Page 1 of 11
What should we do about systemd?
Posted: 2014-10-29 08:34
by edbarx
It is clear there is more than one option regarding systemd. If I understand these are:
a) give up Debian to use another distribution which respects the *nix tradition
b) concentrate on systemd's fork (uselessd) to port it to Debian
c) use sysvinit (INIT) irrespective of its limitations with respect to modern software requirements
d) use another initialisation system like runit
e) accept systemd and continue using Debian
I think, the more logical and practical choice is option b.
Tomazzi has coded a complex package of debugging code that can be used to make uselessd more reliable as it adds debugging features to quickly fix bugs. In Poettering's code there is already some code like Tomazzi's, but I don't know whether I should remove Poettering's code to use Tomazzi's.
However, I have never in my life used such sophisticated debugging code although it would be nice if I learn to. What I usually did was to code in a way to make debugging more accessible, then at the end and when the code worked as smoothly as I intended, I disabled or removed the code. C++ gives the functionality not only to do what I did, but to use very sophisticated mechanisms to improve one's code even in difficult to debug situations, where code can fail due to external factors like hardware and thread latencies and so on.
Regarding the application of such code I will need an example the best of which is an application of the code on uselessd itself.
Re: What should we do about systemd?
Posted: 2014-10-29 08:46
by chrissywissy
Edbarx,
Could you please add an option e) Accept systemd and continue using Debian
I won't necessarily vote for it, but in the interests of balance....
Re: What should we do about systemd?
Posted: 2014-10-29 08:59
by bodiless
I haven't voted because I am between the first two choices. The second option is more appealing to me as I find it difficult to leave debian. However, I have installed Slack in a virtual machine to get myself familiar with it in case I need to move on. I see from other threats that debian is reconsidering the choice of an init system. This puts me at the moment in a stand-by mode until the dust has settled.
Re: What should we do about systemd?
Posted: 2014-10-29 10:21
by kedaha
What should we do about systemd?
I haven't examined the matter in depth but here are my musings:
From what I've read about it, it seems to me that systemd along with its multiple dependences will serve to entrench Gnome3. Now that mate is available in Debian repositories I use this environment in wheezy but, since I use oss4, I wonder if, for example, it'll be so easy to remove alsa and pulseaudio from jessie so as to use the alternative oss4?
I'd be very averse to giving up using Debian - with or without systemd - for a server but, regarding the desktop, an obvious alternative would be to use and contribute to developing
kfreebsd-gnu.
Some users may like systemd, Gnome 3, pulse-audio and I respect their choice.
Other users, self included, prefer sysvinit and, for example, mate, oss4 or whatever.
To quote the
Social Contract:
We will be guided by the needs of our users and the free software community. We will place their interests first in our priorities.
Do the majority of users need systemd? Does the free software community need it to the exclusion of alternatives? Can it really be said that by implementing systemd that users' interests are being placed first? If not, whose interests?
I'm not at all sure whether option b) or d) would be sufficient but I'll vote for the former.
Re: What should we do about systemd?
Posted: 2014-10-29 17:02
by golinux
The problem with #2 - "concentrate on systemd's fork (uselessd) to port it to Debian" - is that it excludes the possibility of other solutions. uselessd might be the best solution or might not. There is strength in diversity and I think many ideas need to be explored. This is the beauty of open source. Eventually the winner(s) will become apparent and then the community can join forces and concentrate on those. It's gonna take time to sort out . . .
Re: What should we do about systemd?
Posted: 2014-10-29 17:30
by fleabus
golinux wrote:It's gonna take time to sort out . . .
Yes. To paraphrase bodiless' post above, I'm also currently in a holding pattern,
with Slackware now on most of my machines along with Wheezy or Wheezy-based.
Re: What should we do about systemd?
Posted: 2014-10-29 22:33
by edbarx
The serious bugs I am encountering whenever I try to install systemd on Jessie are reminding me of Ubuntu's crashes when I was a newbie. To me this is an indication that systemd is an immature software that should be avoided at all costs. This position is also held by knowledgeable people on these fora like Tomazzi who explained why systemd is dangerous to use as it doesn't suffficiently test for proper memory handling which is at the core of any software development.
Although the poll is indicating one to use systemd, technical reasons are pointing into the opposite direction. At the current state, systemd will reduce Debian Stable to the state of the Ubuntu distribution!
Users of systemd expect to experience OS crashes: systemd is still unsuitable to make part of Stable and the freeze is almost with us. This is unfortunately indicating that the viable option is option 1 unless I find a way to bypass systemd to use a more mature replacement.
Re: What should we do about systemd?
Posted: 2014-10-29 23:32
by Bulkley
Indulge my curiosity. Systemd is available in Debian Stable. Anyone tried it in Stable? Does systemd work better in Unstable than in Testing? Is there a progression or is systemd problematic everywhere? Testing is just that; is there a possibility that systemd will be held back from the next Stable?
Re: What should we do about systemd?
Posted: 2014-10-29 23:42
by dilberts_left_nut
No problems to report on my sid.
When it first arrived I discovered an old swap in fstab that shouldn't have been there.
Otherwise business as usual.
Re: What should we do about systemd?
Posted: 2014-10-30 06:40
by fruitofloom
I decided to vote with my feet. Hence my vote is option a).
I could work around systemd, i guess, say with refracta, but i lost trust that the project Debian acts in my very best interest. (my == user).
Re: What should we do about systemd?
Posted: 2014-10-30 09:30
by edbarx
debianxfce wrote:"The serious bugs I am encountering whenever I try to install systemd on Jessie are reminding me of Ubuntu's crashes when I was a newbie"
You are newbie if you parse os by yourself. In this image systemd works fine.
I agree with you: I am a newbie who can modify C/C++ code of essential parts of the OS with tangible proof as my edited version worked. My 'noobishness' makes me so obstinate in my ignorance of the system that I am considering changing parts of the system to continue using what I deem best as an OS's initialisation program. The latter can also evolve as one of my concoctions.
Re: What should we do about systemd?
Posted: 2014-10-30 09:47
by reinob
chrissywissy wrote:Could you please add an option e) Accept systemd and continue using Debian
I won't necessarily vote for it, but in the interests of balance....
I actually voted for that one. When I originally installed systemd-sysv (on sid) a few weeks ago I had a number of problems and none of the expected benefits (e.g. boot time, etc.)
Up until now I've considered systemd as an unnecessary bloat which didn't fit with what I consider Linux (even though I don't use Slackware anymore, I'll always be a Slackware fan).
BUT if you take the time to understand how things are supposed to work with systemd and makes the best of them, you will (perhaps) notice that it's not that bad at all, and it does have a number of advantages. One "problem" (which may also be an advantage) with Linux since the early days was the number of alternatives for doing the same thing. To name a few: power management and network configuration.
I find that with systemd you can IN ONE PLACE set-up when your computer should sleep, hibernate and whatever as well as how your network should be configured (I have a bonded eth0 and wlan0 and it works like a charm). This is what I consider true Linux/Unix. Not having to use graphical tools with limited functionality and dubious quality (Network Manager, WICD and XFCE4 Power Manager are but three examples), but a strong and supported method that works consistently and distribution-independent.
There, I'll stop selling systemd because I am not *that* impressed :)
But my advice is: give it a go. Don't judge Linux programs by the author or maintainer's attitude (otherwise you wouldn't even have a C library! :), even though I personally consider that a bit of a plan (and Lennart seems to have that one covered) is actually a good thing. Perhaps some day we'll have a part of Linux that "just works" without spending a significant chunk of our lives reading HOWTOs and tweaking here and there.
There, I said it.
Re: What should we do about systemd?
Posted: 2014-10-30 11:58
by bodiless
reinob wrote:But my advice is: give it a go.
Well, we all do, don't we? Otherwise we wouldn't be here reading, discussing and still using debian. Its only that, we are seriously worried about how things are developing.
Re: What should we do about systemd?
Posted: 2014-10-30 13:24
by n_hologram
systemd doesn't produce any noticeable errors on my Sid partition, but I have yet to try it without Gnome3 as I simply haven't had the time. That said, I will stop using Debian if Gnome3 dependencies or utilities (pulseaudio, gnome-network-manager, etc.) become unremoveable (I hate pulseaudio with a passion). Gnome3 is "pretty", but also severely inefficient and non-modular for my needs and wants.
I am in favor of forking systemd (ie, uselessd). I don't possess the skills to do so, but that direction has my full support. Again, although I haven't noticed any adverse effects, I stand against it on a philosophical level, because it violates everything that gives Linux its appeal.
Re: What should we do about systemd?
Posted: 2014-10-31 06:51
by edbarx
systemd will kill chroot which has enormous power for all those who use it. This is enough reason for any serious system adminstrator to throw out the infamous abomination that systemd is. According to knowledgeable coders, systemd's C++ code omits the most basic of all coding requirements: memory management. The latter is already giving off its stench, which I am quoting for reference:
Code: Select all
# apt-get install systemd
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following extra packages will be installed:
libcryptsetup4 libpam-systemd libsystemd0
Suggested packages:
systemd-ui
The following NEW packages will be installed:
libcryptsetup4 libpam-systemd libsystemd0 systemd
0 upgraded, 4 newly installed, 0 to remove and 173 not upgraded.
Need to get 0 B/2,834 kB of archives.
After this operation, 12.8 MB of additional disk space will be used.
Do you want to continue? [Y/n] y
Retrieving bug reports... Done
Parsing Found/Fixed information... Done
grave bugs of systemd (→ 215-5+b1) <Outstanding>
#765870 - systemd-logind brings system to knees with RAM consumption
Summary:
systemd(1 bug)
Are you sure you want to install/upgrade the above packages? [Y/n/?/...]
n
Bug 765870, quoted above, should be enough as an eye opener. Why on all the planets in the whole multiverse, should a login daemon eat all RAM bringing down the whole OS?!
The reason is obvious for anyone who has coding experience: careless memory management, which I repeat, also if necessary, ad nauseam, is at the core of all coding.
Software lock-ins and lock-outs as is the case with the rotting abomination of systemd, are effectively treating users like toddlers: handholding is for toddlers; adults, usually, can think for themselves.
software-lock-ins-and-lock-outs != progress
Re: What should we do about systemd?
Posted: 2014-10-31 07:51
by edbarx
This is Jessie with systemd removed. However, some packages depend on functions provided by systemd implying there are two packages providing these functions.
Code: Select all
rc libsystemd-daemon0:amd64 215-5+b1 amd64 systemd utility library (deprecated)
rc libsystemd-journal0:amd64 208-8 amd64 systemd journal utility library
rc libsystemd-login0:amd64 215-5+b1 amd64 systemd login utility library (deprecated)
ii libsystemd0:amd64 215-5+b1 amd64 systemd utility library
rc systemd 208-8 amd64 system and service manager
ii systemd-shim 8-4 amd64 shim for systemd
I have the impression libsystemd0 is not necessary but I will have to check for that.
If anyone knows how to completely remove all components from systemd, don't hesitate to post.
Added Later:
Code: Select all
$ apt-cache policy libsystemd-daemon0
libsystemd-daemon0:
Installed: (none)
Candidate: 215-5+b1
Version table:
215-5+b1 0
500 http://ftp2.fr.debian.org/debian/ jessie/main amd64 Packages
100 /var/lib/dpkg/status
$ apt-cache policy libsystemd-journal0
libsystemd-journal0:
Installed: (none)
Candidate: 215-5+b1
Version table:
215-5+b1 0
500 http://ftp2.fr.debian.org/debian/ jessie/main amd64 Packages
208-8 0
100 /var/lib/dpkg/status
$ apt-cache policy libsystemd-login0
libsystemd-login0:
Installed: (none)
Candidate: 215-5+b1
Version table:
215-5+b1 0
500 http://ftp2.fr.debian.org/debian/ jessie/main amd64 Packages
100 /var/lib/dpkg/status
$ apt-cache policy systemd
systemd:
Installed: (none)
Candidate: (none)
Package pin: (not found)
Version table:
215-5+b1 -1
500 http://ftp2.fr.debian.org/debian/ jessie/main amd64 Packages
208-8 -1
100 /var/lib/dpkg/status
Re: What should we do about systemd?
Posted: 2014-10-31 08:12
by fruitofloom
Me thinks libsystemd0 will swallow all of the other systemd libs (at least that happened in Sid)
I think:
1) ping systemd*
2) install the bsdutils by dzz
3) reinstall all apps which will get removed due to the lack of libsystemd0
1) kinda:
Package: "*systemd*"
Pin: release o=Debian
Pin-Priority: -1
~
2) packaged by dzz (kudos):
http://refracta.freeforums.org/going-wi ... html#p4115
That's as far as i could get. Other solutions are to be found on the net.
---
As far i understood a post at debian-user, Debian per default will enable all compile flags, hence libsystemd0 too. Packages may have it as a Depends, but will work just as well with no *systemd* installed (example for me: wicd).
Re: What should we do about systemd?
Posted: 2014-10-31 09:43
by edbarx
Me thinks libsystemd0 will swallow all of the other systemd libs (at least that happened in Sid)
These are the library's dependencies for Jessie:
Code: Select all
# ldd /lib/x86_64-linux-gnu/libsystemd.so.0
linux-vdso.so.1 (0x00007fff7b7c3000)
librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007f4faaf8f000)
liblzma.so.5 => /lib/x86_64-linux-gnu/liblzma.so.5 (0x00007f4faad6c000)
libgcrypt.so.20 => /lib/x86_64-linux-gnu/libgcrypt.so.20 (0x00007f4faaa8a000)
libresolv.so.2 => /lib/x86_64-linux-gnu/libresolv.so.2 (0x00007f4faa873000)
libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f4faa66f000)
libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f4faa451000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f4faa0a8000)
/lib64/ld-linux-x86-64.so.2 (0x00007f4fab3cf000)
libgpg-error.so.0 => /lib/x86_64-linux-gnu/libgpg-error.so.0 (0x00007f4fa9e96000)
And these are the library's dependencies for SID:
Code: Select all
$ ldd libsystemd.so.0.3.1
linux-vdso.so.1 (0x00007fff17dfe000)
librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007f27c9294000)
liblzma.so.5 => /lib/x86_64-linux-gnu/liblzma.so.5 (0x00007f27c9071000)
libgcrypt.so.20 => /lib/x86_64-linux-gnu/libgcrypt.so.20 (0x00007f27c8d8f000)
libresolv.so.2 => /lib/x86_64-linux-gnu/libresolv.so.2 (0x00007f27c8b78000)
libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f27c8974000)
libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f27c8756000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f27c83ad000)
/lib64/ld-linux-x86-64.so.2 (0x00007f27c96d4000)
libgpg-error.so.0 => /lib/x86_64-linux-gnu/libgpg-error.so.0 (0x00007f27c819b000)
Providing a consistent interface for packages to interface with systemd and any other services is a positive for which Debian should be encouraged rather than discouraged. Stating this doesn't do away with my criticism of systemd, as I consider systemd still ALPHA software unsuitable for a world famous distribution like Debian. I also criticize certain decisions that put undue emphasis on Gnome as a desktop, and on imagining the desktop is the only computer experience to be catered for. The reason for this position is the fact that CHOICE helps the greatest number of people to work in an environment they prefer most maximizing their concentration and output.
Re: What should we do about systemd?
Posted: 2014-10-31 14:21
by PeterB
I think it is the
reverse dependencies you need to investigate. Try;-
Code: Select all
apt-cache rdepends --installed libsystemd0
Re: What should we do about systemd?
Posted: 2014-10-31 15:19
by mmix
in my linux book, majority means nothing if situation gonna like this,
be systemd or not, i don't need systemd/uselessd.