Closed mirenradia closed 5 months ago
Hi Miren - the changes made to initialize m_gramr_pt look good to me but the checkpoint restarts are failing the regression test. Do you want to split the pull request into two separate ones so that the pointer part can be merged while you work on the checkpoint restarts?
Hi Miren - the changes made to initialize m_gramr_pt look good to me but the checkpoint restarts are failing the regression test. Do you want to split the pull request into two separate ones so that the pointer part can be merged while you work on the checkpoint restarts?
I think the regression tests passed on the last commit. The point in the first commit was to show that they would currently fail but then the change in the final 2 commits fixes it.
Thanks - I thought there 3 independent commits and only the third one passed.
This fixes #52 by adding a getter for the
GRAMR
pointer stored by eachGRAMRLevel
object which sets the value if it's null. The pointer has now been madeprivate
so that derived classes use the getter rather than using the pointer directly.I have also added a step to the regression test GH action workflow to restart from the last checkpoint and evolve for a few more steps to prevent this issue (or a similar one) from appearing again.