Bug#591291: some more thought from IRC

Mattia Rizzolo mattia at mapreri.org
Sat Aug 29 19:56:55 UTC 2015


more thoughts about this issue from IRC.
Thanks Andreas Henriksson for the feedback.


From #debian-devel @ 2015-08-29, UTC time:

[07:32:18 PM] <ah> mapreri: experimental pbuilder works fine here.... completely unrelated I've run into issues with /dev/shm though (which was incompletely moved to /run/shm without a compat symlink set up at /dev/shm).... Maybe I could convince you to look into the /dev/shm related bugs... (I pretty much suspect best thing would be to just move the mountpoint back to /dev/shm to avoid breakage).
[07:32:39 PM] <mapreri> cool!
[07:35:27 PM] <mapreri> ah: the /{dev,run}/shm thing is something really fishy. as you can see on bugs.d.o/pbuilder there are several related bugs. on a normal system there is a compat symlink created by some tool i don't remember anymore, which change behavior according to the init system and whatnot. i was thinking about setting a /run/shm → /dev/shm symlink when USERUNSHM=yes and /dev/shm is not already on BINDMOUNT. how does this sound?
[07:41:00 PM] <ah> mapreri: whatever works.... in my chroot the tmpfs was mounted on /run/shm and /dev/shm was just an empty directory.... that doesn't work..... not sure it's worth caring at all about /run/shm, but if you set up compat symlinks then everyone should be happy.
[07:42:50 PM] <mapreri> ah: but i don't really understand is shm now expected on /dev/shm or /run/shm?  can you provide me a simple package (as in: the builds finish in a reasonable time) that requires shm but file if only /run/shm is present?
[07:43:51 PM] <ah> mapreri: /run/shm is a debian-sysvinit invention..... some eager beavers started using /run/shm explicitly so I think systemd might have shipped a /run/shm -> /dev/shm compat symlink...
[07:44:15 PM] <mapreri> it's not systemd, is something else run a boot time
[07:45:19 PM] <mapreri> it's a mess, actually, it's shell script somewhere that do crazy check to catch weird cases i can't even think about, luckily i don't remember where that script is anymore
[07:45:56 PM] <ah> mapreri: /usr/lib/tmpfiles.d/debian.conf creates it if you run systemd....
[07:46:47 PM] <ah> mapreri: shell script mess sounds like the sysvinit/initscripts opposite setup.
[07:47:20 PM] <mapreri> yeah, maybe i read the sysvinit one
[07:47:26 PM] <mapreri> wasn't aware of that file
[07:47:53 PM] <mapreri> ah: does that sounds like #591291 ?
[07:47:58 PM] -zwiebelbot- Debian#591291: pbuilder needs to mount /dev/shm - https://bugs.debian.org/591291
[07:48:37 PM] <ah> mapreri: or #773765
[07:48:41 PM] -zwiebelbot- Debian#773765: pbuilder: Fails to mount /dev/shm in recently created chroot (tgz) - https://bugs.debian.org/773765
[07:49:06 PM] <mapreri> there is also #717022
[07:49:10 PM] -zwiebelbot- Debian#717022: pbuilder does not mount /dev/shm in squeeze chroots - https://bugs.debian.org/717022
[07:49:12 PM] <ah> mapreri: the newer one has some ramblings about the /run/shm vs /dev/shm mess.
[07:49:50 PM] <ah> mapreri: that one predates the move mess... and atleast to me it is mounted... just in the wrong location.
[07:49:56 PM] * mapreri does not like long bug logs
[07:51:27 PM] <mapreri> ah: i have a flight at 5 am to take me to my VACs, so i'm going to do anything RSN.  mind if i drop this irc stuff into #591291?
[07:51:50 PM] <ah> mapreri: feel free... no hurry.
[07:51:56 PM] <ah> mapreri: have a nice vacation
[07:51:56 PM] <mapreri> hopefully i'll wrap my mind around this issue someday
[07:52:28 PM] <mapreri> ah: FYI in jenkins.d.n we have /dev/shm in BINDMOUNDS to workaround it
[07:52:32 PM] <ah> mapreri: IMNSHO just say no to /run/shm.... it's not like we can ever get rid of /dev/shm anyway.
[07:53:03 PM] <mapreri> it's not that easy to do backward-incompatible changes like this :(
[07:53:20 PM] <mapreri> people get angry if you break their setup
[07:53:44 PM] <paultag> people look for things to be angry about
[07:54:02 PM] <ah> mapreri: you're just reverting previously introduced breakage though.. ;)
[07:54:22 PM] <mapreri> sigh


-- 
regards,
                        Mattia Rizzolo

GPG Key: 66AE 2B4A FCCF 3F52 DA18  4D18 4B04 3FCD B944 4540      .''`.
more about me:  http://mapreri.org                              : :'  :
Launchpad user: https://launchpad.net/~mapreri                  `. `'`
Debian QA page: https://qa.debian.org/developer.php?login=mattia  `-
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/pbuilder-maint/attachments/20150829/b0a7beba/attachment.sig>


More information about the Pbuilder-maint mailing list