Re: [Caml-list] Weak array semantics with mutually recursive values.

[Christophe Raffalli <craff73@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1151 bytes --]

Le 24/01/11 13:45, Damien Doligez a écrit :
> On 2011-01-22, at 17:29, Christophe Raffalli wrote:
>
>> for instance. They will become unreachable by the GC at the same cycle.
>> However,
>> if they are both added in a weak array, it seems that they may not be
>> removed at the same time from the array.
>>
>> I have some code that seems to show that most of the time they are
>> removed at the same times, but very rarely this fails.
> As far as I can tell, it should never happen.  Do you have a good
> repro case?
Hello,

I will try to investigate further my example to make sure it is not
another problem. If I get
a simple repro case, I will post it.

What I was considering recently (after my second post) is the following
scenario:

t and u are two inaccessible mutually recursive values, but one is still
in the minor heap (t) while the other (u) has been promoted (if a minor
cycle just ended between the allocation of both values, this seems
possible even if unlikely). In this case, the finalisation of t and its
removal from a weak array might occur long before the finalisation of u ?

Cheers,
Christophe



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 259 bytes --]