rails 6.0 dependency

[dglaser]

Included #710 changes from @serkaniyigun
Since rails 6.0 has been released, I changed the gem depency from rails ~>6.0 beta3 to ~> 6.0

Tests are failing:
TinyTds::Error: Cannot insert explicit value for identity column in table [tasks] when IDENTITY_INSERT is set to OFF

This could be as simple as setting IDENTITY_INSERT to ON/OFF. Doesn't help with activerecord tests that fail.
before do
connection.execute("SET IDENTITY_INSERT on [table] ON;")
...
after do
connection.execute("SET IDENTITY_INSERT [table] OFF;")
....

[ttilberg]

While digging backwards for the identity issue, I'm finding that one potentially related issue is that ActiveRecord::ConnectionAdapters::SQLServerColumn doesn't seem to correctly report whether it is an identity column. I could see this kind of problem causing all sorts of interesting issues when you run the test suite.

=> #<ActiveRecord::ConnectionAdapters::SQLServerColumn:0x000055ae6b885ee0
 @collation=nil,
 @comment=nil,
 @default=nil,
 @default_function=nil,
 @name="id",
 @null=false,
 @sql_type_metadata=
  #<ActiveRecord::ConnectionAdapters::SQLServer::SqlTypeMetadata:0x000055ae6b886070
   @limit=8,
   @precision=nil,
   @scale=nil,
   @sql_type="bigint(8)",
   @sqlserver_options=
    {:sqlserver_options=>
      {:ordinal_position=>1, :is_primary=>true, :is_identity=>true}},
   @type=:integer>,
 @sqlserver_options={}>

Notice specifically that @sqlserver_options is an empty hash, but @sql_type_metadata knows it's an identity column deeper down. The check for identity_required? uses that @sqlserver_options ivar.

Looking at the file lib/active_record/connection_adapters/sqlserver_column.rb -- the initializer is explicitly setting @sqlserver_options to an empty hash, and seemingly never doing anything with it, however these accessors are relying on the data it's supposed to entail.

We can see this was recently introduced at dglaser@b4f558e to handle kwargs, but was never assigned in the body.

I submitted a simple PR to the branch you are attempting to merge.

[ttilberg]

In the test run I'm working on, I'm seeing a ton of errors related to

NoMethodError: undefined method `split' for nil:NilClass

Upon digging into the stacktraces, this seems to be getting caused by something quirky in Minitest, not necessarily the AR adapter. I left some information here: minitest/minitest#831

[dglaser]

The PR you submitted is close to reverting the code I included from PR #710. I feel it would be cleaner to close this PR.

[ttilberg]

Yes, I think it might be a better starting place if the rest works as expected.

[utkarsh2102]

@ttilberg, you might be interested in this: minitest/minitest#831 (comment) :)