[Caml-list] Mismatch between Gc.full_major() documentation and behaviour

[Caml-list] Mismatch between Gc.full_major() documentation and behaviour

[Bertrand Jeannet]

According to the documentation, the sequence 
"Gc.major();Gc.full_major()" should be equivalent to "Gc.full_major()", 
as Gc.full_major() is told to terminate the current major gc and perform 
a complete new cycle.

In my case it is not the case: "Gc.major(); Gc.full_major()" performs 
more calls to finalisation functions than "Gc.full_major()".

My context: I use a weak hashtables containing finalised value (with 
Gc.finalise).

The documentation of Weak module specifies that the finalization 
function is called after the weak hashtbl/array has been set to none.

Does someone know the explanation of this ?

Re: [Caml-list] Mismatch between Gc.full_major() documentation and behaviour

[Török Edwin]

On 12/11/2013 01:46 PM, Bertrand Jeannet wrote:
> According to the documentation, the sequence
> "Gc.major();Gc.full_major()" should be equivalent to
> "Gc.full_major()", as Gc.full_major() is told to terminate the
> current major gc and perform a complete new cycle.

They both perform a minor GC cycle too.
So you run a minor cycle twice with 'Gc.major(); Gc.full_major()', but only one with 'Gc.full_major()'.

> 
> In my case it is not the case: "Gc.major(); Gc.full_major()" performs
> more calls to finalisation functions than "Gc.full_major()".
> 
> My context: I use a weak hashtables containing finalised value (with
> Gc.finalise).
> 
> The documentation of Weak module specifies that the finalization
> function is called after the weak hashtbl/array has been set to
> none.
> 
> Does someone know the explanation of this ?

You can run 'Gc.print_stat stderr' to check how many GC minor/major cycles are run in both cases. The number of minor cycles should be different.

Best regards,
--Edwin