GitLab

cps_recv: Work on issue #207 dealing with wildly bad input time values.

See merge request cds/advligorts!202

Add a filter to the receiver and drop all messages that have a key (time) that is too far from current gps time.
 * defaults to 5 minutes.
 * reads the current gps time from /proc/gps or calculates from system time.

Improve systemd kernel module handling

See merge request cds/advligorts!200

rtcds: added end-of-word check to status scan of lsmod for model names.  Fix had previously been applied to other system models

python3-rtcds: fe_generator: improved binding between front end services, such as making iop module dependent on dolphin nodemgr, binding the dolphin services, and binding each user module to the iop module.  Improves greatly the correctness of the handling of the kernel modules.

inadvertently commented out; reinstated that code.

Edc timing

See merge request cds/advligorts!196

 * When using the gps clock calculate the cycles from the nanoseconds, so that cycle_complete() can be dropped.
 * Increase the sleep time from 2us to 500us.
 * We no longer need to adjust the seconds when we have a negative wait.

This is done on the constructor, not sure why it is needed, what is overwriting it with a 0.

 * Replaced a numeric value with a symbolic constant
 * Removed some commented out code

Use the cycle number to derive the nano-seconds.  The models don't seem to be populating the nanosecond field of the rmIpcStr block structure.

This should allow it to be properly synced with an external IOP or with a gpstime device.

Other minor updates as well to cleanup initialization, remove shadowed variables, ...

Local_dc work on issue #201.

Closes #201

See merge request cds/advligorts!198

 * Added a check that each non-iop model must have the same cycle and second as the iop model, not just the same cycle.
   * When checking for data from the iop the system checks for progression, so we know we are seeing a live iop.
 * Minor cleanups (removing an unused variable, replacing a numeric constant with a symbolic constant)
 * Added a test to check that 'old' dcus are not sent on.

Closes #201

rtcds: change default model build to kernel space instead of user space.

Closes #202

See merge request cds/advligorts!197

The default build can still be set back to user space by setting
USE_KERNEL_MODELS=false
in the environment.

GSC18ai32 module, so fixed as well.

Time to test code still works with standard timing card and backplane.

Rtcds check model

Closes #138

See merge request cds/advligorts!195

rtcds: read in environment before setting flags and change flag tests to explictly look for non-empty strings.

The entire block of code for initializing environment variables has been moved before the command line flag checks.
The USERLAND_FLAG and NO_KERNELSPACE_FLAG default values are determined by the environment, but the /etc/advligorts files need to be read in first.

Command line flags are cleared if empty string, or set if not empty.  This wasn't obvious originally since the test for flags was
just [[ $SOME_FLAG ]].

The tests have been changed to [[ -n $SOME_FLAG ]] or in the negative case [[ -z $SOME_FLAG ]].
This doesn't change the behavior but is less ambigous to the eye.

This code is functional, but requires cleanup and compatability changes
to still support present timing receiver and backplane.

NDS - replacing a gds based reference counting system with a smart pointer.

See merge request cds/advligorts!194

mmap_ptrs aren't copied in the code, so make it non-copyable, change from shared_ptr to unique_ptr.

brought in the daqd raii.hh helpers to give a make_unique that is available at any C++ revision that supports unique_ptr.

Simplified the get and set methods.

Removed the size method as it isn't used.