NH-4013 - SQL-Server batcher failure after CloseCommands.

[fredericDelaporte]

NH-4013 - SqlClientBatchingBatcher CloseCommands contract violated

CloseCommands causes SqlClientBatchingBatcher and MySqlClientBatchingBatcher to be unusable although its xml comment states:

/// Use this method instead of <c>Dispose</c> if the <see cref="IBatcher"/>
/// can be used again.

Test case and fix.

[fredericDelaporte]

SQL-Server batcher and MySql batcher have much in common, but this is not the case of Oracle batcher. So we may refactor this with an additional abstract class, but I am not much inspired by the idea of introducing this for some cases and not for others. (And how to name that? MySqlServerBatcherBase?)

[fredericDelaporte]

Moved to ClearCurrentBatch.

[hazzik]

You still have the same issue as #513

[fredericDelaporte]

#513 looks invalid as an issue to me, since we are not supposed to reuse the session (and hence its batcher) after a failure. So for me there is no point trying to fix such case. Unless we intend someday to enable session recovering from failures.

Here I fix a "no exception" case, where calling CloseCommands causes the batcher to be no more usable while it should still be. This case may happen in almost "normal use" cases, when an user open a reader then commit a transaction without having closed the reader, while not using OnClose connection release mode.

[fredericDelaporte]

And now that is #513 which is "behind", since it will still leave the batcher unusable in case of preparation failure (network error or whatever). I think that is a bit non-sens to change code for an explicitly unsupported use case like re-using the session after an exception. The batcher will still be unusable if the failure occurs outside of DoExecuteBatch by example.

[fredericDelaporte]

Rebased for forcing rebuild, failure on teamcity side (xml report truncated).

[hazzik]

LGTM