Re: beagle-index-gaim-logs killed my system

[Christopher Orr <dashboard protactin co uk>]

Hi all,

On 04/08/04 11:28, Michael A. Koziarski wrote:
> I have quite a few gaim logs, though not an insanely large number, but 
> something about them causes beagle-index-gaim-logs to completely freak out.
>
> It successfully indexes a number of the conversations, then when it hits 
> one of the larger ones it prints the following warning:
>
> GC Warning: Repeated allocation of very large block (appr. size 335548416):
>         May lead to memory leak and poor performance.
>
> And then the poor performance begins.   My system becomes almost totally 
> non-responsive & I have to hit C-c repeatedly till it eventually dies.

I've been meaning to mention a similar situation I had, whereupon 
hitting a certain log file the indexer started consuming vast amounts of 
memory until it just killed itself off.

I traced this to StripTags in Util/ImLog.cs which removes any HTML from 
a given line of input.  However, I found a single occasion in my logs -- 
in old-style gaim log format -- which had a piece of HTML split over two 
lines, thus causing the infinite loop as StripTags tried to search for 
'>' on that line.

I guess the only way to solve this is to allow the log parser to read 
ahead in the log file somehow.  Or in the meantime, use the "--verbose" 
flag with beagle-index-gaim-logs and investigate the conversation it 
gets stuck on.  That is, assuming what you've encountered here isn't 
some new, unrelated problem!

Regards,
Chris