Automatically assigning validations within feature branches

Hi @Martin,

Thank you for your input. Administratively, this approach seems logical.

However, could you provide further details on the process-related advantages it offers? Whether all validations are assigned to oneself or left unassigned, they are both immediately visible upon accessing the validation screen by default. Therefore, I'm questioning the additional functional value in this situation.

The function is that if you merge before all validations have been solved (you shouldn’t, but hey, sometimes it happens ), you can still see to whom these validations belong.

Currently, we have solved this by creating an additional validation step that is executed last and assigns all unassigned validations. For this validation we created a new validation group with the highest sequence no to make sure it is executed after all the other validations.
 

In the code we made an extra check to make sure the validations are only assigned to the developer that is executing the validation job. Not sure if this is the best way to deal with this, but it works for us.  

-- declare parameters
declare @developer_id      name
       ,@cur_validation_id validation_id
       ,@pk_hash           hashbytes_md5

-- get the developer executing the validation
set @developer_id = (
                    select top 1 
                           developer_id
                    from job j
                    where job_type   = 1 -- validation
                      and job_status = 2 -- running 
                      and finish_date_time is null
                      and exists (
                                 select 1
                                 from usr u
                                 where u.usr_id = j.developer_id
                                 )
                    order by 1 desc
                    )

-- if the developer exists then start assigning 
if @developer_id is not null
begin
    -- get all unassinged validations 
    declare cursor_assign_validation cursor for
    select v.validation_id
          ,vm.pk_hash
    from validation_msg  vm
         join validation v
           on v.model_id      = vm.model_id
          and v.branch_id     = vm.branch_id
          and v.validation_id = vm.validation_id
    where vm.model_id  = @model_id
      and vm.branch_id = @branch_id
      and not exists (
                     select 1
                     from validation_msg_assigned vma
                     where vma.model_id      = vm.model_id
                       and vma.branch_id     = vm.branch_id
                       and vma.validation_id = vm.validation_id
                       and vma.pk_hash       = vm.pk_hash
                     )

    open cursor_assign_validation
    fetch next from cursor_assign_validation
    into @cur_validation_id
        ,@pk_hash

    while @@fetch_status = 0
    begin
        -- execute software factory assign task  
        exec task_assign_validation @model_id
                                   ,@branch_id
                                   ,@cur_validation_id
                                   ,@developer_id
                                   ,@pk_hash

        fetch next from cursor_assign_validation
        into @cur_validation_id
            ,@pk_hash
    end
    close cursor_assign_validation
    deallocate cursor_assign_validation
end

To add here, the moment of assigning them to the developer is also up for debate. My suggestion would be: after executing validations, that any messages that come out are automatically assigned to the one executing the validation.

However, developers could opt to not validate at all and merge any changes to another branch, causing validation messages to appear in that branch and being assigned to a developer that does not know where these messages come from. How should the Software Factory deal with this situation in your opinion?

I think part of the requirement for merging is having had a validation run. Or even run it as part of creating the merge session…

‘create-merge-session’ → start validating → warn if there are any and ask if user wants to continue → do the actual merge-stuff.

I mean, it does not make sense to not run your validations before merging. Make it mandatory.

What about automatically assigning the validation to the one that created or mutated the related object? I think that would make much more sense.