need for release?

Filip Konvička filip.konvicka at logis.cz
Mon Feb 11 12:28:14 CET 2008


Dear hackers,
> Hi Filip,
>
> Thanks for your reply. We were discussing your proposal yesterday and 
> we also
> agree that it would be very good to include this fix before releasing. 
> Some
> observations bellow.
I spent some time last week investigating this bug, and I finally 
decided against including this fix now (unless of course you find my 
arguments insufficient...). I closely examined the current code and I 
believe that it is, after all, correct. The error I was getting was very 
likely caused by my mistake: I was not closing sockets after connection 
shutdown, so I caused overflow in the descriptor table (which is 
const-sized).

The code that deals with fd_set is visible in the Win32 SDK headers, and 
I believe that it is used correctly in platform/emulator/os.cc.

I'll wait some time and if I'll see this error again, I'll try to fix 
it. Our server application is used daily in a production environment for 
about 2 years now, and this was the first time I've seen this error, and 
I haven't been able to reproduce it since then. Since it is a 
combination of <long running server; programming error - not closing 
sockets; Windows platform>, I think that no one should be affected by this.

I hope that I did not introduce any delays in the release process.

Cheers,
Filip



More information about the mozart-hackers mailing list