[Python-Dev] Python 2.6.3

[Scott Dial]

Nick Coghlan wrote:
> Scott Dial wrote:
>> I would appreciate this bug being resolved before the next release as it
>> effects me on a daily basis. I have submitted a patch, which reflects my
>> local solution.
> 
> Unfortunately, it's almost certainly too late to get this into 2.6.3. It
> really needed to be brought up back when Barry called for identification
> of significant 2.6 bugs and patches rather than after the RC had already
> been released.
> 

I understand. I didn't figure out the bug until after rc1 landed. It was
only happening spuriously with my mail server, never when I manually
tried to invoke the problem. I was forced to let my cronjob run until
the kernel killed getmail for OOM, giving me the trace shown in the issue.

It is impossible to break anything with this patch, as no program could
proceed. The 3-line patch merely converts it back into the exception
that was originally raised prior to 2.6, so it's not a new behavior in
that respect either. I wouldn't have anticipated this much resistance to
removing an infinite loop from the standard library. I could also for
this patch from the angle that it allows a remote host the ability to
execute a denial-of-service attack (even by accident!) since the
infinite loop appends an empty string to a list on every loop, taking
CPU time and memory with it. Allow me to be naive for a moment and say,
is this not the point of rc1 but to catch bugs that should not be in the
final?

Of course, it's Barry's call.

-- 
Scott Dial
scott at scottdial.com
scodial at cs.indiana.edu