1354 Windows Client Crash on recipe

Case number:845829-2003617
Topic:Crash/Hang
Opened by:actiasluna
Status:Open
Type:Bug
Opened on:Wednesday, March 22, 2017 - 12:21
Last modified:Tuesday, March 28, 2017 - 11:41

Running DRemixW

Crashed for unknown reason and don't see error in log files.

(Mac machine running WIndows 10 on external HD)

AttachmentSize
log.txt1.24 MB
(Wed, 03/22/2017 - 12:21  |  10 comments)


actiasluna's picture
User offline. Last seen 5 hours 58 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Don't see obvious reason. Suspect Thread 4
Attaching correct log.txt (maybe that's why LOL)
and scriptlog

actiasluna's picture
User offline. Last seen 5 hours 58 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Drat! this is a 1355 crash not 1354. (good morning) LOL Can anyone change the original header?

Joined: 09/24/2012
Groups: Go Science

This kind of crash may occur when the recipe calls for an nonexistent function or attribute or variable, but when I look at the scriptlog, it doesn't seem to be the case here (and I never saw this in a tvdl recipe). Moreover, I went to check in the tvdl code around the lines involved and I didn't find anything like this.
And this recipe would have crashed earlier with other puzzles.

So I think the problem is not in the recipe.

actiasluna's picture
User offline. Last seen 5 hours 58 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Thanks Bruno. Just had another crash with another recipe...Banded Worm Pairs, alternate test on the Mac
This time Thread 10 crashed, mac crashlog is the only one I can see "clues" in...
excerpt:

Thread 10 crashed with X86 Thread State (32-bit):
eax: 0x00000000 ebx: 0x00000003 ecx: 0xb048bcdc edx: 0x00000000
edi: 0xb048e000 esi: 0x00000006 ebp: 0xb048bd08 esp: 0xb048bcdc
ss: 0x00000023 efl: 0x00000296 eip: 0x9f8b749a cs: 0x0000000b
ds: 0x00000023 es: 0x00000023 fs: 0x00000023 gs: 0x0000000f
cr2: 0xa5326320

Logical CPU: 0
Error Code: 0x00080148
Trap Number: 132

I honestly don't know if this is one issue (Thread 4 and thread 10 the common threads which crash, which could be several separate issues)... or many. Recipes that run on other puzzles are crash-prone (I know that some of the group recipes I run are old and known to be prone to crashing so I don't report those) ... but recipes that I know to be stable like DRremixW and Banded Worm Pairs, to my mind at least should NOT be crashing, but in these cases are.

I gather that it is a multiple not a single cause. Just off the top of my head and in no particular order they might be:
- recipe authors/editors haven't updated to latest LUA commands
- Bug or typo somewhere in code in the game version(s)
- End user doing something bonkers and not realizing it
- Something to do with network connectivity
- Something to do with how foldit program deals with memory
- The weather
- How the player "holds their mouth" when playing
- developer error of some sort
- Server glitch
- Unreproducible intermittent hardware error somewhere in the whole big network between my machine and foldit central
- The phase of the moon
- A butterfly on the foothills around Mt. Rainier flapping its wings just so... :-)

...so while I'm half serious, the point is that I know it's not a simple thing to debug several platform versions of a game, and even tougher when there's a small staff to do the work and figure out who's not holding the kite string tight enough. :-P

I do wonder (uneducated guess time) that this has something to do with how foldit (both the windows and mac client) handles memory. It seems that after a (somewhat random) period of time, things just go wrong. It can be one day, many days... but often while playing foldit the game will just lock up and the finder freezes. Was able to extricate myself from the latest apparent finder freeze by using the option-CMD-esc key combo and force quitting the applications then force restarting the finder, several times (which I now will use whenever this happens)... but something isn't quite right.

RAM tested fine... HD and Directory structure, permissions, etc. all test fine.
So I just can't figure out what is going on.

The recent devprev does this significantly LESS than prior versions of foldit, which is an improvement.
But it still does crash too much, and many recipes are still not reliable... so more needed. I'm hoping that with the new OS on the foldit Mac test machine and new Xcode that perhaps there will be further improvements!

actiasluna's picture
User offline. Last seen 5 hours 58 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Reproduced Thread 10 Crash

Working in another solid recipe, Softwalker.

Was working megs 73-74 when or just before crash.

This is in devprev, latest update, on Mac.

Only clue I can see is in the first log, the mac crash log.

Hope this helps!

actiasluna's picture
User offline. Last seen 5 hours 58 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Here might lie some clues... another game that has had these kinds of issues...So this is NOT a problem unique to Foldit.
(seems to mirror the kind of errors I'm experiencing.)

http://forum.kerbalspaceprogram.com/index.php?/topic/105378-10-constant-crashing-on-os-x/

and is the bottom line of this part of my mac crashlog any help?

Crashed Thread: 10

Exception Type: EXC_CRASH (SIGABRT)
Exception Codes: 0x0000000000000000, 0x0000000000000000
Exception Note: EXC_CORPSE_NOTIFY

Application Specific Information:
abort() called
*** error for object 0x6c22a860: double free

NinjaGreg's picture
User offline. Last seen 1 day 8 hours ago. Offline
Joined: 05/21/2010
Groups: Go Science

Looking at actiasluna's first log (Repro1355Thread10MacCrashlog.txt): I see that thread 10 is complaining about the memory allocator trying to free a memory chunk that was cast off from a string extension operation, and discovering that the memory chunk that it is releasing has been overrun (you can set up the memory allocator to put "guard bands" around each allocated chunk with a known value in them, which can be checked when the chunk is requested to be freed). Usually caused by an indexing error (also usually an off-by-one error) in an array, in this case a string array. Would need the actual code to see what is going on. It was applying a filter and computing the score further up the call stack.

Just an FYI, nothing we can do about it from the users side.

actiasluna's picture
User offline. Last seen 5 hours 58 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Thanks NinjaGreg. Had the same error maybe exactly on puzzle 1358 so maybe "we" are getting closer to something that will help pinpoint the problem. Thread 10 and Thread 4 are the most common crashers (and sometimes Thread 0) so maybe this is something the dev can work out.

from the 1368 log (crossposted)

Crashed Thread: 10

Exception Type: EXC_CRASH (SIGABRT)
Exception Codes: 0x0000000000000000, 0x0000000000000000
Exception Note: EXC_CORPSE_NOTIFY

Application Specific Information:
abort() called
*** error for object 0x6af54e20: double free

--
Thread 10 Crashed:
0 libsystem_kernel.dylib 0x9f8b749a __pthread_kill + 10
1 libsystem_pthread.dylib 0x9f99aec3 pthread_kill + 90
2 libsystem_c.dylib 0x9f817a44 abort + 141
3 libsystem_malloc.dylib 0x9f9192ae szone_error + 448
4 libsystem_malloc.dylib 0x9f922650 free_tiny_botch + 100
5 libsystem_malloc.dylib 0x9f9225cd free_tiny + 653
6 libsystem_malloc.dylib 0x9f90ec70 szone_free_definite_size + 127
7 libsystem_malloc.dylib 0x9f91bd08 default_zone_free_definite_size + 62
8 libsystem_malloc.dylib 0x9f90e85d free + 391

actiasluna's picture
User offline. Last seen 5 hours 58 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Another few links that might help figure out why this is happening (if the logic translates to foldit's code):

https://github.com/svn-all-fast-export/svn2git/issues/5
https://github.com/svn-all-fast-export/svn2git/commit/7d919c1c988a7eb50d5a0f098af6b194363c5c68

actiasluna's picture
User offline. Last seen 5 hours 58 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters
Sitemap

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons