Upgrade urgency HIGH: Critical fix to Redis Sentinel, due to 3.2.0
regression compared to 3.0.
Hey, this is Redis 3.2.1, and this release should bring some grain of
maturity to Redis 3.2. The list of commits following this note will tell
you the details, but the main things addressed in this release are the
following:
1. A critical bug in Sentinel was hopefully fixed. During the big 3.2
refactoring of Redis Sentinel, in order to implement connection sharing
to make Sentinel able to scale better (few Sentinels to monitor many
masters), a bug was introduced that mis-counted the number of pending
commands in the Redis link. This in turn resulted into an inability to
talk with certain Redis instances. A common result of this bug was the
inability of Redis Sentinel to reconfigure back the old master, after
a failover, when it is reachable again, as the slave of the new master.
This was due to the inability to talk with the old master at all.
2. BITFIELD bugs fixed.
3. GEO commands fixes on syntax errors and edge cases.
4. RESTORE now accepts dumps generated by older Redis versions.
5. Jemalloc now is really configured to save you memory, for a problem a
change in the jemalloc configuration did not really survived when the
3.2.0 release was finalized.
6. TTL and TYPE command no longer alter the last access time of a key, for
LRU evictions purposes. A new TOUCH command was introduced *just* to
update the access time of a key.
7. A bug was fixed in redis-cli, that connected to the instance running on
the port 6379 if there was one, regardless of what was specified.
8. TCP keep alive is now enabled by default. This should fix most ghost
connections problems without resulting in any practical change in
otherwise sane deployments.
9. A Sentinel crash that could happen during failovers was fixed.
Here is the announcement for the stable release after 2 years, which
many features and bug fixes contains.
Summary of the changes since last stable release is:
- - Support the size specific design selection in OS/2 table version 5
- - Allow the modification on 'lang' and 'charset' objects
- - Increase the refcount in FcConfigSetCurrent()
- - some updates in orth files
- - Add --error-on-no-fonts option to fc-cache
- - Use lang=und instead of lang=xx for "undetermined"
- - Add FC_WEIGHT_DEMILIGHT and change from 65 to 55
- - Add FC_COLOR
- - Treat color fonts as scalable
- - no FC_LANG added with FcConfigSubstitute() when it has "und"
- - Hardcode blanks in library
- - Support symbol fonts
- - Unicode 8.0 support
- - Add hintstyle templates and default hintslight
- - GX font support
- - Improve the footprint issue on updating caches
- - Bump the cache version to 6
- - more bug fixes
- - and more...
previous handful of packages like -datatypes or -dblink. They
contain all of the contrib subtree, matching what similar packages
carry in e.g. FreeBSD ports or what PostgreSQL users generally
expect.
This subtree contains porting tools, analysis utilities, and
plug-in features that are not part of the core PostgreSQL system,
mainly because they address a limited audience or are too
experimental to be part of the main source tree. This does not
preclude their usefulness.
[v3.0][] - 2016-03-03
---------------------
This release signifies a major change in libConfuse. On out-of-memory
conditions at run time, invalid API input, and some other odd use-cases,
libConfuse will no longer `assert()`. Instead, `NULL` or `CFG_FAIL` is
returned with an error code for you to handle. For some users this will
completely change how your application works, so heads up! The library
ABI version has also been stepped due to this.
Special thanks in this release goes out to Frank Hunleth, Peter Rosin
and David Grayson for their tireless efforts in helping improve this
library!
**Note:** libConfuse no longer calls `setlocale()` for `LC_MESSAGES` and
`LC_CTYPE`. See the documentation for `cfg_init()` for details.
### Changes
* Support for handling unknown options. The idea is to provide future
proofing of configuration files, i.e. if a new parameter is added, the
new config file will not fail if loaded in an older version of your
program. See the `CFGF_IGNORE_UNKNOWN` flag in the documenation for
more information. Idea and implementation by Frank Hunleth.
* Add public API for removing sections at runtime, by Peter Rosin.
* Allow `cfg_opt_getval()` on options that are `CFGF_MULTI` sections,
by Peter Rosin.
* Add `cfg_setmulti()` and `cfg_opt_setmulti()`, by Peter Rosin.
* Add CLI example of how to manage configuration changes at runtime,
also by Peter Rosin.
* Support for Travis-CI and Coverity Scan, by Joachim Nilsson.
* Use `autoreconf` in `autogen.sh` instead of calling tools separately.
* Powershell script for AppVeyor CI to build libConfuse with MSYS2
by David Grayson.
* Removed calls to `setlocale()` intended to localize messages, with
`LC_MESSAGES`, and region specific types, with `LC_CTYPE`. This is
now the responsibility of the user of the library.
* Reindent to Linux coding style for a clear and well defined look,
this to ease future maintenance. Issue #33
* Add support for `CFGF_DEPRECATED` and `CFGF_DROP` option flags. The
former causes libConfuse to print a deprecated warning message and the
latter drops the read value on input. Idea and implementation by
Sebastian Geiger. Issue #24
* Add `HACKING.md` document to detail maintenance and release checklists
### Fixes
* Do not assert on API input validation, memory allocation, or similar.
Instead, return error code to user for further handling. This change
also includes fixes for a lot of unchecked API return values, e.g.,
`strdup()`. Issue #37
* Protect callers arguments to `cfg_setopt()`, by Peter Rosin
* If new value to `cfg_setopt()` fails parsing, do not lose old value,
by Peter Rosin.
* Fixes to update support for older versions of Microsoft Visual Studio
as well as MSYS2/mingw-w64 by Peter Rosin and David Grayson.
* Issue #45: `cfg_init()` does not report error on multiple options with
the same name. Fixed by Peter Rosin.
* Fixes for memory leaks, invalid expressions, unused variables and
missing error handling, all thanks to Coverity Scan
Version 0.11.2
==============
*released on 15 June 2016*
- Fix typo that would break tests.
Version 0.11.1
==============
*released on 15 June 2016*
- Fix a bug in collection validation.
- Fix a cosmetic bug in debug output.
- Various documentation improvements.
Release 8.39 14-June-2016
-------------------------
Some appropriate PCRE2 JIT improvements have been retro-fitted to PCRE1. Apart
from that, this is another bug-fix release. Note that this library (now called
PCRE1) is now being maintained for bug fixes only. New projects are advised to
use the new PCRE2 libraries.
2016-06-19 meld 3.16.1
======================
Fixes:
* Fix regression in support for file comparisons from pipes (Kai Willadsen)
* Add open build service directories to the version control filter
(Dominique Leuenberger)
* Add darcs to list of loaded plugins (pbgc)
* Fix the dark style (Kai Willadsen)
* Windows build fixes (Kai Willadsen)
Noteworthy changes in version 1.23 (2016-06-15) [C19/A19/R0)
-----------------------------------------------
* Fixes an assertion failure due to es_flush on read/write streams.
* Fixes a bug with a too short memory limit is es_fopenmen.
* Cross-build support for powerpc-unknown-linux-gnuspe and
tilegx-unknown-linux-gnu architectures.
* Interface changes relative to the 1.22 release:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
GPG_ERR_SUBKEYS_EXP_OR_REV NEW.
New Features
Amazon metadata download: Download the series information when available from amazon.com
Amazon metadata download: Recognize common patterns of embedding the series name after the title and remove it automatically
Edit Book: Add support for EPUB 3 to the Table of Contents tool. Now when editing EPUB 3 files using the tool will automatically generate both the EPUB 3 navigation document and the NCX table of contents for backwards compatibility.
EPUB Input: Implement reading of Table of Contents from EPUB 3 files that do not specify a fallback NCX ToC
EPUB metadata: When setting a cover image for an EPUB file that has no metadata cover image defined, add the new cover image as a pure metadata cover instead of aborting.
Bug Fixes
Fix crash on exit if any plugins have loaded code that raises exceptions during shutdown.
Closes tickets: 1592414
Copy to library: Speed up checking for duplicates when copying multiple books to a large library.
Closes tickets: 1593027
Get Books: Update Smashwords plugin for website changes.
Closes tickets: 1590653
Edit Book: Fix a bug in the new text search facility that could cause searches to fail when searching from the current cursor position instead of the top of the file.
Closes tickets: 1588778
Bulk conversion dialog: Fix page margin settings being ignored when converting to DOCX
Make the default shortcut for remove books Backspace on OS X since many mac keyboards have no delete key and the backspace key is labelled as delete
Edit Book: Check Book: Do not warn if a cover image is not referenced in the text
E-book viewer: Fix a bug that could cause incorrect rendering of chapter titles in some rare circumstances.
Windows console output: Handle consoles with non black and white color scheme correctly
Fix a regression in 2.57 that caused auto-removal of borders from covers to become much less proficient at detecting borders
Fix a regression in 2.57 that broke grayscaling of images
Improved news sources
Technology Review (DE)
--------------
Version 2.15b:
--------------
- Added a README section on looking for non-crashing bugs.
- Added license data to several boring files. Contributed by
Jonathan Metzman.
Change Log for New E (NE)
-------------------------
Version 3.02 15-May-12
----------------------
1. The icurrent command did not remove the line number from the line it copied.
2. Cutting and pasting did not always remove line numbers, as it is documented
to do.
3. Some additional information has been added to the output from "show
settings".
Version 3.01 02-Mar-12
----------------------
1. There were bugs in the table the controls what commands may be obeyed in
read-only mode. For example, the "readonly" command was not allowed, so the
mode could not be turned off. Doh! In addition, the single-character
commands such as # ? etc. were not correctly handled in readonly mode;
neither were bracketed groups nor procedures. Using any of these commands
caused segfaults.
Version 3.00 23-Feb-11
----------------------
1. Rather than just fixing an out-by-one bug in eversion.c, I re-designed how
it works. Added the copyright notice to the -help output. Added a lot more
casts for signed/unsigned chars so that it compiles clean with -Wall on today's
compilers (it's been over 5 years since the last release).
2. Renamed configure.in as configure.ac in line with modern practice, and did
some tidying/updating of the configure and make files.
3. Made --help a synonym of -help, and added -v and --version as synonyms of
(the very ancient) -id.
4. Removed the -stream facility, as it seems redundant nowadays.
5. Fixed reference to unset memory in rdargs.c.
6. Removed redundant code for various special (no longer existing) terminal
types, and unused code for the Acorn windowing environment. The only special
terminal type now recognized is xterm. Also removed the unixregexp command,
which used to switch between different regex types, but has been obsolete for
ages.
7. In screen mode, test whether an xterm is set to UTF-8, and adjust output
accordingly. UTF-8 sequences are sent for certain characters whose code points
are greater than 127. Characters that do not occupy exactly one cell on an
xterm display are displayed as the substitute character (defaulting to a
question mark). These include zero-width and double-width characters.
Characters with code points greater than 0xFFFF are also substituted.
8. If -w[idechars] is specified, or the "widechars on" command is obeyed,
recognize valid wide UTF-8 characters in the text being edited, without
modifying the text bytes. A top-bit-set byte that is not part of a valid UTF-8
character is treated as a single-byte character. This affects how data lines
are displayed on the screen and in verification output. You can flip between
modes during an editing session.
9. "Show wordcount" was including line-ending characters in its character
count, contrary to its documentation. Of course, previously it was just a byte
count. It now shows both a byte count and a character count in wide character
mode, in both cases excluding line ending characters. Words are now delimited
by either tabs or spaces (previously it was just spaces). The output is now on
multiple lines.
10. Added new keyboard escapes for inputting Unicode characters, for example,
ESC A ' for aacute. Also recognize ESC U or ESC u followed by up to 5 hex
digits (short-terminated by a non-hex digit or another ESC). This feature
applies only to input in screen mode because in line-by-line mode the commands
are read using standard C input functions.
11. While implementing 8/9/10, several old bugs/infelicities were found and
fixed:
(a) In the code for formatting paragraphs, if the begin/end paragraph
settings did not include indented lines, and some lines of a paragraph
were indented less than earlier lines, characters could sometimes get
scrambled.
(b) Characters deleted by keystroke while in overstrike mode did not get
remembered for recall by "undelete".
(c) Some redundant arguments for certain functions were discovered and
removed (they probably related to code for environments no longer
supported which itself has gone).
(d) If a command line from a .nerc file was recovered by the use of the
"up" key in a screen command line, the final newline was still present,
and displayed as "?".
12. When running in an xterm, NE now recognizes a left-button mouse click in
the window, and moves the cursor appropriately. It also recognizes twiddling
the wheel in a wheel mouse, and for each twiddle scrolls the screen up or down
by an amount that can be set by "set autovmousescroll" (default 1). The current
line is changed only if it would disappear off the screen. The "mouse" command
can be used to turn mouse recognition on and off. Unfortunately, you have to
turn it off if you want to use normal xterm cut-and-paste.
13. NE was maintaining a value in a variable called main_currentlinenumber, but
never actually using it for anything. This must be an historic relic. Removing
it does not seem to have broken anything.
14. When writing a file, NE was not testing for an error on the fclose() call
at the end. Errors can happen here if, for example, the system is buffering the
output and a quota overrun is not detected until fclose().
15. When "show keys" was issued in screen mode, NE was pausing for interaction
more often than actually necessary if the screen or window was relatively deep.
It now notices the available depth.
16. The subchar command now allows for altering the character that is
substituted on screen for undisplayable characters.
Version 2.01 06-Jun-05
----------------------
1. Lines whose length was between 32768 and 65535 were ending up with
negative lengths because I was using short int rather than unsigned short
int. They showed up therefore as empty lines, which was pretty disastrous. As
well as inserting "unsigned", I have removed "short", since memory is plentiful
these days. So line lengths can in principle be *really* big.
2. Change from -lcurses to -lncurses in the Makefile.
3. Insert $(DESTDIR) before all the path names for make install.
Version 2.00 01-Mar-04
----------------------
This version has had the code refactored in several ways, and the compiling
process has been converted into a conventional "configure, make, make install"
paradigm. Only PCRE regular expressions are supported, and the PCRE library has
to be installed. Only Unix-like systems are now supported. The initialization
now uses .nerc instead of an environment variable. A number of other tidies and
minor interface changes were made. The documentation was re-worked to remove
all the obsolete stuff and to incorporate the changes.
