[Twisted-Python] [Twisted] #4378: Error handling in NetstringReceiver broken

Albert Brandl albert.brandl at weiermayer.com
Tue Mar 30 10:14:18 MDT 2010


On Tue, Mar 30, 2010 at 12:27:54PM -0000, Twisted wrote:
>  1. I'm pretty sure this patch, in its current form, will break in the
>  (still supported) Python 2.4, since it uses the 2.5 generator feature
>  where you pass in data.

Yep - I'll fix this.

>  2. Why not initialize the parser in the Protocol's __init__? Less code in
>  a frequently called path.

Good idea, since there's no need anymore to check in every call of dataReceived.

>  3. ``stringReceived`` in docstring violates coding standard, should be a
>  L{} or C{} or something, I forget which. So does received_string variable
>  name.

Oops - this looks like ReST markup :-(.

>  4. In dataReceived, when would we expect to get StopIteration to be
>  raised? Quick read of the generator doesn't show anywhere it actually
>  returns, it seems to be in infinite loop.

When the generator is left by raising an exception, it will raise a StopIteration
the next time someone sends data. But this is irrelevant now, since Python 2.4
does not support sending data to a generator.

>  5. Using long() is unnecessary, you can just use int().

Ok.

>  6. Hooray for more and better tests! I am uncomfortable about use of
>  transport.disconnecting, but we've already gone down the path of using it
>  elsewhere, and not sure what else you could do, so it's probably fine.

I might introduce a new exception (see discussion on the mailing list) - this
would allow to get rid of the check for transport.disconnecting.

>  7. In test_receiveTwoNetstrings(self), perhaps the second string should be
>  different than the first, e.g. "a" and "woo", to catch theoretical bugs
>  where parser repeatedly spits out first thing it parsed.

Good idea.

Thanks & best regards,

Albert
-- 
DI Albert Brandl
Weiermayer Solutions GmbH | http://www.weiermayer.com
4813 Altmünster | Abteistraße 12 | Austria
Fon: +43(0)720 70 30 14 | Fax: +43 (0) 7612 88081




More information about the Twisted-Python mailing list