VMM: The Python 2.4.x compatibility commit - part III.

VMM/account: Check the account's transport only when we have a transport.

VMM: The Python 2.4.x compatibility commit.

pgsql/set-permissions.py: Extended permissions. Postfix needs read access on the postfix_gid view. Dovecot needs read permissions on both mailboxformat columns.

VMM/handler: Restored method Handler.domain_transport(). It went somehow lost with changeset 5806fb74130b.

pgsql: Added helper script: set-permissions.py.

Provide list{addresses,aliases,users,relocated} subcommands The patch provides the list{addresses,aliases,users,relocated} subcommands to the UI. All commands actually take the same path in the code and only one query is run with different parameters for each case. There are still two shortcomings: 1. With alias domains, the output order is not as one might want it, e.g. foo@example.org bar@example.org foo@example.com bar@example.com when it should really be foo@ twice and then bar@ twice. I have not found a way to modify the SQL accordingly. 2. The SELECT queries for Accounts, Alias and Relocated are hard-coded in common.py.

man: escape hyphens Hyphens inside words must be spelt \- instead of just -. This patch fixes up the manpages. * * * .

man: add 'catchall' to domaininfo Since the addition of catchall, the domaininfo subcommand can also take 'catchall' as a [detail] to limit the output of details to the catch-all aliases. Also, the number of catch-all destinations is now included in the simple domaininfo output.

man: vmm1: Use `fqdn' instead of `domain' in catchall* descriptions. Fixed a typo s/cadd/cad/. Replaced double quotes by \(dq.

VMM/cli/subcommands: Use the `fqdn' argument placeholder also for catch-all subcommands.

Fix transport_maps function for non-existent domains The postfix_transport_maps function had a bug causing 2012-04-15 17:40:22 CEST LOG: statement: SELECT transport FROM postfix_transport_map('logcheck', 'domine.madduck.net'); 2012-04-15 17:40:22 CEST ERROR: query returned no rows when the domain was not in the database. This would make did be NULL and make the query fail. This patch moves the tid query until after a check for did. If the latter is NULL, the function RETURNs (rather than fails).

Revert caching of destination interpolation Since 'destination' comes from the table in the query, it cannot be cached across queries! Doh!

pgsql: update_tables_*: Added missing `NOT' to `… tid DROP NOT NULL;'.

pgsql: Create column `note' as text.

Add docstring for common.format_domain_default

Document domainnote in manpages 5806fb74130b did not contain this, so I must have forgotten it. Oops.

Remove domain/account notes from TODO

Include account note in getuser output Modify the getuser output to include an account note, if one is present

Display formatted notes in info output If the user/domain object has an attached note, this patch makes the userinfo/domaininfo output print the note, wrapped to the terminal size.

Ignore *.egg-status build directory created by setuptools

Add user/domainnote commands to manpages

Add note field to Account/Domain and CLI

Modify SQL to add note field to domain_data and users

man: vmm.1: subcommand's name argument is optional now. Based on martin f. krafft's v2 of changeset c340ba4bd544.

Allow username to take no argument to erase value Once a username has been set, it could not be removed via the UI. Now, if no argument is passed to username, the value stored is replaced with NULL.

Add inheritance as a feature to README

Change 'force' argument to reset user records for tid/ssid/qid If domain{quotalimit,transport,serviceset} are run with 'force' and settings are inheritable from the domain, then rather than to copy the ID to the user table, the affected field in the user records should be set to NULL so that the domain default is used.

Change UI to accept 'default' to restore inherited values Instead of explicit values for tid/ssid/qid, the UI now accepts 'default', which removed user-pecific settings and hence causes domain defaults to be used.

Modify userinfo output to indicate when domain defaults are displayed When Account instances reference NULL tid/qid/ssid, the data must come from the associated domain, and this should be indicated. For transport and services, this is easy to do as the string passed in the info dict can simply be modified. For quotalimit, however, another method must be used due to the CLI-side formatting. All approaches use a common formatter outsourced to the common.py file.

Modify Account class to handle NULL references This patch modifies the Account class to defer to using the associated domain's tid/ssid/qid fields if the per-instance fields are None/NULL.

Modify SQL update script to accept NULL fields for users The value NULL in the qid/ssid/tid fields of the user table means that the value from the associated domain record should be used instead. This patch modifies the PL/pgSQL functions used by Dovecot and Postfix accordingly.

man: Add examples for userservices

Reverted previous three changesets (e09139525580, 85517c8fde36, 3acbff727626)

Modify userinfo output to indicate when domain defaults are displayed When Account instances reference NULL tid/qid/ssid, the data must come from the associated domain, and this should be indicated. For transport and services, this is easy to do as the string passed in the info dict can simply be modified. For quotalimit, however, another method must be used due to the CLI-side formatting. All approaches use a common formatter outsourced to the common.py file.

Modify Account class to handle NULL references This patch modifies the Account class to defer to using the associated domain's tid/ssid/qid fields if the per-instance fields are None/NULL.

Modify SQL update script to accept NULL fields for users The value NULL in the qid/ssid/tid fields of the user table means that the value from the associated domain record should be used instead. This patch modifies the PL/pgSQL functions used by Dovecot and Postfix accordingly.

Add list* subcommands to TODO

add domain/account notes to TODO

Add my WIP items to the TODO list

Cache interpolated destination Instead of running the string modification function for destination interpolation twice, store the result once and use the stored result subsequently.

man: Use example.com in catchallinfo example output.

README: reStructuredText fix: ERROR/3 and WARNING/2.

Enable interpolation of alias destinations This patch modifies the virtual_alias_maps function so that the destination address is parsed for the place holders %n, %d and %=, which are replaced with the localpart, the domain or the full address with '@' replaced by '=' of the queried key. In combination with alias domains, this allows for domain-specific recipients. E.g. given example.org and its alias domain example.com, defining an alias postmaster@example.org → postmaster+%d@example.org will cause mail to postmaster@example.*com* to go to postmaster+example.*com*@example.org.

Add catch-all aliases to README/features

Add catchall subcommands to manpages

Include catch-all count in domaininfo output

Make PL/pgSQL function feed back identity for mailboxes/relocated when there are catchall destinations. Without catchall aliases, if no virtual_alias matches, the query can just return NULL and Postfix will later check mailboxes/relocated for the address to rewrite. However, since virtual aliases are handled long before mailboxes/relocated, a catchall alias would also catch mail to mailboxes and relocated addresses, which we do not want. The way to tell postfix to keep delivering is for the virtual alias map to return the search key itself (identity function). This patch changes the postfix_virtual_alias_maps Pl/pgSQL function to do exactly that, but only if there are catchall destinations defined for the domain in question — otherwise it returns NULL when no match is found.

Fix syntax errors

Modify virtual_alias_maps function to check mailboxes/relocated first

Make virtual_alias_maps function search catchall when no aliases are found

Add catchall info to domain_info output

Correct syntax errors

Fixup catchall_info to return iterator, not self

Correct syntax errors

Teach CLI about catchall* commands

Provide catchall_* methods to the Handler class

First version of a CatchallAlias class I based the CatchallAlias class in catchall.py heavily on the Alias class, but by copy, not deriving. The two are functionally related, but the implementations are too different because CatchallAliases have no localpart.

Add database table for catchall destinations Catch-all aliases are stored for a domain. Since there can be multiple destinations per domain, this is done using a 1:n relation on the gid. This commit extends the SQL schema definition and also provides appropriate upgrade DDL.

pgsql: Merged types and functions into the create_tables… files.

postfix: Removed VIEW-based comments from configuration files.

pgsql: Removed most of the VIEWs. There are only two views left for now: - postfix_gid - vmm_domain_info All other views have a faster replacement: plpgsql functions.

Ignore even more Debian-build-related files

Ignore more Debian-build-related files Debian produces log and debhelper files in debian/* that do not need to be known to mercurial. Hence, this patch adds the globs to .hgignore.

Add non-root TODO

VMM/config: Fixed a typo: s/DB_MUDULES/DB_MODULES/g.

Ignore build files related to Debian packaging During Debian packaging, the package is built in ./debian/vmm/, and quilt is used (./.pc/). Both directories contain only generated files and should hence be ignored from the perspective of Mercurial.

Instruct mercurial to ignore build directory Python's distutil builds the package into ./build/, which can and should be ignored with respect to the repository. This patch simply adds the appropriate line to .hgignore.

VMM/handler: Create domain directory with umask 0022.

Modify address check query to obtain well-defined result The way in which UNION does not yield the desired result, because (a) UNION merges results and (b) the result order is undefined. This patch changes the query to select the counts as columns and hence provides a well-defined order.

postfix: Uncommented the query in pgsql-smtpd_sender_login_maps.cf.

postfix: Renamed pgsql-transport.cf to pgsql-transport_maps.cf. Now it matches pgsql-${postfix-parameter}.cf.

po: Updated French translation (translated by Dimitri Duc).

po: Updated Finnish translation (translated by Jorma Karvonen).

VMM/cli/subcommands: Fixed a TypeError in user_info().

postfix: Use pgsql functions in query templates where possible.

pgsql: Updated required privileges for function dovecotpassword.

README: Added section `Installation Prerequisites'. Reworked some parts.

update_config: Write the old package install dir to a tmp file. upgrade: Remove old CamelCase files. Removed some unused variables.

VMM: The pyPgSQL compatibility commit.

VMM: The Python 2.4.x compatibility commit.

VMM/relocated: Don't accept something@localhost as destination.

man1: Updated configuration related descriptions/examples, accordingly to commit 4ff0fa3ba0fa.

Updated the author's e-mail address.

man: Updated the URL of the bug tracker.

man5: Updated the `usable password schemes' part.

man1: Copied new formatted/translated command output into the examples.

VMM/cli/subcommands: Small improvement of domaininfo's output. Pointed out that the displayed quota limit is valid per user.

man1: Removed obsolete commands and added new commands.

man/de/man1: Removed translated argument names.

VMM/cli/subcommands: Use the common output formatting in list_pwschemes().

VMM/handler: Added a comment for the Translation Project.

VMM/cli/subcommands: Subcommand argument placeholders are no longer translatable.

po: Updated German translation (translated by Mario Blättermann).

po: Updated address for message-id bug reports. Looks like as sf.net would have horrible configured mail servers. :-/

merged changes from default(c0e1fb1b0145)