segfaults with decached

Submitted by jeffrice on Wed, 05/18/2005 - 07:29

I'm sorry for such a burst of posts! I don't know if this is a unique problem, but I am getting a segfault if I have any of the wl_ options set to dbcached. They work fine if set to db. Has anyone run into a similar problem?

Forums: 
gps forum

Submitted by mimo on Wed, 05/18/2005 - 15:09.

Do the tables exist? Otherwise run gps once but set mode to init in the config file to create the tables.

Submitted by jeffrice on Wed, 05/18/2005 - 16:44.

Yes, the tables exist. If the table is empty, GPS doesn't segfault if dbcached is set -- it only occurs if the table has entries in it. If I use db instead, the table is read normally. I will try this on a second system -- I had some problems getting GPS to compile -- is it possible that an error occurred somehow during compilation? Otherwise it seems to work fine...

Jeff

Submitted by jeffrice on Wed, 05/18/2005 - 18:03.

Hmm, reading some of the other posts, I am wondering how much dbcache actually matters in my situation anyway. The db is only cached if a new query comes in before the GPS process is killed, right? That's a window of a minute or so, right? I don't have a lot of traffic so most of the time, it would be a new query.

I keep going back and forth between GPS and Gld. Gld seems a bit faster (which I think is natural since it runs as a daemon), but the "reverse" feature of GPS is very appealing.

Jeff

Submitted by mimo on Wed, 05/18/2005 - 18:44.

re compilation.

I installed gps on a fresh system today and have to say: It's not exactly simple.

I think I'll post a mini-howto here sometime soon.

One otehr thing I noticed is that I get segfaults when using the stock libdbd-mysql package in debian. But when I rebuild teh package from source gps runs fine. Is that maybe what you are observing?

Re gld -- I have looked at it. Written in C and certainly much faster than gps. There is always a tradeoff between "features" (bugs) and performance. But I dont think that has to do with the fact it's a daemon. In a way my thinking was that connecting from one daemon (gld) to another daemon (mysql) is a bit redundant.

Submitted by jeffrice on Sun, 05/22/2005 - 00:48.

I noticed a new version of libdb-mysql came down today, and the seg faults appear to have gone away. I think... fingers crossed.