brad miele scribbled on 8/18/06 1:27 PM:
> Hello everyone,

Hi Brad. Hope that Maine weather is fine.

> I have been messing around with incrimental, which i know is 
> "experimental" (how long will it be experimental anyway?).

Guess it'll be experimental as long as it keeps doing things like what you're 
describing. :)


  But i had
> not been getting very far because it seems that the swish-es that i build 
> with --enable-incrimental have this tendency to segfault during the 
> indexing part of things:
> 
> Writing header ...
> Writing index entries ...
>    Writing word text: ...Segmentation fault (core dumped)
> 
> then today, i noticed that if if i didn't use economy mode, things work. 
> so is economy mode and incrimental generally a show stopper? i can through 
> more resources at it in that case i guess. and yeah. incrimental in 
> general.. yes? no? i want to start letting our editors adjust the index in 
> near real time, and we have about 1.5 million records now.

Is it reproducible for a small set of docs? (<100) Could you put together a 
test? The way to track this down is to run it through gdb or similar to find 
where it's segfaulting.

There's been talk off and on about getting a good test suite together for the 
incremental index feature, but as its chief author is no longer on the list, 
there's little momentum there.

-- 
Peter Karman  .  http://peknet.com/  .  peter(at)not-real.peknet.com