[TxMt] Re: mate -w behaviour

[Allan Odgaard]

On 24 Nov 2009, at 13:02, Glyn Normington wrote:

>> Indeed that is because ‘mate’ never gets a “close” signal  
>> from
>> TextMate, as ⌘Q bypass the code which would call that.
>
> That makes sense.
>
> Isn't it odd that this didn't happen on Leopard too?

Application shutdown is mainly handled by the Cocoa framework. Apple  
tries to “optimize” this by not going through normal object  
destruction — sounds like until Snow Leopard, the object responsible  
for talking with ‘mate’ was still properly destructed during exit,  
and thus able to tell ‘mate’ to exit.

>> It isn’t something I’d considered before, the approach taken in  
>> 2.0
>> though does not show this problem.
>
> Interesting. Can you say how the new approach avoids the problem?

In 2.0 communication happens via sockets, if the server goes down, the  
client will learn. That is not the case with distributed objects  
(which 1.x uses).

> Roughly when will 2.0 ship?

When it is ready ;) Still a lot of work to do…