On 03/16/2014 10:12 PM, Andrew Beekhof wrote:
> 
> On 14 Mar 2014, at 12:40 am, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
> 
>>> So I guess the question now becomes why does Python mode conses so
>>> many Lisp objects that it triggers GC so frequently.
>>
>> M-x profiler-start RET mem RET
>> ...reproduce the heavy allocation problem...
>> M-x profiler-report RET
>>
>> might be a good start.
> 
> 
> bwahahaha. I can go one better and give you a reproducer.
> Time wasn't a factor, it was switching to a particular file that triggered the (worst of the) problem.
> 
> http://paste.fedoraproject.org/85549/13948345/
> 
> I had a bug in the file that was causing issues for soemthing:
> 
> @@ -90,7 +92,7 @@ class CTSTest:
>          self.logger.log(args)
>  
>      def debug(self, args):
> -        self.logger.debug(args
> +        self.logger.debug(args)
> 
> 

No repro on trunk.