Quote:
Originally Posted by mazex
I have also experienced the online crash after 20-30 minutes twice (have only had time to fly online twice on the new patch so 100% there). I have mailed the dmp-files with a detailed description to Ilya - but I did notice one interesting thing before killing "launcher.exe" the last time... The 32-bit process took 3Gb of memory. That should be "against the wall" even if it's compiled to be large address aware. Feels like a memory leak to me as an old C++ programmer...
Anyone else that has noticed this?
By the way, I had 20 dmp-files from the same minute it crashed the last time.
|
Got an out of memory again today when changing skins... See my mail below to Ilya (including my mail from yesterday) - reposting it here if anyone else has gotten the same problem (and so you can have an eye open for it!)
Quote:
Hi again Ilya!
I have now repeated the first bug I reported yesterday and can confirm that there is a memory leak regarding skin loading. Thats good news at least!
I was flying on Repka again. Flew a mission in the Spitfire II, shot down a 109, managed to damage my engine trying to catch another with to little radiator and went back and landed. I then tried spawning but got the message from the server "To many planes of this type". Pressed escape which unfortunately does not give you the option to "exit this flight" - so you really leave that Spit II running? Anyway - as there is not "exit this flight" available I guess the server has to "clean out" those planes when a user from the "escape menu" selects create new flight? So I did create a new Spit 1a plane at the same airfield and realised it had that funky "JUNGLE" skin I talk about in my mail below from yesterday. So i did press escape again and set out to "create" a new Spitfire 1a but without the "JUNGLE" skin. So when I klicked the "Plane" button to customize my plane I kept an eye at my Logitech LCD Display where I can see the current CPU and memory usage. I had kept a watchful eye on that as I had the memory leak problems described in "crash number 2" below where the game froze with 3Gb memory used.
So - just after joining the Repka server I checked that launcher.exe used around 1.2Gb of memory, after landing the Spitfire II after the first mission it used 1.7Gb but it decreased a bit back to 1.6... Fine I thought. Then when I changed from the "JUNGLE" skin to the default "1939B" skin I looked at the memory usage (which was 50% then of my total 8Gb). Imediately after selecting the 1939B skin it started going up to 55%. Changed to 1939A and it continued rising. Then I realized it was rising at 1% per second anyway and when it came to something like 75% (very soon) I got a "launcher.exe" has stopped responding error from the OS.
So - it seems you must look at the skin loading routine. There must be some memory that is not released when selecting a new skin and there is something that crasches after a while and just eats all memory in some situations. When I try it in single player after NOT having flown a mission I can not reproduce it though so it seems like you have to have flown before for it to happen!
Good luck /mazex
2011/10/23 xxxx <xxx.xxx@gmail.com>:
> Hi Ilya!
>
> Unfortunately I had two CTD:s while playing online on the Repka 1.11
> Steppes server tonight on the latest official patch (Steam cache
> verifyed etc - computer extremely stable with all latest drivers). The
> first time it happened after crashing and then when selecting a new
> plane (Spitfire 1a) I tried to change the skin. I found a skin called
> "JUNGLE" in the list and selected it (maybe a user skin downloaded
> online?). Immediately after selecting it I got a "Launcher.exe has
> stopped responding", and when I had killed the process I saw a dump
> file that was 0 kb... I enclose the .dds file I found in my cache
> folder that contains the skin. After restarting I could select it
> though without crash...
>
> OK, so I did a second attempt after rebooting my computer to be sure
> no stray processes where running. After taking off I played two
> missions - one for 10 minutes - shot down one me109 in my Spitfire 1a
> and then selected the Spifire II (in limited supply at that server).
> After taking off I battled with one 109 and shot him up badly by he
> managed to crach on his own field (the one at ~E4). After that I saw
> two 109:s low over the edge of the field so I turned against them (I
> was right over their field by then). Then the game "froze" with the
> sound of my Merlin at full speed looping all over, CPU at 0%. When I
> CTRL+ALT+DEL to start the task manager to kill "launcher.exe" I saw
> that the launcher.exe (32-bit) process was taking 3070Mb of memory...
> Tried taking a screenshot of the task manager before killing the
> process but it seems the focus may have been on the dead process in
> some way so the screen shot was just blank. But I am sure it was 3Gb.
>
> So... That must be "against the wall", even if the game is compiled to
> be large adress aware (/LARGEADDRESSAWARE compiler directive has a
> limit of 3Gb?). Feels like a memory leak to me as an old C++
> programmer?
>
> I then looked in the cppcrash folder and found 20! .dmp-files from the
> time of the crash (I had cleaned it after the previous crash)... I
> include an image showing the folder and one of the files.
>
> Good luck with the game - it's really getting good now!
>
> Regards /xxxx (mazex at the 1C forum and online)
>
> My system:
>
> CPU: Intel i7 2600k @ 4.5 Ghz
> GPU: Asus GTX580 Direct CU II with the latest beta drivers 285.38
> Mother Board: Asus P8P67 Pro
> Memory: G-Skill 8GB DDR3 1600 Mhz
> Disks: Intel X25-M G2 SSD (OS) + Raptor 150 (Steam/Games) + 1TB WD (Extra)
> PSU: Corsair CMPSU-HX 750W
> Sound card: Creative X-Fi Fatality Pro (PCI) with the latest 2.18.15 driver
> OS: Windows 7 x64
> TrackIR 4
> Logitech G940 joystick
>
|
__________________
i7 2600k @ 4.5 | GTX580 1.5GB (latest drivers) | P8Z77-V Pro MB | 8GB DDR3 1600 Mhz | SSD (OS) + Raptor 150 (Games) + 1TB WD (Extra) | X-Fi Fatality Pro (PCI) | Windows 7 x64 | TrackIR 4 | G940 Hotas
Last edited by mazex; 10-23-2011 at 09:20 AM.
|