On Thu, Aug 2, 2012 at 22:38:13 +0200, Bertrand Marc wrote: > Le 02/08/2012 20:03, Julien Cristau a écrit : > >>diff -Nru gnunet-0.9.3/debian/changelog gnunet-0.9.3/debian/changelog > >>--- gnunet-0.9.3/debian/changelog 2012-06-20 23:55:23.000000000 +0200 > >>+++ gnunet-0.9.3/debian/changelog 2012-08-01 22:10:06.000000000 +0200 > >>@@ -1,3 +1,21 @@ > >>+gnunet (0.9.3-3) unstable; urgency=low > >>+ > >>+ * debian/control: update Vcs-* to the new repository in collab-maint. > >>+ * Install only the generated binaries on Hurd, thanks to Cyril Roelandt > >>+ (Closes: #670794). > >>+ * Use chmod and chown instead of dpkg-statoverride to set special permissions > >>+ and upgrade properly depending on the previous version (Closes: #673301). > >>+ * Rewrite gnunet-server.init based on /etc/init.d/skeleton and make > >>+ gnunet-server depend on lsb-base to use LSB logging. > > > >This part makes me extremely nervous. I don't think now is the time for > >this kind of change. > > That's why it was not uploaded to unstable, and only to mentors. I > guess we could select the changes we want to see in wheezy: Sorry, I meant the init script change. > - The fix to build on Hurd seems harmless to me. > - Getting rid of dpkg-statoverride would be nice, since its use was > quite wrong. In particular, in the current package, there is no way > to know whether the administrator or the gnunet package set the > dpkg-statoverrides. It is in Debian only for a few months, and it > would be nice not to mess up with dpkg-statoverride on Debian/stable > boxes. Ack, that's fine. > - The rewrite of gnunet-server.init is purely cosmetic: the current > package doesn't use LSB logging, so gnunet is the only line with the > old look at boot. I also added the status command, making lintian > happy. Clearly it is the less "urgent" change. > This is the one I would rather see deferred. > >>+ * Fix "gcc-4.6: [sparc] compiler fails to align stack-allocated > >>+ struct, with array of uint32-values to 32-bit boundary": > >>+ new patch sparc_alignment.patch, taken from upstream: > >>+ https://lists.gnu.org/archive/html/gnunet-svn/2012-07/msg00548.html > >>+ Thanks to Jurij Smakov for the analysis. > >>+ (Closes: #670578) > >>+ > >>+ -- Bertrand Marc<beberking@gmail.com> Wed, 01 Aug 2012 22:09:49 +0200 > >>+ > >> gnunet (0.9.3-2) unstable; urgency=low > >> > >> * Clean properly dpkg-statoverride in gnunet-server.postrm > >[...] > >>diff -Nru gnunet-0.9.3/debian/gnunet-server.postinst gnunet-0.9.3/debian/gnunet-server.postinst > >>--- gnunet-0.9.3/debian/gnunet-server.postinst 2012-06-07 23:13:17.000000000 +0200 > >>+++ gnunet-0.9.3/debian/gnunet-server.postinst 2012-07-07 15:50:27.000000000 +0200 > >>@@ -52,6 +52,24 @@ > >> echo " done." > >> fi > >> > >>+ # this can go away after wheezy > >>+ if dpkg --compare-versions "$2" le "0.9.3-2"&& dpkg --compare-versions "$2" ge "0.9.2-1"; then > >>+ for file in /usr/bin/gnunet-helper-exit \ > >>+ /usr/bin/gnunet-helper-fs-publish \ > >>+ /usr/bin/gnunet-helper-nat-client \ > >>+ /usr/bin/gnunet-helper-nat-server \ > >>+ /usr/bin/gnunet-helper-transport-wlan \ > >>+ /usr/bin/gnunet-helper-vpn \ > >>+ /usr/bin/gnunet-helper-dns \ > >>+ /usr/bin/gnunet-service-dns > >>+ do > >>+ if dpkg-statoverride --list $file>/dev/null 2>&1 > >>+ then > >>+ dpkg-statoverride --remove $file > >>+ fi > >>+ done > >>+ fi > >>+ > > > >How does this interact with statoverrides set by the local admin? > It doesn't. The use of dpkg-statoverride was wrong in the first > place, so there is no way to know whether the local admin set > statoverrides or whether the gnunet package did. And since postrm > removes the gnunet group and user, we need to be sure we remove > every statoverrides before postrm. Again, it is like this only since > April or May, and it has not reached Stable, yet. > Why does postrm do that? Is there any harm in leaving the user and group around? See also http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=621833 Cheers, Julien
Attachment:
signature.asc
Description: Digital signature