+ baseconn: established a rule for how should Open() end: OpenFinished/ReadError/exception
+ atconn: notifying listeners about errors
+ tcpconn, serialport, serialadatpter: exception if Open() is called on connection being already opened and not closed

reksio authored 9 years ago and Aleksy Barcz committed 9 years ago

If the response contains invalid transaction id just ignore it, instead
of progressing with queries. Ignoring such request makes more sense
because usually that happens when a device is lagging behind with
its responses, so it makes sense to wait.

This also fixes a bug where response with wrong transaction id was leaving
deadliner timer active, which could lead to a crash.

Also - don't do anything on deadline timer tick if the unit is
in idle state.

One day we will have a unit test for this code...

Mateusz Morusiewicz authored 9 years ago and Aleksy Barcz committed 9 years ago

This change needs more testing on various architectures.

Tomasz Pieczerak authored 9 years ago and Aleksy Barcz committed 9 years ago

* files moved to another repository.

Now printing makes one vertical axis for params of the same min/max and unit. The lowest precision is considered.

Few pointers were not deleted properly. wxLog initialized on its own to avoid leak.

clean up input buffer on connection (re-)establishment to make
sure that no stale data lingers in the buffer

all started daemons are stopped.

+ If sender query is NO_DATA (one or more send is NO_DATA) wait
for it to be filled with data before write.

* Introduce true sender access in BaseDaemon.
* Needed changes in S7Dmn.
* S7Dmn: Do not send NO_DATA to device (maybe optional?).

Few pointers were not deleted properly. wxLog initialized on its own to avoid leak.

there is new heartbeat added which role is to communicate
meaner4 health

the meaning of the existing one is changed, its is now following
timestamp of parcook messages

+ if file doesn't exist, don't break the uninstallation

receive correctly formatted response that still gets
ignored (because of e.g. having wrong transaction id
or unit id)

Updating szarp will now trigger a config reload.
Also, iks-server should be restarted properly by szarp-iks now.

* lost+found/ exclusion is needed for configurations where there are
  any locations mounted in /opt/szarp/<prefix> directory (like szbase/
  located on a separate partition). This does not cause 'szrsync' to
  fail, but generates a misleading warning message.

we throw exception