Re: [arch-general] /var/run should not be contained in the package file list.
On Tue, Feb 28, 2012 at 6:01 PM, Christian Stadegaart wrote: > What exactly is bind-mounted? Current mount information shows: > > /run on /var/run type tmpfs (rw,nosuid,nodev,relatime,mode=755) > > I presume this is a bind-mount? Correct. > So what you're saying is that I should boot > with SysV once to convert it into a symlink? "should" might be a bit strong ;-) You can do it, and at least I find the result somewhat tidier. It can be argued that it does not matter at all... -t
Re: [arch-general] /var/run should not be contained in the package file list.
Op 23-02-12 23:45, Tom Gundersen schreef: On Thu, Feb 23, 2012 at 11:40 PM, Damjan wrote: g: * /var/run is a symlink (created on boot) to /run. This should be changed in the future so the symlink is shipped with the filesystem package, but we have not figured out the transtion yet. How would that work with systemd where /run is bind mounted on /var/run (so it requires a /var/run directory) ? That's not a problem. systemd checks if /var/run is a symlink, and if so leaves it alone (this is the preferred way of doing it). Only if /var/run is a directory, it will by bind-mounted, giving more or less the same result as the symlink. Deleting /var/run and replacing it with a symlink is a bit of a hassle, so even if you are using systemd it might be worth-while to reboot once with initscripts so this will be done for you. -t copy was sent correctly> Hi, I'm running systemd. What exactly is bind-mounted? Current mount information shows: /run on /var/run type tmpfs (rw,nosuid,nodev,relatime,mode=755) I presume this is a bind-mount? So what you're saying is that I should boot with SysV once to convert it into a symlink? Thanks, Christian.
Re: [arch-general] /var/run should not be contained in the package file list.
On Thu, Feb 23, 2012 at 11:40 PM, Damjan wrote: > g: >> * /var/run is a symlink (created on boot) to /run. This should be >> changed in the future so the symlink is shipped with the filesystem >> package, but we have not figured out the transtion yet. > > How would that work with systemd where /run is bind mounted on /var/run > (so it requires a /var/run directory) ? That's not a problem. systemd checks if /var/run is a symlink, and if so leaves it alone (this is the preferred way of doing it). Only if /var/run is a directory, it will by bind-mounted, giving more or less the same result as the symlink. Deleting /var/run and replacing it with a symlink is a bit of a hassle, so even if you are using systemd it might be worth-while to reboot once with initscripts so this will be done for you. -t
Re: [arch-general] /var/run should not be contained in the package file list.
> Currently we have the following: > * /var/run is a symlink (created on boot) to /run. This should be > changed in the future so the symlink is shipped with the filesystem > package, but we have not figured out the transtion yet. How would that work with systemd where /run is bind mounted on /var/run (so it requires a /var/run directory) ? -- дамјан
Re: [arch-general] /var/run should not be contained in the package file list.
On Feb 22, 2012 7:05 PM, "Leonid Isaev" wrote: > > On Wed, 22 Feb 2012 18:33:24 +0100 > Tom Gundersen wrote: > > > On Wed, Feb 22, 2012 at 5:52 PM, Leonid Isaev wrote: > > > The package-owner it 'filesystem'. But I disagree with the proposal because > > > > I'm not really sure what you are disagreeing with... > > > > > (1) /var/run is a symlink, not a separate FS and (2) initscripts should > > > accomplish the minimum required operations and should mount only absolutely > > > necessary FS (not /tmp, /media or the like). > > > > Currently we have the following: > > * /var/run is a symlink (created on boot) to /run. This should be > > changed in the future so the symlink is shipped with the filesystem > > package, but we have not figured out the transtion yet. > > Ah, OK my information seems to be old. But fs package (2011.12) used to have a > symlink, right? Why was this changed? No, it was only in svn, but had to revert it until we figure out the upgrade path. > > * /run is a tmpfs, so if packages contain files in /var/run or in > > /run, they will not survive a reboot. They should use the tmpfiles > > mechanisem which we added for this purpose. > > * traditionally rc.sysinit deleted the contents of /var/run, > > /var/lock and /tmp on boot, we have now a simpler and cleaner > > situation since /var/run and /var/lock are on tmpfs. We did not force > > /tmp to be on tmpfs as it does not matter from the point of view of > > early boot. /media is probably going away in the long-run anyway, but > > does at any rate not have anything to do with boot, so no danger of > > that being touched. > > > > -t > > > -- > Leonid Isaev > GnuPG key ID: 164B5A6D > Key fingerprint: C0DF 20D0 C075 C3F1 E1BE 775A A7AE F6CB 164B 5A6D
Re: [arch-general] /var/run should not be contained in the package file list.
On Wed, 22 Feb 2012 18:33:24 +0100 Tom Gundersen wrote: > On Wed, Feb 22, 2012 at 5:52 PM, Leonid Isaev wrote: > > The package-owner it 'filesystem'. But I disagree with the proposal because > > I'm not really sure what you are disagreeing with... > > > (1) /var/run is a symlink, not a separate FS and (2) initscripts should > > accomplish the minimum required operations and should mount only absolutely > > necessary FS (not /tmp, /media or the like). > > Currently we have the following: > * /var/run is a symlink (created on boot) to /run. This should be > changed in the future so the symlink is shipped with the filesystem > package, but we have not figured out the transtion yet. Ah, OK my information seems to be old. But fs package (2011.12) used to have a symlink, right? Why was this changed? > * /run is a tmpfs, so if packages contain files in /var/run or in > /run, they will not survive a reboot. They should use the tmpfiles > mechanisem which we added for this purpose. > * traditionally rc.sysinit deleted the contents of /var/run, > /var/lock and /tmp on boot, we have now a simpler and cleaner > situation since /var/run and /var/lock are on tmpfs. We did not force > /tmp to be on tmpfs as it does not matter from the point of view of > early boot. /media is probably going away in the long-run anyway, but > does at any rate not have anything to do with boot, so no danger of > that being touched. > > -t -- Leonid Isaev GnuPG key ID: 164B5A6D Key fingerprint: C0DF 20D0 C075 C3F1 E1BE 775A A7AE F6CB 164B 5A6D signature.asc Description: PGP signature
Re: [arch-general] /var/run should not be contained in the package file list.
On Wed, Feb 22, 2012 at 5:52 PM, Leonid Isaev wrote: > The package-owner it 'filesystem'. But I disagree with the proposal because I'm not really sure what you are disagreeing with... > (1) /var/run is a symlink, not a separate FS and (2) initscripts should > accomplish the minimum required operations and should mount only absolutely > necessary FS (not /tmp, /media or the like). Currently we have the following: * /var/run is a symlink (created on boot) to /run. This should be changed in the future so the symlink is shipped with the filesystem package, but we have not figured out the transtion yet. * /run is a tmpfs, so if packages contain files in /var/run or in /run, they will not survive a reboot. They should use the tmpfiles mechanisem which we added for this purpose. * traditionally rc.sysinit deleted the contents of /var/run, /var/lock and /tmp on boot, we have now a simpler and cleaner situation since /var/run and /var/lock are on tmpfs. We did not force /tmp to be on tmpfs as it does not matter from the point of view of early boot. /media is probably going away in the long-run anyway, but does at any rate not have anything to do with boot, so no danger of that being touched. -t
Re: [arch-general] /var/run should not be contained in the package file list.
2012/2/22 郑文辉(Techlive Zheng) : > 2012/2/22 Allan McRae : >> On 22/02/12 12:24, 郑文辉(Techlive Zheng) wrote: >>> Due to /var/run now is a tmp filesystem, so the path /var/run/* should >>> not be packaged anymore. Instead, rc.script should take care aware >>> create needed /var/run stuff. Otherwise, there are a missing wraning >>> while running "pacman -Qk". >>> >> >> Bugs should be reported to the bug tracker... Which package is this? >> > > On my machine,runing "pacman -Ql | grep "/var/run" | cut -d' ' -f1 | > sort -u",got following: > > apache > avahi > dbus-core > filesystem > proftpd > samba > courier-authlib As Allan said, please file bugs against the relevant packages (the maintainers are probably not even reading this list). -t
Re: [arch-general] /var/run should not be contained in the package file list.
On Wed, 22 Feb 2012 10:55:53 +0800 郑文辉(Techlive Zheng) wrote: > 2012/2/22 Allan McRae : > > On 22/02/12 12:24, 郑文辉(Techlive Zheng) wrote: > >> Due to /var/run now is a tmp filesystem, so the path /var/run/* should > >> not be packaged anymore. Instead, rc.script should take care aware > >> create needed /var/run stuff. Otherwise, there are a missing wraning > >> while running "pacman -Qk". > >> > > > > Bugs should be reported to the bug tracker...Which package is this? > > > > On my machine,runing "pacman -Ql | grep "/var/run" | cut -d' ' -f1 | > sort -u",got following: > > apache > avahi > dbus-core > filesystem > proftpd > samba > courier-authlib The package-owner it 'filesystem'. But I disagree with the proposal because (1) /var/run is a symlink, not a separate FS and (2) initscripts should accomplish the minimum required operations and should mount only absolutely necessary FS (not /tmp, /media or the like). -- Leonid Isaev GnuPG key ID: 164B5A6D Key fingerprint: C0DF 20D0 C075 C3F1 E1BE 775A A7AE F6CB 164B 5A6D signature.asc Description: PGP signature
Re: [arch-general] /var/run should not be contained in the package file list.
2012/2/22 郑文辉(Techlive Zheng) : > 2012/2/22 Allan McRae : >> On 22/02/12 12:24, 郑文辉(Techlive Zheng) wrote: >>> Due to /var/run now is a tmp filesystem, so the path /var/run/* should >>> not be packaged anymore. Instead, rc.script should take care aware >>> create needed /var/run stuff. Otherwise, there are a missing wraning >>> while running "pacman -Qk". >>> >> >> Bugs should be reported to the bug tracker... Which package is this? >> > > On my machine,runing "pacman -Ql | grep "/var/run" | cut -d' ' -f1 | > sort -u",got following: > > apache > avahi > dbus-core > filesystem > proftpd > samba > courier-authlib i think something like pacman -Ql | grep -P '/var/run/.+' is more appropriate in this case.
Re: [arch-general] /var/run should not be contained in the package file list.
2012/2/22 Allan McRae : > On 22/02/12 12:24, 郑文辉(Techlive Zheng) wrote: >> Due to /var/run now is a tmp filesystem, so the path /var/run/* should >> not be packaged anymore. Instead, rc.script should take care aware >> create needed /var/run stuff. Otherwise, there are a missing wraning >> while running "pacman -Qk". >> > > Bugs should be reported to the bug tracker...Which package is this? > On my machine,runing "pacman -Ql | grep "/var/run" | cut -d' ' -f1 | sort -u",got following: apache avahi dbus-core filesystem proftpd samba courier-authlib
Re: [arch-general] /var/run should not be contained in the package file list.
On 22/02/12 12:24, 郑文辉(Techlive Zheng) wrote: > Due to /var/run now is a tmp filesystem, so the path /var/run/* should > not be packaged anymore. Instead, rc.script should take care aware > create needed /var/run stuff. Otherwise, there are a missing wraning > while running "pacman -Qk". > Bugs should be reported to the bug tracker...Which package is this?