SVN revision 12994 - Crashing

[eracc]

Hi y'all,

I just had a crash while playing this recent checkout from SVN. Here is what was left in the xterm window following the crash:

Code: Select all

Loading a starsystem
Loading Star System Crucible/Cardell
 Next To: Crucible/Stirling
 Next To: Crucible/Everett
 Next To: Crucible/Cephid_17
 Next To: Crucible/17-ar
 Next To: Crucible/Maat
 Next To: Redemption/Bernards_star
Unit is the active player character...changing scene graph
Crucible/Everett  IS NOW AT  Crucible/Cardell
Crucible
Launching bases for Crucible/Cardell
Assumed exit on unit Diplomatic_Center(if not quitting, report error)
Traceback (most recent call last):
  File "/home/gene/VegaStrike/data/modules/missions/privateer.py", line 34, in Execute
    i.Execute()
  File "/home/gene/VegaStrike/data/modules/random_encounters.py", line 299, in Execute
    un = self.decideMode ()
  File "/home/gene/VegaStrike/data/modules/random_encounters.py", line 267, in decideMode
    if (self.DifferentSystemP()):
  File "/home/gene/VegaStrike/data/modules/random_encounters.py", line 283, in DifferentSystemP
    self.NewSystemHousekeeping(self.cur.lastsys,cursys)
  File "/home/gene/VegaStrike/data/modules/random_encounters.py", line 78, in NewSystemHousekeeping
    fg_util.launchBases(newsystem)
  File "/home/gene/VegaStrike/data/modules/fg_util.py", line 627, in launchBases
    launchBase(fg[0],fg[1],fac,sys,sig_units,shipcount)
  File "/home/gene/VegaStrike/data/modules/fg_util.py", line 613, in launchBase
    launchSingleBase (type,faction,sig_units[which])
  File "/home/gene/VegaStrike/data/modules/fg_util.py", line 586, in launchSingleBase
    launchBaseOrbit (type,faction,sig.Position(),radpct,speed,sig)
  File "/home/gene/VegaStrike/data/modules/fg_util.py", line 568, in launchBaseOrbit
    bas=VS.launch("Base",type,faction,"unit","default",1,1,Vector.Add(loc,R),'')
RuntimeError: Cannot open "techniques/default.technique"
terminate called after throwing an instance of 'boost::python::error_already_set'
Aborted

If you need to see anything else, let me know.

[pheonixstorm]

Looks like it couldn't find default.technique (which doesn't actually exsist) and died. Klauss???

Post your config setting for your video. Shader, detail, etc. Maybe we can figure out what setting is looking for default and shoo people away from it.. unless they all use it.

[CLoneWolf]

I did a checkout and recompiled these days, while trying to find answers to sebaspons' questions, and had exactly the same issue: all fine until I jumped to a system containing a diplomatic center.

I assumed it was because of my "No Shader" setting (Chuck don't kill me, I know my jurassic PC is way below par
), and tried changing it via the vssetup, and all worked fine; even later, when reverting to "No Shader" again via the vssetup, all worked; so I think it's a rogue legacy setting in the vegastrike.config file and not a lack in the current code/data/config set.

Other notes: even with the "No Shader" setting, the diplomatic center kills my FPS down to 7, with an eye perception coherent with the HUD. The upside instead is for busy scenes that bring my FPS down to the 20's: with earlier releases, my eyes counted less FPS than the HUD said, it was very noticeable; now, when the HUD says 20, it's still as smooth as a real 20 should be.

Meanwhile I've also kept my system updated (kernel, drivers, task switch settings) so I can't really say what drove that change the most.

[Turbo]

I couldn't get the SVN game to load because it seemed to have a problem with every data field in units.csv. So I created a brand new folder, checked out Win32 from SVN (rev 13000), and ran setup.exe then vegastrike.ese. It starts to load the game (no music and a red blank screen), then freezes. STDERR.TXT looks like this:

Code: Select all

GOT SUBDIR ARG = 
Found data in C:\Program Files\Vegastrike\bin/../data
Using C:\Program Files\Vegastrike\data as data directory
USING HOMEDIR : C:\Program Files\Vegastrike\data/.vegastrike As the home directory 
CONFIGFILE - No config found in home : C:\Program Files\Vegastrike\data/.vegastrike/vegastrike.config
CONFIGFILE - No home config file found, using datadir config file : C:\Program Files\Vegastrike\data/vegastrike.config
MISSION_NAME is empty using : main_menu.mission
Attempt to call ReadFull on a bad file units.csv -1 C:\Program Files\Vegastrike\data/units/units.csv
'import site' failed; use -v for traceback
running import sys
print sys.path
sys.path = [r"C:\Program Files\Vegastrike\data\modules\builtin",r"C:\Program Files\Vegastrike\data\modules\quests",r"C:\Program Files\Vegastrike\data\modules\missions",r"C:\Program Files\Vegastrike\data\modules\ai",r"C:\Program Files\Vegastrike\data\modules",r"C:\Program Files\Vegastrike\data\bases"]
testing VS randomrunning import sys
print sys.path
  Initializing renderer...
Min (0.000000, 0.000000, 0.000000) Max(0.000000, 0.000000, 0.000000) MinLumin 1.000000, MaxLumin 1.000000Read In Star Count 0 used: 2000
Min (0.000000, 0.000000, 0.000000) Max(0.000000, 0.000000, 0.000000) MinLumin 1.000000, MaxLumin 1.000000Read In Star Count 0 used: 38
Loading a starsystem
Loading Star System Special/Empty
FOUND MODIFICATION = player FOR PLAYER #0
Compilation of technique fixed failed... trying fixed5
Cause: Error compiling program vp:"fixed8" fp:"fixed"
Compilation of technique fixed5 failed... trying fixedgl
Cause: Error compiling program vp:"fixed5" fp:"fixed"
Compilation of technique 5_ps3.0/default failed... trying ../4_ps2.0/default
Cause: Error compiling program vp:"zwrite" fp:"zwrite"
Compilation of technique ../4_ps2.0/default failed... trying fixed
Cause: Error compiling program vp:"zwrite" fp:"zwrite"
Compilation of technique 5_ps3.0/default_simple failed... trying ../2_ps1.4/default_simple
Cause: Error compiling program vp:"zwrite" fp:"zwrite"
Compilation of technique ../2_ps1.4/default_simple failed... trying fixedgl
Cause: Error compiling program vp:"fixed5" fp:"fixed_simple"

[shenle]

Uh oh... again the red screen. That is what I get on my win7/64 system just before I get the runtime error (see the other thread). Unfortunately I cannot replicate this on any other computer I have access to. And even the same machine under Linux runs VS fine - so I'm willing to bet that in my case it's something about either the video drivers (though I've tried many revisions, all behave the same) or the way the 64-bit VC++ runtime behaves. I was unable to compile a real debug version and get serious about finding what the error is about.

[pheonixstorm]

Turbo, post your windows flavor (and bit) and dxdiag please. Would be nice if we could track this one down and kill it...

[eracc]

Looks like it couldn't find default.technique (which doesn't actually exsist) and died. Klauss???

Post your config setting for your video. Shader, detail, etc. Maybe we can figure out what setting is looking for default and shoo people away from it.. unless they all use it.

Sorry, for the delay. I have attached my entire vegastrike.config file as created by vssetup. Maybe someone will see a setting I should try changing.

[eracc]

For the record, I know I "need" a better video card than this 7200 GS for the game to run better. Unfortunately eating, paying the rent and paying the utilities still appears to be a priority for me. I must not be a "hard core" gamer.

[pheonixstorm]

is that an add-on card or onboard video?

Anyway. Haven't looked at the config yet but will try to in the next day or so.

[shenle]

For the record, I know I "need" a better video card than this 7200 GS for the game to run better. Unfortunately eating, paying the rent and paying the utilities still appears to be a priority for me. I must not be a "hard core" gamer.

I am running vegastrike (or at least trying to) on a Radeon 3200 which is in many respects worse than the 7200. Don't worry about it - I too bought the cheapest laptop they had at Best Buy... It's good enough to run Oolite and JJFFE on it though

[eracc]

is that an add-on card or onboard video?

Anyway. Haven't looked at the config yet but will try to in the next day or so.

This is an add-in card from EVGA (PDF Document): http://www.evga.com/products/pdf/512-P2-N430.pdf

[eracc]

For the record, I know I "need" a better video card than this 7200 GS for the game to run better. Unfortunately eating, paying the rent and paying the utilities still appears to be a priority for me. I must not be a "hard core" gamer.

I am running vegastrike (or at least trying to) on a Radeon 3200 which is in many respects worse than the 7200. Don't worry about it - I too bought the cheapest laptop they had at Best Buy... It's good enough to run Oolite and JJFFE on it though

Heh, well, I build computers (I do not buy from Best Buy. ) and I know what are considered "good" gaming adapters. I just cannot justify the $300 to $500 to plunk down on just the video card for my own PC. The typical custom gamer box my company builds usually has a $400+ video card in it. On some systems that is roughly one third to one half of the cost of the entire system. Admittedly, since I moved the company to a Linux, Unix and eComStation only business we have not built many "gamer" boxes. Most gamers are slaves to Microsoft. I will not build Microsoft boxes any more, the market is too full of folks that do that already.

[pheonixstorm]

Heh, well, I build computers (I do not buy from Best Buy. ) and I know what are considered "good" gaming adapters. I just cannot justify the $300 to $500 to plunk down on just the video card for my own PC. The typical custom gamer box my company builds usually has a $400+ video card in it. On some systems that is roughly one third to one half of the cost of the entire system. Admittedly, since I moved the company to a Linux, Unix and eComStation only business we have not built many "gamer" boxes. Most gamers are slaves to Microsoft. I will not build Microsoft boxes any more, the market is too full of folks that do that already.

You can always use ReactOS in place of windows. Its still inthe early stages of development but its a nice open source windows clone. ReactOS They have a live cd you can probably check out. Still looks like win2k but eventually it might look more like xp.. who knows when though...

[eracc]

You can always use ReactOS in place of windows. Its still inthe early stages of development but its a nice open source windows clone. ReactOS They have a live cd you can probably check out. Still looks like win2k but eventually it might look more like xp.. who knows when though...

Heh, yeah, I think ReactOS is an interesting idea that may work well in the long run and may not. For the time being it is not quite ready for "prime time" use.

Code: Select all

Please bear in mind that ReactOS 0.3.12 is still in alpha stage, meaning it is not feature-complete and is not recommended for everyday use.

[shenle]

Heh, well, I build computers (I do not buy from Best Buy. )

I build my own, just not the laptops
(I did rebuild the cheapo laptop though - upgraded the CPU from single to dual-core; that was a complete tear-down/reassembly)

[klauss]

I finally found the problem. I just grepped the MESHES (bfxm files) for "default", and it so happens that the diplomatic center forces the use of the default technique.

Go figure.

I have two choices: create a default technique, or modify the bfxm.

We should really get our act together on this regard: what to do? having all technique names on all technique folders is a maintainance drag, but makes us able to have bfxm files that specify their own technique (which adapts itself to detail settings).

I think we should have a default.technique in every technique folder, but I just don't like the maintainance overhead that implies, and IIRC, svn doesn't allow us to commit symlinks (at least they won't work in windows... will they?)

[pheonixstorm]

Ive only tried them in linux.. well as those annoying windows 7 commercials go: To the google! (stupid ms clouds... grid computing FTW) Plan9 Plan9 Plan9!

[charlieg]

Can't you just simply interpret 'default' as meaning 'match any'?

Then modellers can create models that will look the same regardless of shader settings, and can then later update and optimize them to work with different shader settings.

Or am I making a pig's ear of that? My input is probably a bit superfluous here.

[klauss]

Empty ("" or not specifying anything) already means the configured default. But if we were to have it my way, we would never use the default.

I imagine the author of the BFXM wanted specifically that technique, the one called "default" at the time.

It must have been one of the earliest models using the techniques feature, so it makes some sense we didn't really know what we were doing

In any case, technique names specify the "kind of effect" the model is looking for, rather than exactly how to get it. The config specifies which technique folder to use, and each of those folders would have a different implementation of the same effect, with varying levels of sophistication tailored for the various GPUs.

So, rather than default and whatnot, the technique would be something like "cinemut", "fireglass", "earth", "gas_giant", "rocky", etc... rendering techniques suited for various kinds of objects:

• cinemut = supershader for most solid materials
• fireglass = supershader for glass and other transparent materials
• earth = planets with earth-like atmosphere, which means all terrestrial planets of any kind (rocky ground, seas, any kind of atmosphere, clouds, etc...)
• gas_giant = gas giant planets, which are balls of clouds.
• rocky = planets with rocky ground and no atmosphere.
• etc

So, for onboard GPUs, the techniques would render a very crude approximation of the effect, and for highend GPUs, the actual .technique file would reference other, more complex shaders, so choosing a shader setting merely amounts to choosing a different techniques folder.

But since some techniques really have no throttling (no way to simplify them any more, or perhaps they're simple enough already), many folders have mere copies of other .technique files, and that's why I'd like to use symlinks: having to maintain copies of technique files spread all over is just messy and prone to synchronization issues, when someone changes one copy and forgets to change the other.

I've read how SVN is supposed to treat symlinks, and even forgetting the fact that there's a bug in SVN that makes it crash with symlinks in windows, the way it handles them wouldn't work in windows.

So I guess I'll implement some kind of explicit symlinking on the XML level, kind of <technique reference="some other file.xml">.

[charlieg]

Symlinks are not a cross platform solution.

I think it might be better to consider only storing the actual implementations in the folder hierarchy (i.e. forget the symlinks) and have the .technique file store a reference to the substitute technique instead. That should be pretty straightforward to implement.

[safemode]

Why couldn't you use the vs method of overloading the path so that each request for a technique first looks in the directory it does now, but you have an additional path appended that's the same for everything (a common directory) that contains a "default" or shared techniques so that if the required named technique is not found in the first path it is looked for in the second? Similar to how we retrieve images when a bxfm requests a texture name.

Why is this not possible? This would remove the need for symlinks or copying the same files for all units and such.

Say all bases have a technique that's shared but it's not shared in ships or planets and such. So you have a directory in /data for /techniques/bases/. same for ships and planets. The top level /techniques can have shared techniques for everything, and inner directories would have group shared. Then individual units can supply their own very specific techniques. Then the bxfm requests techniques the same as it does images, and we search first the unit directory, then group directory for that unit (supplied by bxfm possibly since the game doesn't distinct between bases and ships and such on this level) and finally if it can't find it in there it just uses the top level shared techniques.

[klauss]

Why couldn't you use the vs method of overloading the path so that each request for a technique first looks in the directory it does now, but you have an additional path appended that's the same for everything (a common directory) that contains a "default" or shared techniques so that if the required named technique is not found in the first path it is looked for in the second? Similar to how we retrieve images when a bxfm requests a texture name.

It already does something like that...
...I'm not sure why it doesn't work with this particular case, there was a reason though.

...Memory fleeting...

Say all bases have a technique that's shared but it's not shared in ships or planets and such. So you have a directory in /data for /techniques/bases/. same for ships and planets. The top level /techniques can have shared techniques for everything, and inner directories would have group shared. Then individual units can supply their own very specific techniques. Then the bxfm requests techniques the same as it does images, and we search first the unit directory, then group directory for that unit (supplied by bxfm possibly since the game doesn't distinct between bases and ships and such on this level) and finally if it can't find it in there it just uses the top level shared techniques.

In any case, I wouldn't want to hardcode the search progression on the engine.

Rather have it controlled by an explicit chain of fallbacks/includes on the xml files.

Matter of taste