Mnesia Bug with select and disc_only_copies tables

[Rudolph van Graan]

>
> See the manual:
>
>  For efficiency
>           the NObjects is a recommendation only  and  the  result
>           may  contain  anything from an empty list to all avail-
>           able results.


Does "Empty list" here mean "empty" even though the query matched  
results? Or does it mean "Empty" when there was no results? For me  
"recommendation" means that the query should return *at most*  
NObjects if any results matched, or if the continuation mechanism  
fails (for dets) *all* the data. But not nothing.

My experience shows that by only removing the NObjects argument, the  
query worked, otherwise it did not return *any* data. The same query  
without NObjects did return data.

I haven't had time to dig around too much, but your code runs here as  
well. My table has indices, which might be part of the issue. I am  
happy if the query *ignores* the NObject, but not for it to return no  
data.

Do you want me to try and narrow this down to a reproducable scenario?


>
> (it is written because of dets :-)
>
> It works here:
>
> 1> mnesia:start().
> 2> mnesia:create_table(a, []).
> {atomic,ok}
> 6> mnesia:change_table_copy_type(schema, node(), disc_copies).
> {atomic,ok}
> 7> [mnesia:dirty_write({a, N,N}) || N <- lists:seq(1,1000)].
> 8> mnesia:change_table_copy_type(a, node(), disc_only_copies).
> {atomic,ok}
> 14> P = ets:fun2ms(fun({a,_,A}) when A > 990 -> A end).
> 16> F = fun('$end_of_table', _, Acc) -> Acc;
>            ({Recs,Cont}, Fun,Acc) ->
>            R = mnesia:select(Cnt),
>                Fun(R, Fun, Recs ++Acc)
>         end.
> 17> mnesia:transaction(fun() ->
>                  Cont = mnesia:select(a, P, 10, read),
>                  F(Cont, F, [])
>                end).
> {atomic,[996,997,1000,998,994,992,993,995,999,991]}
Regards,

Rudolph