State of the 0.7.1 release

[Hynek Hanke]

Hello all,

Speech Dispatcher 0.7.1:

This is currently being developed in Git branch speech-dispatcher-0-7 .

This branch contains bug fixes as well as major improvements that go
further in the field of session integration. We have deprecated the
SPEECHD_PORT environment variable and instead established
a SPEECHD_ADDRESS variable which can handle both
communication methods including their parameters and is extensible
with a third DBUS interface in the future.

Autospawn was highly improved in that it first checks whether
auto-starting the server will really lead to establishment of the
desired communication channel. As such, autospawn will not be done
if the client tries to connect to a remote server or if the communication
channel he choses is not the one which Speech Dispatcher is configured
for. Autospawn will also relay any error messages reported by
the server back to the client, so in case something fails, the user
can directly see it and doesn't have to study server logs.

Numerous bug fixes were committed. We did a study of
memory leaks in the server and in the espeak module.
We did so by running repeated tests including automatic
tests as well as real Orca usage. The memory usage was
checked over time and when increase was found, this
was investigated with Valgrind Memcheck and the major
leaks fixed.

We also found increases in memory consumption which
are not memory leaks in their strict definition. Some information
related to the client sessions was kept after the client disconnects
and the also a table of file descriptors in use was growing
unnecessarily. Both problems were fixed.

We can now report that although minor memory leaks
still exist, particularly on startup and in error cases, the
memory usage over a 24 hour testing period seems to be relatively
stable (<0.1MB growth) in server and in the espeak module.
There are some problems inherent in this kind of testing,
such as swapping and various reorganizations of the memory
performed by the kernel at certain times. If somebody wants
to occupy himself with further investigations on memory leaks,
we will welcome such a contribution, but we think it can hardly
be considered as a real usage problem anymore.

We are also still considering inclusion of the SpeechTech TTS
module. It will however be disabled by default in 0.7.1.

What remains is to modify the spd-conf debug mode to reflect
on the new state of unix socket communication and user session
integration and to update the documentation.

Best regards & thanks for all your contributions,
Hynek Hanke