-----BEGIN PGP SIGNED MESSAGE-----
Though, making this class Serializable is not the only difference
between the tarball version and the SVN version. So you are perfectly
right, taking the whole ThreadPool.java from the SVN fixes the problem,
though only making the class Serializable in the tarball code does not.
Anyway, it works for me now, thanks.
On 05/09/2013 01:58, Sean Luke wrote:
> What's curious about this is that I also tried the tarball on the website, and after replacing the
ThreadPool.java file with the properly working one (on SVN), it works great.
> On Sep 4, 2013, at 5:06 PM, Marc Segond wrote:
>> Dear Sean,
>> I tried with the SVN version, and indeed it works.
>> I tried again with the .tar.gz version, and I confirm that
checkpointing doesn't work when using multithreading. It generates
correctly the checkpoints, put when you try to resume from a checkpoint,
it loads all the history, but then fails to start again the evaluation
without printing any error or raising any exception.
>> As far as I could trace it, it stays stuck waiting for a thread.
>> Conclusion: use the SVN version ;)
>> Best regards
>> Le 4 sept. 2013 à 17:36, Marc Segond <[log in to unmask]> a écrit :
>>> On 04/09/2013 17:29, Sean Luke wrote:
>>>> On Sep 4, 2013, at 11:03 AM, Marc Segond wrote:
>>>>> - - launch a run with checkpointing AND more than one thread
>>>>> - - try to load from any checkpoint
>>>>> It gets stuck after loading the checkpoint, seems that it does not
>>>>> succeed to run the evaluation.
>>>> Per earlier discussion on this list, did you make the ThreadPool
>>> Yes, when it is not Serializable, it raises an error at execution time,
>>> which is not the case.
>>>> Marc, try the SVN repository version of the code rather than the
>>> tarball on the website.
>>> I will try.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
-----END PGP SIGNATURE-----