Problems starting with speech module

[darwinscusp]

Hi,

I'm running nwn 1.65 on RedHat Enterprise Linux 4. I compiled NWNX2 successfully, and also managed to compile the speech plugin as well (though not easily, as my mysql libs and includes are not in a "standard" location. Odd since I compiled it from source ).

Anyway, each time I start nwn, it appears to come up ok at first, then crashes with a Segmentation Fault. Here's the exact message:

[darwinscusp]

Ok, not to be a dweeb and follow up on my own question, but I've done some additional work and isolated the problem somewhat.

I completely deleted and reinstalled NWN, complete with all my haks, modules, etc. I then deleted all the nwnx stuff from my build tree, and unarchived and rebuilt all those components as well, verifying everything linked to mysql (and in the case of the speech plugin, that -lpcap was used).

I'm now able to start modules, and run with normal database interaction just fine. However, whenever I issue a command (like *talk*) or anything that the speech plugin would grab, the server Seg Faults, and crashes.

I did not compile nwnx with debug symbols, but can do so if that will help diagnose the speech plugin issue.

Thanks,

darwinscusp

[JeroenB]

Have you done everything thats stated in the document (also in the OnLoad event)? And did you also imported the ERF of Lanthar? If you haven't this can cause undefined behaviour.

[darwinscusp]

I don't recall anything specific to the speech plugin that required modification of the onload event.

I'm using a module that I've previously developed on Windows. I'd already imported Lanthar's erf into this module, and it worked there fine.

The only difference in aps2.erf that I noticed, between the Windows and Linux versions, was the addition of the NWNX!INIT line in SQLInit() for the linux version. I'll reimport Lanthar's erf, as well as aps2.erf, recompile, and try again.

Thanks for the response JeroenB. I'll post my results.

[darwinscusp]

Ok, I reinstalled the aps_include and aps_onload scripts from aps2.erf from the linux-version of nwnx2 (the most recent one). I then saved, exited, reopened, recompiled.

When I launched the server, everything appeared normal, and database interaction worked as I'd expect (I have a DB-based spawn system that worked fine).

However, again, whenever I issued *help* or any string the speech plugin might interact with, the server segfaults, and crashes.

I did an strace, to see exactly what's going on, but it's not particularly helpful:

[JeroenB]

In the onload you must specifiy the name of the server, otherwise it doesn't know where to look (it early .. so I hope I don't talk nonsens here)..

[darwinscusp]

Ahhh, true enough. Since I reimported the lds module, I need to add those two lines back. I'll give that a try this evening.

Thanks for the response, and sorry I keep asking so many questions. I have all this working in VMware running in windows, but I'd really prefer to get away from windows completely if possible.


Thanks again.

[darwinscusp]

Ok, I readded the code to set the module name, but I continue to get the segfault.

So far, I've completely reinstalled NWN from scratch, installed nwnx and the speech plugin, compiling and installing both. I've also reimported the aps2.erf and the Talus erf, adding code to the lds_mod_load (or whatever it's called), to include the two lines mentioned in the speech plugin documentation about the server name.

NWNX is still working like a champ, but the speech plugin is not getting along with something, causing a segfault each time it's invoked, and crashing the server.

[JeroenB]

I will have a look at my configuration of it soon.

[darwinscusp]

Thanks JeroenB.

One thing I've noticed, is my speech plugin seems to compile libpcap and libmysqlclient statically, rather than shared. At least that's my guess -- here's what makes me think so:

your nwnx_speech.so:

[darwinscusp]

Because I was concerned that perhaps the shared library issue mentioned above might be causing some problems, I recompiled libpcap and mysql (was upgrading anyway), and made sure to compile with shared libraries. I then recompiled all of nwnx and the speech plugin. ldd nwnx_speech.so now lists libpcap and libmysqclient properly. But I still experience the segmentation fault issue.

Hope that helps maybe narrow the issue some.

If I can do anything further to help troubleshoot this, or provide additional info, I'm all ears.

Thanks,

darwinscusp

[darwinscusp]

In the process of troubleshooting my nwnx_speech.so problem, I created a (new) Makefile.in for the speech plugin. Using the following steps, once can configure,make, and install it, via the nwnx2 steps.

Here are the steps.

1) tar xzf linnwnx2-2.5.3-rc1.tar.gz.
2) cd linnwnx2.
3) mkdir speech.
4) cd speech.
5) unzip /your/full/path/nwnx_speech_linux_01.zip.
6) create a file called Makefile.in containing the code below.
7) cd .. # You should now be in /blah/blah/linnwnx2
./configure --prefix=/usr/local/nwn --with-extraplugins=speech
9) make
10) sudo make install

I've been screwing with this code tree for a while, so it's possible I've overlooked something. Feel free to let me know.