[Trivial Psychic]

Howdy all:

I'd like to express my irritation with a certain SCP bug that seems to be getting more prevalent with each successive test build, since about early February.  The bug has already been posted on Mantis by Goober as bug 165.  For me, the irritation factor is now exceeding that of the old dreaded "Y-targetting" bug.  In my experience, it results when a player is flying a mission with multiple events and sequences, and that mission has been played through most of these events.  After this, once the player tries to replay the mission (either after dying or pressing escape), or completes the mission and enters subspace, the game will crash to desktop, with no error messages.  I call it CTD on Mission Termination, or CoMT for short... although its also known as "Aw CRAP!"  This is more anoying than the "Y" bug, since it prevents the player from advancing to the next mission.  Once, I even had the CTD occur as soon as I saw my ship dying, and it never even loaded the options window.  At first I thought it was my missions, but I just encountered it when playing an Inferno mission for the first time in months (see Inferno forum for more info on that).  I have had occasions when a mission such as this didn't crash, but I had only played the first few seconds before choosing to abort, so that's why a requirement is that some of the mission events be completed before it will trigger.

The first and most obvious suggestion one would make to me, would be to use a debug build and find the error message, but debugs always crash if I try to do anything beyond the lobby (if I even make it THAT far), so that is not an option for me.

I ask that since the coder team is in "Bug Hunt" mode, that you make this particular bug a priority.

In case it matters:

AMD Athlon 800
A7V 100Mhz.
256MB SDRAM-133
ASUS V7100 GF2-MX 32MB AGP
WinME
Latest nVidia drivers for WinME

Build:  Any build for the past month at least.
Command flags:  -glow -spec -jpgtga -fps

Later!

[redsniper]

Originally posted by Trivial Psychic
WinME

Problem found.

[ChronoReverse]

Does it happen with the retail missions?

[Bobboau]

reading the info in mantis, I'm thinking this might be sort of mine, it said that the error was in message_mission_shutdown, I added the following code to that function

Code: [Select]

   for (i=0; i            if (Message_avis[i].anim_data != NULL) {
                    for (j=0; jref_count; j++) {
                            anim_free(Message_avis[i].anim_data);
                    }
            }
            Message_avis[i].anim_data = NULL;
    }



to fix a memory leak, however this is not my code, this was taylor's sudgestion, and code (and I mentioned this in the function), perhaps I copied it incorectly or missed some sort of subtle thing. this was just before Taylor became a member wich is why I added it and not him.
and to make it clear, I'm not placeing blame or anything, more likely than not it was something I didn't do corectly as I just blindly copied the changes.

[StratComm]

Originally posted by redsniper

Problem found.

While I can surely sympathize with the suckyness of ME, it's not the problem.  Happens on my XP box too.

[ShadowDrakken]

Funny, ME was the most stable of the Win9x series for me, but that's unimportant.  I've had this same crash problem in FS2 and FSO both in WinXP.  And it gets worse the more missions played in a single session until some event all of a sudden just dumps the game.  Sometimes when a menu is pulled up (like the death menu, or trying to restart the mission) and other times such as a new ship warping in

As a side note, with FS2 (not FSO) this also happened to me in Win95, Win98, Win98SE and WinME... so don't try and blame it on XP either

Current System Config:
AMD Athlon XP 1800+ (1.533gHz)
1gb PC2100 DDR
VisionTek GF4 Ti4600 128mb AGP 4x
Turtle Beach SantaCruz

The only thing that's remained the same all this time, is that I've always had an AMD CPU.

[Slasher]

Odd, I've encountered something similar in The Babyon Project.  Starting with the mid-March build, I think.  In my version, though, some missions will spontaneously crash while loading while other times they load just fine and I can play them through without error.  Similarly, this event seems to have grown more frequent over the past month so that the majority of the time certain missions crash.  Maybe 1 out of 3 times it will load and play flawlessly.

[WMCoolmon]

I've run into this bug playing the vanilla main campaign + current mediapack using the fs2_openT-20040405 build. (More info in the Recent builds thread for it).

[Goober5000]

It happens all the time, but taylor said he couldn't reproduce it.  He's working on putting a fix together.

[Bobboau]

you know I think I remember getting this while traceing through the code once, wich means it probly won't be too hard to fix (as opposed to the 'Y' bug)

[Trivial Psychic]

I just started playing through the Port again under SCP.  I made it as far as the 3rd mission with the supply depot attack, and was able to do a successful mission restart after losing an objective.  I completed all objectives the 2nd time around, and just as the screen went black after I entered subspace, BANG... CoMT!  As I said in my first post, the mission needs to have some guts to it, and the player needs to be most of the way though for this to occur.

[Col. Fishguts]

I get this error with the 2nd mission from TBP R1, as soon as I end the mission...CoMT.
I get a error message that says somthing along the lines of "FS2_open caused blablabla......while trying to acces Register FFFFFFFF...blablabla" in case that helps.
Isn't FFFFFFFF the highest memory adress there is ?

[taylor]

Everybody with this problem try this new build: 20040423-win32.zip

There is a last minute Windows OpenGL thing I have to fix but it won't cause any crashes, just be aware of some wierd screen updates if you use OGL.  Bobboau did copy the code correctly so this one's all on me.  This same code has been in icculus.org CVS for over a year so I don't know why it's only causing problems in FSO.  Please let me know pretty quickly if this fixes it or doesn't.  If this doesn't fix the bug then I'm going to remove the code in question.  That would open up a bad memory leak but it's better than a CTD.

[Trivial Psychic]

Well Taylor, I tried your build on the same mission as I posted above, and it did make a difference... this time it crashed with an error message:


Ddhelp

Ddhelp has caused an error in
Ddhelp will now close

If you continue to experience problems,
try restarting your computer.


I doubt that'll help much... sorry.

[Slasher]

I didn't even get the error message.  But since my problems almost always spring up during loading I'm wondering if I have a different problem entirely...

[ShadowDrakken]

you mention OpenGL users, but OpenGL isn't even usable... in Nebula missions all you get is a black screen and all you see is glow maps flying around... it's not much fun

[taylor]

Trivial Psychic: I played that mission 17 times without a crash.  I'll keep working on it but misbehaving computers tend to work perfectly when I walk into a room so maybe it's some wierd karma thing.  Anyway, I have a debug build that will save out some info that I need if you want to try that.  When it crashes, just send me (taylor AT icculus.org) your errorlog.txt and the newest DBUG*.txt file.  Hopefully that will be all I need to figure out what is going on.

ShadowDrakken:  Hmm, Neb missions in OGL should work fine with this build.  What video card do you have?

Also, I don't suppose anyone has noticed a change in load times with this build...

[ShadowDrakken]

Taylor: the same one listed in my first post of this thread
I haven't tried this build yet, but the 4/05 build had the nebula problems

[Slasher]

I've downloaded the debug build and will send you the two text files later today, if I can get that far.

[taylor]

D'oh! < hand abruptly stops against forehead >  I should have remembered that ShadowDrakken, sorry 'bout that.  I've got a Ti4200 and it works pretty good for me with the 04/05 build.  I have made a few fixes since then to make it better but I doubt it's anything that would fix the problem you're describing.

I really don't want to keep dumping builds on all of you but I'm down to my 2nd to last option (the last option opens a memory leak).  I still need the bebug info to try and figure out what the real problem is but if someone wants to test a pretty hack send me an e-mail and I'll get you a link to the build.  I'm not going to post the build here and if this last one doesn't fix it then I'm going to open up the memory leak again.  If no one responds within about 6 hours of this post then I'll release a true test build (new lightning arcs(!) and some major OGL bugs fixed) in the recent builds forum with this code removed.  It's the lesser of two evils in this case and I need this bug resolved in order to get the rest of my fixes in CVS for 3.6.