FreeOrion

I'm running into a rather annoying bug, and I'd like some help getting some more details on it so I can report it.

The bug comes up when I change between the categories in the production menu. The program freezes and the entire computer becomes sluggish. The bug doesn't always occur, I've only noticed it when a system isn't selected, but I haven't tested it for that reliably yet.

I am pretty suspicious that it isn't a problem with the FreeOrion code, but instead with the proprietary nVidia display driver I have. (The bug didn't seem to occur before I installed this driver) This is the main reason I haven't posted this as a bug report yet.

It's been rather tricky tracking down the exact source of the bug because the system gets so unresponsive I can't perform any actions after the freeze.

Here are the steps I'm planning on taking, any suggestions would be greatly appreciated:

-Update my nVidia driver to the version released a couple days ago
-Update to the most recent SVN version of FO
-Try to catch the bug in gdb (Do I have to do anything to create debug targets, or is the normal scons enough?)
-Try to trace into the bug in gdb
-Try to catch any bad memory accesses in Valgrind
-Get rid of the nVidia driver and see if it still happens.

You may have picked a bad time to ask for programming advice, as I think tzlaine will be away for a few weeks starting today...

However, tzlaine and I have both been making changes to the production screen and production-related code recently... When did this start happening precisely, and how frequently have you been updating from SVN?

Did it work before, but just start hanging recently, or you just tried recently and it hung the first time you tried?

You might want to try RC4 precompiled (if on Windows) or a SVN revision from a few versions ago... I'm not that familiar with SVN, but I assume you can check out older versions without too much difficulty.

And: do you have all the art files?

Edit: Just occured to me that you might be forgetting to recompile GG after updating... and if necessary copying the resulting libraries / dlls to where FO can find the updated versions.

That's okay, I'm in no hurry.



I have a version from SVN before all of those changes. I'm updating to the newest SVN version to see if it makes matters worse/better.



Most of the time it works. I am not 100% sure but it seemed to start happening after I added the new video driver. It did however, happen on the only version I've installed on this computer...



I'm running Fedora 4, but I can try older SVN versions...


Yup, I got that, I haven't had any problems displaying artwork.


Hmm... I will have to look into this, but this is probably not likely since I have thus far only had one version...

Thanks for your help so far!

If the only thing you changed was the video driver, and the version you had was before the recent changes, then it seems reasonable to assume the problem isn't related to the recent changes, so older versions from SVN probably won't help.

It's odd that it's only on the Production screen when switching between categories though... It's definitely not happening on the Research screen (which has category buttons), or when just the sidepanel is open (maybe switching between explored systems via the droplist) ?

If you can't debug after the crash, you could add a few extra lines of:

Logger().debugStream() << "some debug text"

to the code, and set the debug level to DEBUG in config.xml, then check the console output to see how far you get before the crash happens.

I have updated to the latest SVN and the newest version of my graphics driver.
Upgrading to the newest SVN version didn't seem to have any effect. I'm having some problems selecting a planet in the production screen, but from the bugs I've seen listed on the bugtracker this doesn't have anything to do with the particular bug I'm struggling with.
Upgrading my graphics card did have a significant affect however. The bug is now much less common and sometimes it actually manages to recover. This makes me think it may be mostly or entirely a problem with the driver, but I still want to look into it a bit more.
I ran it under Valgrind and could not reproduce the bug. While this is not really conclusive, it is troubling...
My attempts to trap it with gdb have all failed so far... The only time I have managed to catch it in progress was in a graphics thread in nvopengl.so or something like that. Which either means it is a driver problem, or possibly I just didn't catch it in the right place...

Sorry for the long post, but this bug is quite frustrating...