Re: flawed LuaSocket select()/tcp.receive() interaction ?

[lua@...]

* On 2006-11-20 Diego Nehab <diego@tecgraf.puc-rio.br> wrote  :

> Hi,
> 
> >Yes, I know that page - can you please specify what was
> >not clear about my question, this will surely help me ask
> >smarter ones next time.
> 
> I meant the fact your subject claims there is a flaw
> somewhere, rather than considering the possibility there is
> a problem with your understanding of the issue.
>
> I am obviously overreacting here. To protect yourself from this kind
> of reaction, simply approach the potential problems in a more subtle
> way.

Yes, you are completely right about this one, my apologies. Although I
added a question mark to indicate I am open to other causes, of course :)
 
> 
> >No, I'm afraid this is not the case. From the linux man page of recv:
> >"The receive calls normally return any data available, up to the
> >requested amount, rather than waiting for receipt of the full amount
> >requested."
> 
> You are right. Even the POSIX standard agrees with you (man pages are
> not reliable enough in multi-platform applications).  

True, but manpages are fast and nearby, and mostly without bugs.

> >Nonblocking mode should never be necassery when using select(),
> >because of the proper behaviour of recv() as I described above.
> 
> See the remark on send()...

Yes, that's a good reason.

Anyway, thanks again, my problem is fixed !
 
-- 
:wq
^X^Cy^K^X^C^C^C^C