay and probably officially, at
> least five years out.  But a decade out, all bets are off!

Do you really believe that 2-3 years from now systemd will still
exist? Not systemkit, then syskit, then ...


-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

[gentoo-dev] Re: Global Systemd USE Flag

[Duncan]

Michał Górny posted on Thu, 09 Aug 2012 09:24:26 +0200 as excerpted:

> On Thu, 9 Aug 2012 06:57:17 +0000 (UTC)
> Duncan <1i5t5.duncan@cox.net> wrote:
> 
>> So I really expect people to be switching to systemd 2-3 years from
>> now, and that it'll be the gentoo default in 3-5 years, tho openrc will
>> almost certainly be supported in /some/ form, at least comparable to
>> the kde- sunset overlay and probably officially, at least five years
>> out.  But a decade out, all bets are off!
> 
> Do you really believe that 2-3 years from now systemd will still exist?
> Not systemkit, then syskit, then ...

Were any of the *kits, etc, Lennart projects?  IDR reading about pulse, 
anyway, switching names.

Meanwhile, Isn't systemd close to two years old now, already?  I'm about 
a month behind on LWN, but I've kept up with my other community news and 
haven't read anything of a change yet, and it'd take /some/ discussion 
and then time to switch to some other name/framework, etc.  Given I've 
not read of any such thing yet, and I suspect Lennart would attempt a 
veto, in practice I think we're locked in to nearing two years anyway, at 
least a year.

Plus, they're still in the build-up phase with systemd ATM, integrating 
things like logging (binary-format, no less! <shudder>!), etc.  I'd 
expect them to play with their toys a bit before getting bored and 
throwing them out of the tram. =:^)

But in the 3-5 year timeframe I think it's possible.  Thus the mention of 
hal.  And certainly out beyond five years... FLOSS predictions of any 
sort tend to get fuzzy out past five years; given what seems to be 3-5 
year worldview changer event period, but a built-in system reaction time 
buffer that never-the-less gives you /some/ prediction safety out to five 
years, /maybe/ seven for the real broad strokes.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

Re: [gentoo-dev] Re: Global Systemd USE Flag

[Peter Stuge]

Duncan wrote:
> given that a number of gentoo devs support larger installations of
> gentoo and aren't likely to be wanting to switch servers, etc, to
> systemd just because it's there

I think that once they've learned systemd, they will want to switch
those servers fast.

I use it on one sort-of production server for about a year. Had I set
up any new servers since trying it out they would also have used it.

As I wrote before, I've not used init scripts on any system for over
10 years, including the maybe 50 servers that I've ran during that
time. Although systemctl is a stupid long command the benefits of
systemd make it an easy choice for me.

systemd isn't at all unstable in my experience, the only thing that
it is lacking is experience among administrators.

I agree that as long as there is anyone interested in maintaining and
supporting openrc, it will be available in Gentoo. The alternatives
and choice is what makes Gentoo great, and I think this fundamental
principle will never change, no matter what Red Hat or anyone else
wants to do.

systemd in itself is no big problem, Gentoo already solves much
bigger problems. Yes, we may have to rethink the service file policy,
perhaps integrate the init system with eselect, or have some other
tooling - who knows, so far I don't know if Gentoo has ever supported
replacing openrc. I think it's a good idea to not require re-emerging
every package just to get them, but there are other solutions.

Don't worry, it will all be fine.

As long as you and I run no systemd maybe it's OK if we still
download it's tarball in order to build and install the udev files.
If you feel very strongly not OK, after much experience with systemd
and udev, then by all means do fork udev. But I *seriously* don't see
the point. Please choose another project. :)


Kind regards

//Peter

[gentoo-dev] Re: Global Systemd USE Flag

[Duncan]

Peter Stuge posted on Thu, 09 Aug 2012 14:29:42 +0200 as excerpted:

> systemd isn't at all unstable in my experience, the only thing that it
> is lacking is experience among administrators.

FWIW, the "unstable" I was referring to wasn't necessarily crashing or 
refusing to do its init job, but "design unstable" -- continuing to add 
and change functionality such as logging, where the format is 
specifically declared to be unstable and subject to change.

When all that stuff gets worked out and functionality begins to settle 
down into predictable, and when an admin can read a systemd log with his 
choice of tools from a well documented and well supported by multiple 
tools format...

THEN I believe is when gentoo will likely switch to systemd by default, 
along with most users, altho based on upstream and leading distro 
predictions, various bits of nice desktop GUI functionality will likely 
be broken without systemd somewhat before then, so those gentoo desktop 
users who like their automatic GUI bells and whistles such as automount, 
etc, may find themselves switching somewhat before that...

Which is precisely where crashing-instability may begin to show its ugly 
head as well, with various desktop users unfortunately not so well 
prepared to fix things... yes, even on gentoo.  Hopefully not, but...

> I agree that as long as there is anyone interested in maintaining and
> supporting openrc, it will be available in Gentoo. The alternatives and
> choice is what makes Gentoo great, and I think this fundamental
> principle will never change, no matter what Red Hat or anyone else wants
> to do.

Hear, hear! =:^)

> Don't worry, it will all be fine.

While I've little doubt most will ultimately navigate the hazards (tho 
the process may take years) unfortunately, that sounds WAY too much like 
what we heard in both the kde4 and gnome3 cases.  IOW, at least to folks 
who have gone thru that, your "reassurances" will likely make many even 
more sure than they were before, that systemd is an iceberg with users 
looking like the Titanic!  IOW, stay FAR FAR away from it!

I know that wasn't your intent, but given recent events on the desktop, 
and the assurances WAY too much like the above that came before them...

> As long as you and I run no systemd maybe it's OK if we still download
> it's tarball in order to build and install the udev files.

I don't see a problem with that.  After all, many packages, especially 
those from kde's monolithic tarballs, already work that way.

But if we have to build systemd as part of the udev build, only to delete 
it before final qmerge... which appears from what I've read to be the 
case despite promises that it'll work just as before...  (But that angle 
has been discussed already and other than fork, it seems there's little 
choice unless upstream changes its mind.  And so far, it hasn't seemed 
enough to actually do that fork, tho the mdev alternative project 
certainly seems to be getting some traction because someone did actually 
step up and take responsibility for making it happen  But that too has 
been discussed...)

> If you feel
> very strongly not OK, after much experience with systemd and udev, then
> by all means do fork udev. But I *seriously* don't see the point. Please
> choose another project. :)

As I said, I think gentoo will eventually default to systemd.  But I do 
hope openrc stays the default until systemd dev-stabilizes quite a bit 
from its current state.  Fortunately, it does seem openrc will remain the 
gentoo assumed default for at least the near future, anyway, and 
hopefully by the time that changes, systemd really has reasonably matured 
and stabilized.  Prospects look reasonable so far! =:^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

Re: [gentoo-dev] Re: Global Systemd USE Flag

[Ian Stakenvicius]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 09/08/12 02:57 AM, Duncan wrote:
> Jason A. Donenfeld posted on Thu, 09 Aug 2012 06:33:02 +0200 as
> excerpted:
> 
> Consider... five years ago was 2007. Android hadn't been released
> yet at this point in 2007 (November, according to the LWN 2007
> timeline I'm looking at [1]).  Nokia releases the N800.  KDE4 was
> yet to come out (January, 2008) altho the PR machine had been in
> full swing for awhile. Gnome3 was still a ways out.  GPLv3 is
> released and Bruce Perens among others predicts Linus and Linux
> will switch after kicking and screaming for a couple years.  Kernel
> 2.6.20-2.6.23.  Con Colivas quits the kernel, but eventually
> releases the brainfuck scheduler without attempting a mainline
> merge.  The MS/Novell agreement and controversy.  AMD announces 
> that it is opening up ATI graphics documentation.  SCO files for
> chapter 11 (which just now it's trying to convert to chapter 7)
> bankruptcy.  One- laptop-per-child XO goes into mass production.
> SAMBA gets access to the MS protocol docs...
> 

Yep, those were the days; it's been all down hill from there.. ;P
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)

iF4EAREIAAYFAlAjzDgACgkQ2ugaI38ACPAGDAD9G1aa7QecEzfBKG3nHsRj17gq
yOVJjTVcSCiNGneJM4kA/ifMmQBjvQuq/HUufEiOwfxZzW2qqqYRxD71g7UwFgMi
=0ecq
-----END PGP SIGNATURE-----

Re: [gentoo-dev] Global Systemd USE Flag

[Jason A. Donenfeld]

On Wed, Aug 8, 2012 at 4:31 PM, Patrick Lauer <patrick@gentoo.org> wrote:
>
> equery f udev | grep udevd
>
> /usr/lib/systemd/systemd-udevd
>
>
> And as long as our maintainers refuse to use the proper paths this is
> just one of the little things that makes life more exciting for us.
>
> Can we please add some sanity back?
>

The gods heard your call, and have replied:
> Yes, udev on non-systemd systems is in our eyes a dead end, in case you
> haven't noticed it yet. I am looking forward to the day when we can drop
> that support entirely.
-- Lennart  [1]

[1] http://lists.freedesktop.org/archives/systemd-devel/2012-August/006066.html

Re: [gentoo-dev] Global Systemd USE Flag

[Luca Barbato]

On 8/12/12 6:10 PM, Jason A. Donenfeld wrote:
> The gods heard your call, and have replied:
>> Yes, udev on non-systemd systems is in our eyes a dead end, in case you
>> haven't noticed it yet. I am looking forward to the day when we can drop
>> that support entirely.
> -- Lennart  [1]
>
> [1] http://lists.freedesktop.org/archives/systemd-devel/2012-August/006066.html

In the face of the statement given when udev tree got merged into 
systemd tree.

lu