intermittent misbehavior in /Logger/Channels/0/Settings/ODB dump

[Amy Roberts]

When taking runs that (1) use a byte limit for the subruns and (2) use a delayed transition to stop, we sometimes see issues with ODB dump in the last and second-to-last dump files in the run.

The second-to-last file will have two end-of-run events (ID 0x8001) at the end of the file, while the last file will have no end-of-run event.

In all the cases I've looked at so far, the last file has only a beginning-of-run event (ID 0x8000) and no other data. It's not clear if data should be written, but isn't - or if Midas is printing spurious begin-run and end-run events.

[dd1]

need to test this using deferred transition code from Thomas. K.O.

[dd1]

the sequence is expected to run like this: logger checks the byte count, decides to stop the run, executes cm_transition(), which returns CM_DEFERRED_TRANSITION, (logger should continue as if nothing happened), the deferred frontend eventually executes the run stop transition, which calls mlogger to close the files, etc. The 0x8001 odb dump should be written at this time, once, into the latest open file. obviously something goes wrong somewhere to have the 0x8001 odb dump written twice into the wrong file. K.O.

[dd1]

cannot duplicate, insufficient information, bug is too old. please reopen of additional information becomes available. K.O.