LoadEventNexus bug for instruments with detectors missing from IDF

[Russell Taylor]

Load a BASIS file and you will observe an anomalously high number of events in the first event list/workspace index. What is happening is that the BASIS IDF has gaps in the sequence of detector ids, yet those detectors appear to not only physically exist but to sometimes have events in them. Because of some unfortunate choices for default values within the code, these events are ending up being put in workspace index 0 instead of being discarded.

This is also the reason that we are having failures with loading a BASIS file when a mutex for the event lists is missing (see #7489) - multiple threads are ending up holding references to the first event list when it should only be possible for one to operate on it.

[Russell Taylor]

Several commits are already there in the branch bugfix/7524_basis_loading_error but haven't shown up here.

[Russell Taylor]

Re #7524. Additional testing for getDetectorIDToWorkspaceIndexVector.

Changeset: 9cbb55f14c18213c1d482711bafed650629451c8

[Russell Taylor]

Re #7524. Additional test on LoadEventNexus to avoid regression.

Changeset: 5138a888092aee8c3359800fcb0c866fd6525a29

[Russell Taylor]

Re #7524. Add log message about discarded events.

Changeset: 0f00daae44199a3215984164e8026f07a05279e4

[Russell Taylor]

Tester:

Before this fix, the first spectrum of a loaded BASIS dataset (there's one each in AutoTestData & SystemTests/Data) would have an anomalously high number of events in it. Now it will have the correct number (I verified this by looking into the nexus file itself with HDFview).

Also, there were unit (ModeratorTzeroLinear) and system tests (BASISAutoReduction & LoadLotsOfFiles) that have been failing (crashing) on master because of this bug. In develop, they should have been behaving.

[Andrei Savici]

git test pass did not close the ticket

[Stuart Campbell]

This ticket has been transferred to github issue ​8369