[Twisted-Python] Weekly Bug Summary

[John Santos]

On 5/1/2019 02:01 AM, Glyph wrote:
>> On Apr 28, 2019, at 8:01 AM, John Santos <john at egh.com 
>> <mailto:john at egh.com>> wrote:
>>
>> I've been having ongoing problems accessing the TwistedMatrix site.  
>> (Tried both Firefox and Chrome from my home PC.)
>>
> The site was having some issues and the server badly needed a reboot 
> during the time you were trying to access it.
>>
>> I've been trying to log in to update issue 9629 below to report it is 
>> a duplicate of a known problem.  (Can't tell you WHICH known problem 
>> because I can't search the Twisted tickets.)
>>
> Hopefully the search function is back?  But also you can try searching 
> with Google, using 'site:twistedmatrix.com/trac/ticket' 
> <http://twistedmatrix.com/trac/ticket'>.
>>
>> On Monday evening, when I submitted this issue, I filled in the form 
>> and nothing appeared to happen.  (I may have even submitted it twice.)
>>
>> I searched the tickets and discovered 1) it didn't appear (so no 
>> ticket?) and 2) it seemed to be the same as a previously reported 
>> problem, so I added a comment to the previous ticket that I was seeing 
>> the same issue and a little context that might help.  (I only started 
>> seeing it with Twisted 19.2.0, but the original report said it was 
>> under 18.4.0 which I had been running uneventfully since last summer.)
>>
>> Anyway, I got no feedback, so I assumed my initial attempt to create a 
>> ticket had failed.
>>
>> Someone with access, please mark #9629 as a duplicate and close.
>>
> Done.  Looks like https://twistedmatrix.com/trac/ticket/9462 was the 
> issue in question; I see your comment on it.
> 
> Thanks for writing the email to follow up :).
> 
> -g

Yes, that's it.  Thanks, Glyph.  (I always get a warm feeling when I see 
a bug* get stomped even when it's just because it's a duplicate!)

I also saw your later comment that it might be related to #7926.  I 
suspect you are correct, though it makes my brain hurt to try to follow 
it.  I think TLS is trying to deal more sensibly with connections 
getting intentionally shut down by either end, but issues still arise 
when shutdowns happen in the middle of negotiations.  (Which broaches 
the issue "can this be leveraged into a DOS attack?", on top of just 
trying to deal with it correctly.)  In my case, though, I think the 
issue pops up when my server times out and closes the connection, but 
the client end (a vanilla web browser such as Firefox or Chrome) doesn't 
do exactly what the TLS server code expects.  Which side (client or 
server) is behaving incorrectly?  I have no idea.

[*] a computer bug**, not a real bug, for any hard-core Buddhists or 
Jains lurking... ;-)
[**] I've seen the original Mark II/Grace Hopper bug.  It was actually a 
large moth, scotch-taped to the console log...  Found a picture at 
https://www.atlasobscura.com/places/grace-hoppers-bug  (Maybe you can 
tell I'm in full work-avoidance mode.)

PS the web site is working much better today, thanks.

-- 
John Santos
Evans Griffiths & Hart, Inc.
781-861-0670 ext 539