Bug#619650: File descriptor 4 (pipe:[556425]) leaked on lvs invocation. Parent PID 17535: /bin/sh

[Colin Watson]

On Fri, Mar 25, 2011 at 03:26:10PM -0500, Jonathan Nieder wrote:
> Typical experimental amd64 system.  Some warnings from today's
> upgrade.  This might be a duplicate of bug#592834 and bug#549976, but
> I'm filing it separately just in case.
> 
> | # cupt full-upgrade
> [...]
> | Setting up grub-pc (1.99~rc1-5) ...
> | Installation finished. No error reported.
> | Generating grub.cfg ...
> | Found background image: /usr/share/images/desktop-base/desktop-grub.png
> | Found linux image: /boot/vmlinuz-2.6.38-rc6-00039-g61d2fb7
> [... etc ...]
> | Found initrd image: /boot/initrd.img-2.6.37-trunk-amd64
> | File descriptor 4 (pipe:[556425]) leaked on lvs invocation. Parent PID 17535: /bin/sh
> | File descriptor 5 (/var/lib/dpkg/status (deleted)) leaked on lvs invocation. Parent PID 17535: /bin/sh
> | File descriptor 6 (/var/lib/apt/lists/mirrors.kernel.org_debian_dists_squeeze_main_binary-amd64_Packages) leaked on lvs invocation. Parent PID 17535: /bin/sh
> | File descriptor 7 (/var/lib/apt/lists/security.debian.org_dists_squeeze_updates_main_binary-amd64_Packages) leaked on lvs invocation. Parent PID 17535: /bin/sh
> | File descriptor 8 (/var/lib/apt/lists/mirrors.kernel.org_debian_dists_squeeze-proposed-updates_main_binary-amd64_Packages) leaked on lvs invocation. Parent PID 17535: /bin/sh
> | File descriptor 9 (/var/lib/apt/lists/mirrors.kernel.org_debian_dists_sid_main_binary-amd64_Packages) leaked on lvs invocation. Parent PID 17535: /bin/sh
> | File descriptor 10 (/var/lib/apt/lists/mirrors.kernel.org_debian_dists_experimental_main_binary-amd64_Packages) leaked on lvs invocation. Parent PID 17535: /bin/sh
> | File descriptor 11 (/var/lib/cupt/lock) leaked on lvs invocation. Parent PID 17535: /bin/sh
> | done
> | (Reading database ... 124124 files and directories currently installed.)
> 
> Bugscript output attached.  Any hints for tracking this down? 

Looks to me like cupt's fault for leaking all those fds through to
subprocesses, rather than marking them close-on-exec.  Do you agree?  If
so, it should be reassigned to cupt.

(lvs is pretty whiny about this, and it isn't strictly necessary for it
to be, but it does occasionally show up real bugs.  Though it's slightly
odd that you see this since os-prober sets the environment variable
that's supposed to suppress this.)

-- 
Colin Watson                                       [cjwatson at debian.org]