[Twisted-Python] Re: [Twisted-commits] r14922 - Fix to replace a lambda with something that's safe in the scope when it's

[Jean-Paul Calderone]

On Thu, 27 Oct 2005 15:27:36 +0100, Matt Goodall <matt at pollenation.net> wrote:
>Matt Goodall wrote:
>> Jean-Paul Calderone wrote:
>>
>>>On Thu, 27 Oct 2005 03:45:59 -0600, Matt Goodall
>>><mg at wolfwood.twistedmatrix.com> wrote:
>>>
>>>
>>>>Author: mg
>>>>Date: Thu Oct 27 03:45:58 2005
>>>>New Revision: 14922
>>>>
>>>>Modified:
>>>>  sandbox/mg/laxdb.py
>>>>Log:
>>>>Fix to replace a lambda with something that's safe in the scope when it's
>>>>actually executed .... just in case anyone else is actually using this
>>>>thing.
>>>>
>>>
>>>Note that this changes the behavior from constructing Failure instances
>>>in the reactor thread to creating them in a worker thread.
>>
>>
>> Yes, very true. Thanks for raising it. Does it matter in this case?
>>
>> All the the args passed to Failure are worked out in the thread and I
>> think that means Failure's __init__ does basically nothing. Did I miss
>> something?

It may work fine now.  The API isn't guaranteed to be threadsafe, though.

>
>Also, I think it's now pointless digging the exception information out
>of Python to pass to the Failure.__init__. That's what Failure will
>*now* do on its own when passed nothing.

As it stands in SVN now, I agree.  The Failure was constructed that way before since it was being instantiated in a different thread than the one in which the exception occurred.

Jp