Two changes to defer.py

1) get rid of 'logError' as default errback. Tv complained about this, and I agree that it's silly -- logging should happen at the end of the chain, not at random points during the execution. instead of using 'logError', it now just uses (lambda x: x).

2) Get rid of old PB compatibility thing. We had a check for StringType in the result of a call/errback, because PB errors used to be strings. They're not any more. I committed the Failure patch months ago ;-)

And one more thing that I propose, but that's not in this patch: Enforcing the arguments to errbacks to be Failures. Very confusing things may happen if you pass something other than a Failure to an errback, and besides, passing something other than a Failure is like using 'raise "string"' rather than 'raise MyError("string")'.

Here's the patch:

Index: twisted/internet/defer.py =================================================================== RCS file: /cvs/Twisted/twisted/internet/defer.py,v retrieving revision 1.12 diff -u -r1.12 defer.py --- twisted/internet/defer.py 24 Jul 2002 19:04:56 -0000 1.12 +++ twisted/internet/defer.py 1 Aug 2002 00:42:15 -0000 @@ -103,7 +103,7 @@ These will be executed when the 'master' callback is run. """ cbs = ((callback, callbackArgs, callbackKeywords), - (errback or logError, errbackArgs, errbackKeywords)) + (errback or (lambda x: x), errbackArgs, errbackKeywords)) if self.default: self.callbacks[-1] = cbs else: @@ -228,14 +228,8 @@ callbackArgs=(0,), errbackArgs=(1,)) break - if type(self.result) != types.StringType: - # TODO: make this hack go away; it has something to do - # with PB returning strings from errbacks that are - # actually tracebacks that we still want to handle as - # errors sometimes... can't find exactly where right - # now - if not isinstance(self.result, failure.Failure): - self.isError = 0 + if not isinstance(self.result, failure.Failure): + self.isError = 0 except: self.result = failure.Failure() self.isError = 1