1.70, crash when dying or purchasing new ship and more

ArkanoiD · Post by **ArkanoiD** » Wed Dec 12, 2007 12:42 am

RedHat Linux, repeatable 100%, 2Gb RAM, p4-3GHz, Nvidia GoFx5200 32Mb VRAM

Post mortem:
2007-12-10 22:50:16.993 oolite[3833] [shipEntity.bug.subEntityRetainUnderflow]: Subentity died while still in subentity list! This is bad. Leaking subentity list to avoid crash.

Trying to buy a new ship:
2007-12-10 23:36:56.811 oolite[4094] *** NSTimer ignoring exception 'NSInvalidArgumentException' (reason 'Tried to init array with nil object') raised during posting of timer with target 0x8a446f8 and selector 'performGameTick:'
2007-12-10 23:37:35.188 oolite[4094] [shipEntity.bug.subEntityRetainUnderflow]: Subentity died while still in subentity list! This is bad. Leaking subentity list to avoid crash.

There are also random crashes at start:

2007-12-10 22:16:55.737 oolite[3763] [texture.load.png.warning]: ***** A PNG loading warning occurred for /home/ark/.Oolite/AddOns/all-stars.oxp/Textures/starhawk_auv.png: Profile size field missing from iCCP chunk
2007-12-10 22:17:20.197 oolite[3763] [texture.load.png.warning]: ***** A PNG loading warning occurred for /home/ark/.Oolite/AddOns/all-stars.oxp/Textures/starbelly2.png: Incomplete compressed datastream in iCCP chunk
2007-12-10 22:17:20.198 oolite[3763] [texture.load.png.warning]: ***** A PNG loading warning occurred for /home/ark/.Oolite/AddOns/all-stars.oxp/Textures/starbelly2.png: Profile size field missing from iCCP chunk
2007-12-10 22:17:22.675 oolite[3763] [texture.load.png.warning]: ***** A PNG loading warning occurred for /home/ark/.Oolite/AddOns/all-stars.oxp/Textures/aafreighter_auv.png: Incomplete compressed datastream in iCCP chunk
2007-12-10 22:17:22.675 oolite[3763] [texture.load.png.warning]: ***** A PNG loading warning occurred for /home/ark/.Oolite/AddOns/all-stars.oxp/Textures/aafreighter_auv.png: Profile size field missing from iCCP chunk
2007-12-10 22:17:27.188 oolite[3763] [entity.bug]: -[Entity subEntityReallyDied:] called for non-ship entity 0xb4b00020 by 0xa89cd40
2007-12-10 22:17:27.189 oolite[3763] [entity.bug]: -[Entity subEntityReallyDied:] called for non-ship entity 0xb4b00020 by 0xa8d40c8

and a stiuation when oolite is either crashes or ends up "running" without any controls and objects in the ooniverse:

2007-12-10 22:29:45.263 oolite[3763] *** NSTimer ignoring exception 'NSInvalidArgumentException' (reason 'Tried to init array with nil object') raised during posting of timer with target 0x9b4f398 and selector 'performGameTick:'
2007-12-10 22:30:09.728 oolite[3763] [unclassified.MyOpenGLView]: Creating a new surface of 1280 x 800
2007-12-10 22:30:11.808 oolite[3763] [unclassified.MyOpenGLView]: Creating a new surface of 1280 x 800

Commander McLane · Post by **Commander McLane** » Wed Dec 12, 2007 6:08 am

The ship-buying crash is there since 1.69.1 and has to do with your current ship having subentities.

@Ahruman: I would guess, as both are involving playership's subentities, it's related to bug 011686, probably using the same underlying buggy routine, and crashing in that.

ArkanoiD · Post by **ArkanoiD** » Fri Dec 14, 2007 12:28 am

Commander McLane wrote:
The ship-buying crash is there since 1.69.1 and has to do with your current ship having subentities.

@Ahruman: I would guess, as both are involving playership's subentities, it's related to bug 011686, probably using the same underlying buggy routine, and crashing in that.

Still there in 1.71-1271
found a Swift for sale, so i wait for the issue to be resolved

ArkanoiD · Post by **ArkanoiD** » Tue Dec 18, 2007 4:11 am

Any chances this bug will get fixed soon?

Post by **Cmdr James** » Fri Dec 28, 2007 12:15 am

What ship do you have that you are trying to sell (the one you have currently, not the one you want to buy).

I have seen this problem with the urutu III. In this case it was caused by the first 3 of the 27 subentities being nil. The first 3 were the exhaust plumes, the next 24 are normal ship parts.

If you can get me the details, and maybe a save game file, I can do some more analysis and see if I can actually find the underlying problem. No promises, but Ill have a shot.

It also appears to be an exhaust problem with the Velocity150. I guess the exhausts are being released somewhere and the other subentities are not. I have looked quite a bit, and so far I cannot see anywhere obvious that this could be happening.

ArkanoiD · Post by **ArkanoiD** » Fri Dec 28, 2007 3:00 am

Kukri mkII dragster (kukrimk2-player, racers.oxp).

I am going to sell it because i've got nice offer, the faster ship, Swift. But i am afraid it is uber.

Cmdr James wrote:
What ship do you have that you are trying to sell (the one you have currently, not the one you want to buy).

Commander McLane · Post by **Commander McLane** » Mon Dec 31, 2007 10:46 am

Cmdr James wrote:
What ship do you have that you are trying to sell (the one you have currently, not the one you want to buy).

Nice if you could track it down!

I have the problem with my ship being an Imperial Courier (version 2). I could also produce it by giving me another ship with subentities in my save-file. I think it was a DW-Cobra, but am not sure anymore.

But, anyway, for finding out which ships cause the problem, it's a good idea to give yourself one of the suspicious ships by hacking a save-file of yours and then try to sell it.

Post by **Cmdr James** » Mon Jan 07, 2008 12:23 am

Well, I did some refactoring of setupfromdictionary, so that it is simpler for me to work on, and as far as I can tell I no longer get this problem in my local build. This is kind of weird as I had not actually started looking at the problem. I can buy and sell the kukri, and all other ships that are in the shipyard when I test.

It could be fixed by another_commander as I have updated from his latest source, but I didnt see anything that looked likely in the change set.

I am waiting for some more memory for my MacBook so that I can run Instruments without going into swap. If I can confirm that it is fixed, Ill pm someone with access to svn to change trunk.

I am not too happy with this, as I do not know how I fixed it, Ill try to get some more info on it.

Post by **another_commander** » Mon Jan 07, 2008 8:41 am

I can confirm the existense of the bug in 1.70 on Windows. It had not been fixed in the trunk... up to now. Fix will hit SVN very soon. Looks like Cocoa and GNUstep have slightly different ideas on how memory release should work, as a similar fix was applied not too long ago to smash another type of memory related crash, which seemed to occur on Windows only.

Edit: Hey, what do you know, we have two birds with one stone! Berlios Bug 11686 (crash on respawn player when player ship contains subentities) is also fixed. Tested both with the Kukri MKI and Kukri MKII and they work fine now.

Commander McLane · Post by **Commander McLane** » Mon Jan 07, 2008 11:22 am

another_commander wrote:
Edit: Hey, what do you know, we have two birds with one stone! Berlios Bug 11686 (crash on respawn player when player ship contains subentities) is also fixed. Tested both with the Kukri MKI and Kukri MKII and they work fine now.

As I predicted just a few posts above, if I may say so.

ArkanoiD · Post by **ArkanoiD** » Tue Jan 08, 2008 7:46 pm

rsynced to nightly build, still there

Post by **another_commander** » Tue Jan 08, 2008 10:50 pm

Please send me your save file just before selling the ship. Email is nikbar2004 at yahoo dot com.
I am curious to see if I can reproduce it. It works for me after the fix. Do you still get the same error message at crash time?

ArkanoiD · Post by **ArkanoiD** » Tue Jan 08, 2008 10:59 pm

another_commander wrote:
Please send me your save file just before selling the ship. Email is nikbar2004 at yahoo dot com.
I am curious to see if I can reproduce it. It works for me after the fix. Do you still get the same error message at crash time?

Sent. Diagnostic messages differ slightly this time:

2008-01-08 22:58:08.227 oolite[13731] [entity.bug]: -[Entity subEntityReallyDied:] called for non-ship entity 0xaf4fa010 by 0x11ce0dd0
2008-01-08 22:58:08.228 oolite[13731] [entity.bug]: -[Entity subEntityReallyDied:] called for non-ship entity 0xaf4fa010 by 0x9694248
2008-01-08 22:58:30.575 oolite[13731] [shipEntity.bug.subEntityRetainUnderflow]: Subentity died while still in subentity list! This is bad. Leaking subentity list to avoid crash.
Fatal signal: Segmentation Fault (SDL Parachute Deployed)

Post by **another_commander** » Tue Jan 08, 2008 11:00 pm

Thanks. Please remind me, which ship it is you want to buy? Is it the Swift?

Edit: Arkanoid, I tried your saved game. Works fine here. I was able to buy each and every ship that was in the shipyard, no problem whatsoever. At this point, I am starting to think about differences in GNUstep versions. Which version do you have installed? Also, it would be very helpful if you can let me know exactly what you do to crash it. I just loaded your saved game, and started buying one ship after the other. Do you do anything specific that crashes the game?

Also, can you post what happens if you take the Kukri MKII from your saved game and crash it on the station? Do you get a crash during respawn?

Sorry for the questions storm, but it looks like there is more to this than I thought and all possible information will be required. In the meantime, I am reopening bug 011686 at Berlios, with a note that the bug seems to be now Linux only.

ArkanoiD · Post by **ArkanoiD** » Tue Jan 08, 2008 11:58 pm

Yes, it crashes on respawn as well. I use GNUStep bundled with compiled binaries on rsync repository.