Dovecot v2.3.14 released
Hi!
We are pleased to release v2.3.14 of Dovecot.
IMPORTANT NOTE:
We have removed some components from the software, please review changelogs carefully prior upgrading.
Please find source tarballs at https://dovecot.org/releases/2.3/dovecot-2.3.14.tar.gz https://dovecot.org/releases/2.3/dovecot-2.3.14.tar.gz.sig
Binary packages in https://repo.dovecot.org/ Docker images in https://hub.docker.com/r/dovecot/dovecot
Aki Tuomi Open-Xchange oy
- Added new aliases for some variables. Usage of the old ones is possible, but discouraged. (These were partially added already to v2.3.13.) See https://doc.dovecot.org/configuration_manual/config_file/config_variables/ for more information.
- Optimize imap/pop3/submission/managesieve proxies to use less CPU at the cost of extra memory usage.
- Remove autocreate, expire, snarf and mail-filter plugins.
- Remove cydir storage driver.
- Remove XZ/LZMA write support. Read support will be removed in future release.
- doveadm -D: Add timestamps to debug output even when LOG_STDERR_TIMESTAMP environment variable is not set. Timestamp format is taken from log_timestamp setting.
- If BROKENCHAR or listescape plugin is used, the escaped folder names may be slightly different from before in some situations. This is unlikely to cause issues, although caching clients may redownload the folders.
- imapc: It now enables BROKENCHAR=~ by default to escape remote folder names if necessary. This also means that if there are any '~' characters in the remote folder names, they will be visible as "~7e".
- imapc: When using local index files folder names were escaped on filesystem a bit differently. This affects only if there are folder names that actually require escaping, which isn't so common. The old style folders will be automatically deleted from filesystem.
- stats: Update exported metrics to be compliant with OpenMetrics standard.
- doveadm: Add an optional '-p' parameter to metadata list command. If enabled, "/private", and "/shared" metadata prefixes will be prepended to the keys in the list output.
- doveconf: Support environment variables in config files. See https://doc.dovecot.org/configuration_manual/config_file/config_file_syntax/... for more details.
- indexer-worker: Change indexer to disconnect from indexer-worker after each request. This allows service indexer-worker's service_count & idle_kill settings to work. These can be used to restart indexer-worker processes once in a while to reduce their memory usage.
- auth: "nodelay" with various authentication mechanisms such as apop and digest-md5 crashed AUTH process if authentication failed.
- auth: Auth lua script generating an error triggered an assertion failure: Panic: file db-lua.c: line 630 (auth_lua_call_password_verify): assertion failed: (lua_gettop(script->L) == 0).
- configure: Fix libunwind detection to work on other than x86_64 systems.
- doveadm-server: Process could crash if logging was done outside command handling. For example http-client could have done debug logging afterwards, resulting in either segfault or Panic: file http-client.c: line 642 (http_client_context_close): assertion failed: (cctx->clients_list == NULL).
- dsync: Folder name escaping with BROKENCHAR didn't work completely correctly. This especially caused problems with dsync-migrations using imapc where some of the remote folder names may not have been accessible.
- dsync: doveadm sync + imapc doesn't always sync all mails when doing an incremental sync (-1), which could lead to mail loss when it's used for migration. This happens only when GUIDs aren't used (i.e. imapc without imapc_features=guid-forced).
- fts-tika: When tika server returns error, some mails cause Panic: file message-parser.c: line 802 (message_parser_deinit_from_parts): assertion failed: (ctx->nested_parts_count == 0 || i_stream_have_bytes_left(ctx->input))
- lib-imap: imapc parsing illegal BODYSTRUCTUREs with NILs could have resulted in crashes. This exposed that Dovecot was wrongly accepting atoms in "nstring" handling. Changed the IMAP parsing to be more strict about this now.
- lib-index: If dovecot.index.cache has corrupted message size, fetching BODY/BODYSTRUCTURE may cause assert-crash: Panic: file index-mail.c: line 1140 (index_mail_parse_body_finish): assertion failed: (mail->data.parts != NULL).
- lib-index: Minor error handling and race condition fixes related to rotating dovecot.index.log. These didn't usually cause problems, unless the log files were rotated rapidly.
- lib-lua: Lua scripts using coroutines or lua libraries using coroutines (e.g., cqueues) panicked.
- Message PREVIEW handled whitespace wrong so first space would get eaten from between words.
- FTS and message PREVIEW (snippet) parsed HTML &entities case-sensitively.
- lib-mail: When max nested MIME parts were reached, IMAP BODYSTRUCTURE was written in a way that may have caused confusion for IMAP clients and also Dovecot itself when parsing it. The truncated part is now written out using application/octet-stream MIME type.
- lib-oauth2: HS512 and HS384 JWT token algorithms crash when you try to use them: Panic: file hmac.c: line 26 (hmac_init): assertion failed: (meth->context_size <= MAC_MAX_CONTEXT_SIZE).
- event filters: NOT keyword did not have the correct associativity. NOT a AND b were getting parsed as NOT (a AND b) instead of (NOT a) AND b.
- Ignore ECONNRESET when closing socket. This avoids logging useless errors on systems like FreeBSD.
- event filters: event filter syntax error may lead to Panic: file event-filter.c: line 137 (event_filter_parse): assertion failed: (state.output == NULL)
- lib: timeval_cmp_margin() was broken on 32-bit systems. This could potentially have caused HTTP timeouts to be handled incorrectly.
- log: instance_name wasn't used as syslog ident by the log process.
- master: After a service reached process_limit and client_limit, it could have taken up to 1 second to realize that more client connections became available. During this time client connections could have been unnecessarily rejected and a warning logged: Warning: service(...): process_limit (...) reached, client connections are being dropped
- stats: Crash would occur when generating openmetrics data for metrics using aggregating functions.
- stats: Event filters comparing against empty strings crash the stats process.
- Aki Tuomi, 04.03.21 11:21
Hi!
We are pleased to release v2.3.14 of Dovecot. [...]
Hi,
Just a minor thing I noticed by chance: the Wiki documentation that is included in the source tarball is rather outdated. The timestamp of the files in dovecot-2.3.14/doc/wiki is 2019-06-19, which would be somewhere between 2.3.6 (2019-04-30) and 2.3.7 (2019-07-12), according to dovecot-2.3.14/NEWS.
I suggest either refreshing it with the current content, or simply replacing it with a small textfile pointing to wiki.dovecot.org and/or doc.dovecot.org. While 90% of it is probably still valid, there has been quite a bunch of tweaks, fixes and feature additions and drops that might lead to WTF moments and a bit of head-scratching, if one follows these offline docs rather than their corresponding online version.
Bye, Thomas
On 08/03/2021 12:20 Bjoern Franke bjo@schafweide.org wrote:
Hi,
- Remove XZ/LZMA write support. Read support will be removed in future release.
Will XZ/ZMA Read support be removed in the next release or just in some future release?
Kind Regards Bjoern
It will be removed in some future release. Not next release.
Aki
- Bjoern Franke bjo@schafweide.org:
Hi,
- Remove XZ/LZMA write support. Read support will be removed in future release.
Will XZ/ZMA Read support be removed in the next release or just in some future release?
I'm a bit behind the curve: Why is it being removed at all?
- reduce code size
- better compression algorithms available
- other
-- Ralf Hildebrandt Geschäftsbereich IT | Abteilung Netzwerk Charité - Universitätsmedizin Berlin Campus Benjamin Franklin Hindenburgdamm 30 | D-12203 Berlin Tel. +49 30 450 570 155 | Fax: +49 30 450 570 962 ralf.hildebrandt@charite.de | https://www.charite.de
On 09/03/2021 08:20 Ralf Hildebrandt ralf.hildebrandt@charite.de wrote:
- Bjoern Franke bjo@schafweide.org:
Hi,
- Remove XZ/LZMA write support. Read support will be removed in future release.
Will XZ/ZMA Read support be removed in the next release or just in some future release?
I'm a bit behind the curve: Why is it being removed at all?
- reduce code size
- better compression algorithms available
- other
It has memory safety issues and is very slow compared to the other alternatives.
Aki
participants (5)
-
Aki Tuomi
-
Aki Tuomi
-
Bjoern Franke
-
Ralf Hildebrandt
-
Thomas Zajic