FWD » Core Development » Database

• oo.Test1:
  

  class oo.Test1:
      def private temp-table tt1 before-table btt1 field fa as int.
  
      method public void m1().
         def var ht as handle.
         def var hb as handle.
         def var hf as handle.
  
         create temp-table ht.
         ht:create-like("tt1").
         ht:temp-table-prepare("tt2").
         hb = ht:default-buffer-handle.
         hf = hb:buffer-field("fa") no-error.
         message hf:name error-status:error error-status:get-message(1).
      end.
  end.
  
  

• oo.Test2
  

  class oo.Test2
  inherits oo.Test1:
      def private temp-table tt1 before-table btt1  field f1 as int.
  
      constructor public Test2().
         m1().
  
         def var hb4 as handle.
         def var hf as handle.
         hb4 = temp-table tt1:before-table.
         hf = hb4:default-buffer-handle:buffer-field("f1") no-error.
         message hf:name error-status:error error-status:get-message(1).
      end.
  end.
  

• testrun.p
  

  def var l as oo.Test2.
  l = new oo.Test2().
  

• temp-tables during the constructor are registered with BufferManager using without the defining type. This creates collisions - the DMO definition has the type where is defined, but is not being used to register the 'pending temp-table' with BufferManager
• AbstractTempTable.computeTableType does not consider the defining type of the temp-table - it will use just the before-table's name, and look into BufferManager. This can end up with the wrong before-table instance being resolved.

These problems were found in the #8363 app.

DATASET:CREATE-LIKE("<dataset-name>") needs to be checked, too.

Ovidiu, please review this, I'll get ETF testing tomorrow and if is OK, I'd like to merge this tomorrow.

Review of 8554a, r15120-15122.

I am OK with the code. I'm sorry I made a mess with condition codes which do and do not generate errors. I hope there are no others.
Is it possible to reduce the visibility of g/setDefiningType() for datasets only to package or protected?

You can merge to trunk now.

Hynek: 8152a can merge after 8554a.

Branch 8554a was merged into trunk as rev. 15125 and archived.