Hi! On Fri, May 26, 2017 at 11:28:48AM +0200, Ivo De Decker wrote: > If a package leaves the system in a broken state, that is very much release > critical. So I'm upgrading the bug to serious. It should probably be > 'critical' (breaks the whole system), but as both are RC, that doesn't matter > that much.
That is correct and I'm not arguing that per se. > > As the init system is a rather fundamental component of a Linux > > distribution, it affects many other packages, directly or indirectly > > and it's therefore too much of a burden to provide support for all > > init systems available in Debian. Although runit is available in > > Debian, it does not mean that it has to be fully supported. > > If an init system is shipped in a stable release, it has to be supported. > Otherwise it should not be in a stable release. Well, there is also ulibc being shipped with Debian stable. Yet, when someone tries to use it and breaks their system, it's not supported either. So, I don't think this policy can be sweepingly applied to every package. > > The fact that it is in Debian is merely of the result of Debian's policy to > > not limit packages from entering the distribution unless the license or > > other serious concerns prevent it. > > No. If it's broken, it should not be in a stable release. We can still remove > this package. Which hasn't happened yet. The package has been open with an RC bug for longer than 10 days and yet it wasn't removed from testing. > > This policy is a result of Debian's decision to adopt systemd as > > its default init system [1] as well as the follow up general > > resolution [2] where Debian Developers decided that providing > > support for alternative init systems was not mandatory. > > > Furthermore, as also already explained, the problem you have run > > into cannot really be trivially solved as the installing runit > > does not replace the running instance of systemd with runit so > > it is to be expected for commands like 'poweroff' and 'reboot' > > to not work until the machine has been rebooted. > > The fact that problems cannot be trivially solved doesn't mean they are not > RC. Also, I don't see how it can be 'expected' that reboot doesn't work until > you reboot. How is this supposed to work? Doing a hard reset of the machine? I'm rather curious to hear how you suggest to resolve this issue then. If there was a trivial way, then there'd be a patch for this already. You cannot replace the running init system without rebooting the machine, simply because the kernel will panic the moment PID 1 exits. What the original bug reporter here is asking to work simply cannot work. You install runit, it removes systemd and overwrites the symbolic links for 'shutdown' and 'poweroff' while systemd is still running. It is therefore inevitable that these commands don't work until you reboot the machine. And, yes, if a user desires to introduce such a breaking change to their system, they are on their own. We *did* decide to focus on systemd after all, so I don't see why we should care if users intentionally decide to break their systems. > > A possible solution would be to modify the runit postinst scripts > > in a way that it does not automatically overwrite the symlinks > > for the the above commands until the machine has been rebooted > > (e.g. by placing a script which is run only once after the system > > has been first rebooted with runit) so that the 'poweroff' and > > 'reboot' commands are still sent to systemd. However, the lack of > > a reply of the runit maintainer to this particular bug report seems > > to indicate that there is currently no interest for such a solution. > > If the maintainer isn't interested in making sure that this package works as > expected, it isn't fit for a stable release... I fully agree. However, runit is one of the packages which is not automatically removed. > > Thus, in order to prevent this bug report from blocking the release > > of Debian Stretch, I have reduced its severity to 'normal'. You > > are still welcome to propose a patch to address this issue though, > > it's just not relevant for the upcoming Debian release. > > This is not a good reason to downgrade a bug. Again, Debian has decided to adopt systemd as the standard init system, the same way we have decided to adopt glibc and the Linux kernel as the standard C libraries and kernels. You really cannot expect a fundamental component like an init system to be easily replace by the end user the same way they can swap their default text editor. Thanks, Adrian -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer - glaub...@debian.org `. `' Freie Universitaet Berlin - glaub...@physik.fu-berlin.de `- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
