[ Raphaël Hertzog ] * Fix dpkg-source regression in "3.0 (quilt)" source packages while unapplying patches that remove all files in a directory. Closes: #683547 * Fix segfault in field format parsing on empty strings, affecting «dpkg-query -W -f ''» and «dpkg-deb -W --showformat=''». LP: #1035512 * Fix dpkg's French usage string which was missing the final “s“ in --print-foreign-architectures. Closes: #685863
[ Guillem Jover ] * Use “statoverrides” instead of “statusoverrides” in dpkg-statoverride. Closes: #686995 * Comment out dpkg(1) documentation about disabled --command-fd option. Closes: #685677 * Cleanup dpkg-divert unit-test environment to avoid build failures. Closes: #687656 * Fix update-alternatives test suite to behave correctly on non-Debian binary paths. Known to be affecting at least Gentoo and Mac OS X. * Do not leak subcall command arguments in update-alternatives. * Fix segfault on update-alternatives when passing --slave without any action at all. LP: #1037431 * Fix memory leak in dpkg filesavespackage(). * Do not print garbage (or worse) on dpkg shared conffile debug output. * Use a hash instead of a ref to a hash for keys() in Dpkg::BuildFlags get_feature_areas(). This causes compilation failures with older perl versions, which can be an issue with partial upgrades. * Fix filter subpattern debug output format string to print an actual value instead of just blanks. * Ignore trailing filter subpattern slashes on reinclusion comparison. This makes sure to reinclude directories previously excluded so that contained files marked for inclusion do not fail to unpack due to a missing directory. Closes: #688416 * Do not consider obsolete conffiles as actively owned by the package. This ensures conffile entries are not misshandled nor mixed up when configuring packages owning the non-obsolete conffiles. Closes: #689836 Based on a patch by Andreas Beckmann <debian@abeckmann.de>. * Properly mark in the database obsolete conffiles on package replaces. * Sync the Conffiles field values for all package instances. Because only the first package instance being configured will have a *.dpkg-new conffile available to be processed, the subsequent ones need to use the hash from the previously processed entries. * Fix logic for previously configured conffiles, so that the shared conffile checks actually work on reinstallation. Closes: #684776 * Avoid info database corruption and bogus accesses on unknown format values, by always reading the format file and validating it. * Clarify that the most probable reason for multiarch database inconcistencies is due to upgrades from unofficial dpkg versions. * Only satisfy a dependency on a “Multi-Arch: foreign” if arch-unqualified. * Take architecture into account in virtual packages on remove and configure dpkg actions. Closes: #683411 * Update update-alternatives --query format and examples in man page to match the implementation. * Add two missing 3rd person ‘s’ in dpkg-gensymbols(1). Closes: #689863 Thanks to Paul Menzel <pm.debian@googlemail.com>. * Fix regression on old-style binNMUs for packages that specify an explicit binary version to dpkg-gencontrol, by always fixing up the source version. Closes: #690823