do not use a device register as a trigger any more

allow using the TestFacility without the testable mode, which can be useful for some tests of the framework itself

According to ChimeraTK/DeviceAccess#61 Devices will no longer close on their own (e.g. when throwing an exception). The exception handling now takes into account that the device always stays open

TriggerFanOut must catch exceptions when reading from the TransferGroup, since this cannot be handled by the decorator

Martin Killenberg authored 5 years ago and Martin Christoph Hierholzer committed 5 years ago

added more exception test code which should work but is but. Commented because it completely blocks the test.

Martin Killenberg authored 5 years ago and Martin Christoph Hierholzer committed 5 years ago

Creation of fanOuts is based on externalTriggerImpl (=trigger consumer), not externalTriggerNode (=trigger feeder).
- Fixes connection network but not exception handling.
- Lots of debug output left in the code for further investigation.

Martin Killenberg authored 5 years ago and Martin Christoph Hierholzer committed 5 years ago

extended test with a second device. check that it is still working if the first one is throwing is not working with the old impl.

Martin Killenberg authored 5 years ago and Martin Christoph Hierholzer committed 5 years ago

  no change to the test sequence yet

fixed a potential invalid read/write error, access to object and its potential destruction was done in wrong order

Supports specifying modules in xml config.

the swicth because order is not important for these.

ModuleList serves as a container for VariableGroups. The VariableGroups
mirror xml modules in the config file.

Xml parsing logic factored out to class ConfigParser.

 - Have a ZMQ push-type variable from a DOOCS-backed device that is NOT
   used as a trigger (otherwise the bug does not happen)
 - Connect that to an ScalarPushInput in an ApplicationModule
 - call readAnyGroup() from the ApplicationModule's main method

=> ReadAnyGroup.finalise() will cause an infinite recursion between
ExceptionHandlingDecorator::doReadTransferAsync and
TransferElement::readAsync because of wrong dispatching to the decorated
type

- Make doxygen pick up markdown.
- introduced: set_default_build_to_release.cmake

set_default_build_type_to_release.cmake -> set_default_build_to_release.cmake

- add test for full exception handling

  DeviceModule and ControlSystemModule now do not take the optional prefix argument in their constructors any more. Use the [] operator instead to get the submodule.

- Add ExceptionHandlingDecorator and decorate all device variables with it.