Hmm, good catch. I think you should call this out as a separate change in the changelog, since this isn't just a "remaining change" but a fix to multiarch support for the runtime lib package.
--- debian/lvm2.postinst 2009-10-08 18:17:43 +0000
+++ debian/lvm2.postinst 2012-08-15 17:05:32 +0000
@@ -5,7 +5,9 @@
case "$1" in
configure) vgcfgbackup >/dev/null 2>&1 || :
- invoke-rc.d lvm2 start || :
+ if [ -x /etc/init.d/lvm2 ]; then
+ invoke-rc.d lvm2 start || :
+ fi
if [ -x /usr/sbin/update-initramfs ]; then update-initramfs -u
fi
Since the lvm2.init is being dropped, this probably shouldn't be conditional - it should probably be removed entirely. In fact, when this regression was introduced during the lucid merge, the previous postinst was doing this:
if test -f /etc/init.d/lvm2; then
update-rc.d -f lvm2 remove >/dev/null 2>&1 || true
rm -f /etc/init.d/lvm2
fi
We probably want to be doing that again, possibly with better handling of modified conffiles.
BTW, perhaps it's worth checking with Kees to see if there was a reason he thought this init script should be restored when merging.
This is unfortunate, but it happens... In this case, it appears that dmeventd used the old symbol, so we want to have a Breaks against the old version of dmeventd. No other packages in Ubuntu that depend on libdevmapper-event1.02.1 appear to use that symbol, so this Breaks is the only fix needed.
> - debian/ {clvmd. ra,clvm. init}:
> - create /var/run/lvm if it doesn't exist.
debian/rules now points at /run/lvm; this should be updated to match.
--- debian/ libdevmapper- dev.install 2010-12-07 08:08:45 +0000 libdevmapper- dev.install 2012-08-18 04:00:45 +0000 libdevmapper. h libdevmapper- event.h libdevmapper. so pkgconfig/ devmapper. pc pkgconfig/ devmapper- event.pc libdevmapper* */pkgconfig/ devmapper*
+++ debian/
@@ -1,5 +1,2 @@
-usr/include/
-usr/include/
-usr/lib/
-usr/lib/
-usr/lib/
+usr/include/
+usr/lib/
Hmm, good catch. I think you should call this out as a separate change in the changelog, since this isn't just a "remaining change" but a fix to multiarch support for the runtime lib package.
--- debian/ lvm2.postinst 2009-10-08 18:17:43 +0000 lvm2.postinst 2012-08-15 17:05:32 +0000
vgcfgbackup >/dev/null 2>&1 || : update- initramfs ]; then
update- initramfs -u
+++ debian/
@@ -5,7 +5,9 @@
case "$1" in
configure)
- invoke-rc.d lvm2 start || :
+ if [ -x /etc/init.d/lvm2 ]; then
+ invoke-rc.d lvm2 start || :
+ fi
if [ -x /usr/sbin/
fi
Since the lvm2.init is being dropped, this probably shouldn't be conditional - it should probably be removed entirely. In fact, when this regression was introduced during the lucid merge, the previous postinst was doing this:
if test -f /etc/init.d/lvm2; then
update-rc.d -f lvm2 remove >/dev/null 2>&1 || true
rm -f /etc/init.d/lvm2
fi
We probably want to be doing that again, possibly with better handling of modified conffiles.
BTW, perhaps it's worth checking with Kees to see if there was a reason he thought this init script should be restored when merging.
--- debian/ libdevmapper- event1. 02.1.symbols 2012-04-14 03:19:00 +0000 libdevmapper- event1. 02.1.symbols 2012-08-18 04:00:45 +0000 event.so. 1.02.1 libdevmapper- event1. 02.1 #MINVER# daemon_ fini_fifos@ Base 2:1.02.74 daemon_ init_fifos@ Base 2:1.02.74 daemon_ talk@Base 2:1.02.74
+++ debian/
@@ -1,11 +1,13 @@
libdevmapper-
Base@Base 2:1.02.20
- daemon_talk@Base 2:1.02.67
+ dm_event_
+ dm_event_
+ dm_event_
This is unfortunate, but it happens... In this case, it appears that dmeventd used the old symbol, so we want to have a Breaks against the old version of dmeventd. No other packages in Ubuntu that depend on libdevmapper- event1. 02.1 appear to use that symbol, so this Breaks is the only fix needed.
- dm_event_ get_version@ Base 2:1.02.67 get_version@ Base 2:1.02.74 handler_ create@ Base 2:1.02.20 handler_ destroy@ Base 2:1.02.20 handler_ get_dev_ name@Base 2:1.02.67 handler_ get_dev_ name@Base 2:1.02.74
+ dm_event_
dm_event_
dm_event_
- dm_event_
+ dm_event_
Those version bumps are unnecessary in Ubuntu, but mostly harmless.
@@ -22,5 +24,3 @@ handler_ set_uuid@ Base 2:1.02.20 register_ handler@ Base 2:1.02.20 unregister_ handler@ Base 2:1.02.20
dm_event_
dm_event_
dm_event_
- fini_fifos@Base 2:1.02.67
- init_fifos@Base 2:1.02.67
These symbols are also only used by dmeventd. (This makes sense, as all three symbols appear to be due to a Debian-specific patch for dmeventd.)
The rest looks good to me.