v0.9.2

This is a bugfix and maintenance release.

Windows:

• add files in lcm/windows that were missing from 0.9.1

v0.9.1

This is a bugfix and maintenance release.

C:

• update comment docs to describe start_timestamp option

C++:

• change ReceiveBuffer forward declaration to struct. Fix MSVC error.

C# / .NET:

• lcm-dotnet: fix publishing fragmented messages
• modify one of the c# examples so that it can be easily used to test
  fragmented messages (by increasing num_ranges)

Java:

• install jar file as share/lcm.jar, not share/lcm-x.y.z.jar with a
  share/lcm.jar as a symlink.
• fix the export snippet to work with two bookmarks that were created out of
  order. also make it default to selecting the beginning/end to bookmark if
  only 1 bookmark exists
• expose flush() method in lcm.logging.Log

lcm-gen:
Python:

• python encode checks fingerprints of nested types (re #47)
• apply patch from tprk77 bugfix in init.py import generation. (re #50)
• don't raise UnicodeError when decoding non-utf8 strings.
  Java:
  • apply patch from jamoozy to reduce eclipse compiler warnings. (re #49)
    C++:
  • remove --cpp-cpath option

Windows:

• in WinPorting.h, #define strtoll _strtoi64 (re #46)
• add LCM C++ files to VS project file
• remove redundant WinSpecific/include/lcm dir.
• move WinSpecific/Win*.{cpp,h} -> lcm/windows/
• adjust vcproj files accordingly
• update WinSpecific/README.txt and remove dead directory

lcm-logger:

• add --flush-interval option to periodically flush log file to disk.
• close and re-open logfile when SIGHUP is received. Should make it easier
  to use lcm-logger with logrotate.

liblcm-test:

• lcm-source win32 bugfix. sleep convert to ms correctly.

v0.9.0

This release adds a constructor to the C++ API, some command line options to
lcm-logger, and incorporates a number of bugfixes.

Java:

• workaround to accomodate recent changes in Automake. In particular, fixes
  lcm-java for Ubuntu 12.04.
  lcm-c:
• add start_timestamp option to log file provider
• fix multicast TTL issue #45
• Fixes for fragmentation of larger messages. Big messages were previously
  fragmented into 1400B packets. However, since lcm_udp uses a short for the
  number for fragments there can only be 64K fragments. The makes LCM use
  the 1400B packet size if the message would be broken up in to less than 64K
  fragments. Otherwise it increases the packet size to split the message into
  64K packets
  lcm-c++:
• add a constructor to create a C++ instance wrapping arond an existing lcm_t
  C instance.
  Python:
• fixes to build in Windows using MS VS 2008.
  C#:
• bugfixes when generating default namespace.
  lcm-logger:
• add --auto-split-mb and --auto-split-hours options.

v0.8.0

This release changes the C++ API, adds some features to lcm-logplayer-gui, and
fixes a few bugs.

lcm-c:

• add new macros LCM_MAJOR_VERSION, LCM_MINOR_VERSION, LCM_MICRO_VERSION to
  check LCM version from C/C++.
  lcm-c++:
• rename LCM::fileno() to LCM::getFileno()
  fileno is implemented as a macro on some platforms, and is thus effectively
  a reserved keyword.
  lcm-gen:
  C++:
  • don't try to dynamically resize fixed-size arrays (re issue #42)
    Python:
  • bugfix in decoding multivariable arrays
    lcm-logplayer-gui:
    add command line parameters to:
  • start logplayer with the log paused
    • filter out specific channels from being played back by default.

v0.7.1

This is a bugfix and maintenance release.

lcm-gen:
C++:
- fix decoding of zero-length primitive arrays (re issue #38)
- computeHash() omit parameter name if unused
- don't #include a type's own header file for recursive types
- fix memory leak in LCM::publish() (re issue #35)
- inline methods for generated code
- use static local variable for getHash() to reduce liblcm dependency of
generated C++ code
- #include if a type has a string field
C:
- don't #include a type's own header file for recursive types
Java:
- add --with-jardir=DIR option to configure script to allow configuring
lcm.jar destination directory.
General:
- update lcm-gen manpage to document --c-no-pubsub

lcm-java:

• ClassDiscoverer doesn't try to traverse unreadable files/directories

v0.7.0

This release introduces support for native C++ bindings, and includes a number
of bugfixes and documentation updates.

lcm-gen:

• Added the -x / --cpp option, which generates native C++ language bindings.
  See examples/cpp for examples, and the documentation for a tutorial and
  API reference.

lcm-c:

• fix a minor memory leak in lcm_tcpq.c (closes issue #33)

General:

• Switched documentation from GTK-Doc to Doxygen. To build most documentation,
  run doxygen from the docs/ directory. Java, C#, and Python API reference
  docs are still built separately.

v0.6.0

This release includes a number of bugfixes and performance enhancements

lcm-c:

• Change internal buffering from a fixed size ringbuffer to
  per-subscription FIFO queues. Subscriptions start with a default maximum
  queue length of 30 messages, which can be adjusted using the new function
  <message_type>_subscription_set_queue_capacity()
  For example, to set the queue size to 5 for a message type example_t:
  example_t_subscription_t* subs =
  example_t_subscribe(lcm, "EXAMPLE", message_handler);
  example_t_subscription_set_queue_capacity(subs, 5);
• Explicitly disallow recursive calls to lcm_handle.
• fix synchronization issues when allocating receive resources and
  conducting multicast self-test. see issue #30

lcm-logplayer-gui:

• expose remote log player play and pause commands

lcm-logger:

• bugfix when creating subdirectories for output files

lcm-spy:

• bugfix when LCM URLs contain equal signs
• searches directories on classpath for lcmtype .class files, in
  addition to .jar files.

lcm-gen:
Python
- bugfix. remove extraneous typename
- Constructor initializes valid fixed-size arrays
- python initialize float, double to 0.0 instead of 0

Java
- emit_java now appends 'f' for const float

General:
- flush stdout before failing on parse / semantic error

General:
- add example for using LCM with GLib event loop
- add example program for using LCM with select()
- minor updates to API documentation

v0.5.2

This release includes a number of bugfixes and performance enhancements

lcm-logger:
exit when disk full
use GLib regexes if they're available.
Add --invert-channels / -v flag

lcm-java:
Make LCMTypeDatabase public
LCM fragmentation: guard against receiving same fragment twice
lcm.lcm.LCM: allow passing NULL or the empty string to the constructor to
explicitly specify default LCM provider.
lcm.lcm.LogFileProvider: bugfix for write mode
remove lcm.test.SlowLCMSubscriber

lcm-gen:
general
warn if struct has 'int' member type instead of 'intN_t'
add options --csharp-strip-dirs and --csharp-root-nsp
Python - generate valid code on empty LCM type
only emit new_parents in _get_hash_recursive if necessary
fix init.py imports.
Check for "from import " instead of "import "
lock init.py when writing
C#
miscellaneous bugfixes
Java
trust explicit array length field, not .length attribute.
fix copy which copied in the wrong direction
faster arraycopy in several instances.

miscellaneous:
add lcm-lite implementation
modify examples -- place LCM types in package "exlcm"
add tutorial-dotnet.sgml to Makefile

Windows & .NET:
LCM.NET - fix unsubscribe bug
portability fixes

v0.5.1

This is a bugfix release, affecting only the Windows port.

Windows:
Fix lcmgen Visual Studio project file - missing emit_csharp.c
Add a .NET tutorial and some some C# examples

v0.5.1

This release introduces experimental support for C# / .NET, provided by
Jan Hrbáček. Bug reports and patches are greatly appreciated.

This release also introduces a small python API change, and renames a program
in windows.

Additional bugfixes are included.

Windows:

• Now builds lcm-gen.exe, not lcmgen.exe

.NET:

• The lcm-dotnet/ directory contains a publish-subscribe library for
  .NET applications.

lcm-gen:

• (C#) A new set of options are available for generating C# bindings
• (Python) change init.py import statements. If a package named 'proj'
  contains the LCM type msg_t, then the init.py file now contains:
  from msg_t import msg_t
  and the recommended usage pattern is:
  import proj
  instance = proj.msg_t()
  see googlecode issue #21
• (Windows, Java) fix pathname bug

Other:

• minor build system bugfixes