[Pan-users] Re: Filtering unusable slow

[Greg Lee]

On Thu, 16 Aug 2007 21:04:12 +0000, Frederik Himpe wrote:

> I cannot really imagine how another application can be still be
> "considerably worse" than this.

Searching headers for a pattern in a high traffic group was extremely
time consuming in Tin.  After 15 minutes of my system being almost 
completely tied up with both cores pushing 100% usage, I would hit
'q', Tin would ask if I wanted to abort the search, I would say yes, but
after a half hour, it still couldn't pull itself out.  Opening a new
group that Tin didn't yet have an overview file for was in some cases
just impossible. I would let it work over night and wake to find it
taking over two gigabytes of my 3 of ram.  A few more hours and it would
take more, plus several gigabytes of swap.  Then the kernel would oom my
X window system. I'm a stubborn person, but I just had to give up.
 
> I just tried filtering for CFS in one
> month of linux kernel mailing list, and I killed it again after several
> minutes of 100% CPU time. Pan has become completely unusable for me now:
> it was certainly not like that before, otherwise I would never used it
> for months without noticing and complaining... Could it be a regression
> in 0.132?

It could be, I suppose, but the number of headers I keep around tends
to grow over time, naturally increasing search times, so without doing
actual tests, I couldn't say about this, one way or the other.

-- 
Greg