21
|
|
|
Martin Pitt |
111 |
13 years ago
|
|
|
20
|
|
|
Martin Pitt |
110 |
13 years ago
|
|
|
19
|
|
|
Martin Pitt |
109 |
13 years ago
|
|
|
18
|
|
|
Martin Pitt |
108 |
13 years ago
|
|
|
17
|
|
[ Peter Eisentraut ] * Add initial support for 9.0 (pre-release snapshots for now): - cleanpg: Remove 9.0 clusters. - pg_createcluster: Clone configure_8_4() as configure_9_0(). - pg_upgradecluster: Migrate obsolete/changed configuration parameters. * pg_wrapper: Fix spelling. * In maintainer scripts, instead of "#!/bin/sh -e" use a separate "set -e", per lintian. * debian/control: Bump Standards-Version to 3.8.4; no changes necessary.
[ Martin Pitt ] * cleanpg: Do not remove the /var/log/postgresql/ directory itself, only its contents, to avoid logcheck falling over. * t/020_create_sql_remove.t: Do not consider $PG_GRANDPARENT_PID an unsafe environment, it's set by 9.0 servers. * t/060_obsolete_confparams.t: Add a full configuration for 8.4, so that upgrades to 9.0 can be tested. * t/052_upgrade_encodings.t: Specify 'C' locale when creating the ASCII encoded test database for versions >= 8.4, otherwise we have a locale/encoding mismatch. * t/TestLib.pm: Abort tests on FAILURE=shell and the shell exits with nonzero. This avoids endless followup error shells once it's clear that the following ones are doomed anyway. * PgCommon.pm: Add new function get_db_locales() which reads lc_ctype and lc_locales for a particular database. This replaces the get_cluster_locales() function for >= 8.4 server versions (which moved from global to per-database locales). * pg_upgradecluster: Use get_db_locales() instead of get_cluster_locales() for upgrades from >= 8.4. * pg_ctlcluster: Drop check for valid locales. We can only determine them (using pg_controldata through get_cluster_locales()) for <= 8.3 clusters, which are obsolete in sid/lenny. For >= 8.4 clusters, we need to start the server to find out the locales, at which point the check is pointless. * pg_upgradecluster: Run pg_restore with --disable-triggers when running in --data-only mode. (Closes: #579768) * t/020_create_sql_remove.t: We now assume that /var/log/postgresql/ always exists, and is writable by the "postgres" user. Thus if we delete the log, running "pg_ctlcluster start" should be able to recreate it. This fixes the 020_create_sql_remove.t test failure in "umask 077" mode, and uncovers some more inconsistencies wrt. log directory permissions. * pg_createcluster, testsuite: Fix permissions of /var/log/postgresql/ if this script creates it.
|
Martin Pitt |
107 |
14 years ago
|
|
|
16
|
|
* t/001_packages.t: Check for ".utf8" locale instead of ".UTF-8". This seems to be the new canonical spelling now. * t/TestLib.pm, check_clean(): Require that /var/log/postgresql/ still exists. If we remove it entirely, current versions of logcheck complain loudly. This was already fixed a while ago in squeeze, this will ensure it stays so. (Closes: #576180) * t/040_upgrade.t: Check upgrade of database and table ACL, an ALTER DATABASE option, and correct upgrading of a custom pg_hba.conf. This reproduces #543506. * pg_upgradecluster: Rework logic to use "pg_dumpall -s" instead of "pg_dumpall -g", to also catch ACLs, global settings, and the like. Drop manual database creation and encoding special-cases, since those are being taken care of by the -s commands already. (Closes: #543506, #562676) * pg_upgradecluster: Re-enable connections after the upgrade is done, not in between. * t/120_pg_upgradecluster_scripts.t: Upgrade scripts should not create tables in the "init" phase, since they will be overwritten during dump/restore. Change the test case to move table changes into the "finish" phase, and update documentation in pg_upgradecluster to point out this potential problem. * debian/init.d-functions, status(): More appropriate output formatting for "status" init script commands. (Closes: #522679)
|
Martin Pitt |
106 |
14 years ago
|
|
|
15
|
|
|
Martin Pitt |
105 |
14 years ago
|
|
|
14
|
|
|
Martin Pitt |
104 |
14 years ago
|
|
|
13
|
|
* Move the lsb-release dependency of p-common to a recommends of p-client-common. (Closes: #562642) * Drop Oliver from Uploaders:, he has not been active any more for years. * t/005_PgCommon.t: Add '.' to library search path, so that this prefers testing PgCommon.pm from the source tree. * t/005_PgCommon.t: Add test cases for {set,disable,replace}_conf_value(). * PgCommon.pm, replace_conf_value(): Do not append the new parameter if the old is not found. * PgCommon.pm, {set,disable,replace}_conf_value(): Intercept errors on writing data (which could happen when running out of disk space). Write into a new file and rename it in the end, instead of directly writing into the original file. Thanks to Yann Dirson for the original patch. (Closes: #549206) * pg_createcluster: Change effective gid as well when setting the socket directory, to avoid moving postgresql.conf's group to root (which would happen with the previous change). * Add t/042_upgrade_tablespaces.t: Test upgrading a cluster with tablespaces. Since this is not currently supported (and hard to do in an automated fashion due to the nature of tablespaces), just check that this errors out with a sane message and does not do any damage. * pg_upgradecluster: Add an early check for tablespaces. (Closes: #523574) * pg_lsclusters: Fix undefined value if owneruid cannot be determined (which can happen in special setups). Thanks to Daniel Pittman! (Closes: #541434) * t/020_create_sql_remove.t: Verify permissions of the data and configuration directories. * pg_createcluster: Make /etc/postgresql/<version> and /var/lib/postgresql/<version> owned by 'postgres', so that they can be administrated with 'postgres' privileges. Update t/030_errors.t accordingly. (LP: #236704, Closes: #525294) * t/020_create_sql_remove.t: Check that a missing log file is recreated by pg_ctlcluster (if it has enough permissions). * pg_ctlcluster: Make re-creation of log file actually work. (LP: #391119, #372476)
|
Martin Pitt |
103 |
14 years ago
|
|
|
12
|
|
|
Martin Pitt |
102 |
14 years ago
|
|
|
11
|
|
|
Martin Pitt |
101 |
14 years ago
|
|
|
10
|
|
|
Martin Pitt |
100.1 |
14 years ago
|
|
|
9
|
|
|
Martin Pitt |
100 |
14 years ago
|
|
|
8
|
|
|
Martin Pitt |
|
15 years ago
|
|
|
7
|
|
|
Martin Pitt |
|
15 years ago
|
|
|
6
|
|
* debian/supported-versions: Add "Debian 5.0" aka Lenny. (Closes: #509144) * debian/README.Debian: Document port handling, and point to relevant manpages and tools. (Closes: #508977) * debian/README.Debian: Fix "confident" typo. (Closes: #512648) * Drop pg_autovacuum handling, which was only necessary for 7.4 (8.0 and above have internal autovacuuming). This was kept for Lenny to allow Etch backports. This also gets rid of pg_maintenance and /etc/cron.d/postgresql-common. (Closes: #425914, #481025) * Add debian/postgresql-common.preinst: Remove obsolete conffiles (cronjob and /etc/postgresql-common/autovacuum.conf) on upgrade. * Drop support for pre-8.1 clusters, together with all hacks and workarounds for those. Add Conflicts: to postgresql-{7.4,8.0}, to ensure that this version isn't used with ancient servers any more. * t/030_errors.t: Check that clusters on the same port can run side by side if they are using different Unix socket directories and different TCP addresses. This reproduces #514132. * pg_ctlcluster: Replace overly harsh port conflict check (which broke clusters on the same port, but different Unix/TCP namespaces) with a more modest one which just checks conflict on the same Unix socket directory. Thanks to Bernd Helmle for the patch! (Closes: #514132, #472627) * debian/postgresql-common.postinst: Do not call pg_updatedicts with full path (DP 6.1). * pg_lsclusters, pg_upgradecluster: Fix forgotten "=back" after itemize list in the POD. Thanks lintian. * debian/compat, debian/control: Bump compat level to 6. * pg_updatedicts: Ensure generated tsearch dictionaries are world readable when being generated under umask 077. * debian/README.Debian: Point out incompatibility between using the upstream tools (initdb) and the Debian tools (pg_createcluster) and give some recommendations. (LP: #138793) * debian/maintscripts-functions: Unset $GREP_OPTIONS. Thanks to Carlo Calderoni for noticing!
|
Martin Pitt |
|
15 years ago
|
|
|
5
|
|
|
Martin Pitt |
94 |
15 years ago
|
|
|
4
|
|
|
Martin Pitt |
93 |
15 years ago
|
|
|
3
|
|
|
Martin Pitt |
91 |
15 years ago
|
|
|
2
|
|
|
Martin Pitt |
71 |
17 years ago
|
|
|