[ CCing debian-dpkg ]

i'm sorry Steve but a bug is sitting somewhere.

i made some tests using the following scripts:

$ cat check_boost 
#! /bin/sh

dpkg --get-selections | grep "libboost.*-dev" | awk '{print $1}' | xargs dpkg 
-L | for f in `cat -`; do if [ ! -e "$f" ]; then echo "$f"; fi; done
$ cat remove_boost
#! /bin/sh

dpkg --get-selections | grep libboost | awk '{print $1}' | xargs dpkg --purge
$

installing boost 1.32.0 from scratch with no older versions around lets
the system in the wanted state. all the files provided by the boost
packages are in place and check_boost doesn't print anything.

upgrading from 1.31.0 to 1.32.0 instead leaves some holes and not all
the files owned by boost packages are really present on the filesystem.
check_boost prints the list of missing files.

directly using dpkg to upgrade boost packages (ie. dpkg *1.32.0-2*.deb)
produces a different list of missing files than the one produced
upgrading with dselect. so probably the order in which the packages
are upgraded affects the list of missing files.

i saw the bug on all the 4 i386 and 1 sparc i upgraded. i'm able to
reproduce the bug on my own i386 but i still have to check the sparc
and at least another i386.  it looks reproducible and deterministic. i'd
like to see if missing files lists are the same across different archs
and boxes.

i'm suspecting of a dpkg bug but i need some further analysis before
i'm sure it is.

i attached the list of missing files i obtained performing upgrades
using both dpkg and dselect twice.

please double check the packages in order to exclude our (my)
responsability in this. i'll do the same.

cheers
domenico

-----[ Domenico Andreoli, aka cavok
 --[ http://people.debian.org/~cavok/gpgkey.asc
   ---[ 3A0F 2F80 F79C 678A 8936  4FEE 0677 9033 A20E BC50

Attachment: missing_files_dpkg1.txt.gz
Description: Binary data

Attachment: missing_files_dpkg2.txt.gz
Description: Binary data

Attachment: missing_files_dselect1.txt.gz
Description: Binary data

Attachment: missing_files_dselect2.txt.gz
Description: Binary data

Reply via email to