sorry for the dupe..the forum seems to have lost my original post. i’ll attach the crash output this time.
i am having very worrisome random crashes. my patch runs stable for maybe 45 minutes, maybe an hour and then crashes max with the message below. i’m aware of the threads that discuss crashing due to java and pattrstorage; i’ve removed all of that stuff from my patch – still crashing. i got the the point where i even wiped my drive and freshly installed everything. memtest in single user mode reports no errors.
os x 10.6.4
i have a show coming up…please, can anybody help me understand what the problem might be? cv.jit bug? overheating?
Without a patch (and maybe if you had since I am having my own crashing issues on a patch for a show) it is hard to tell.
I would look and see if you can figure out the objects mentioned in the below are and if their is any way to rework them/make them simpler.
0 com.cycling74.MaxMSP 0x0002fb91 getbytes + 67
1 com.cycling74.MaxAPI 0x0451cf23 sysmem_newptr + 31
2 com.cycling74.JitterAPI 0x1ee50067 jit_getbytes + 32
3 com.cycling74.JitterAPI 0x1ee23698 jit_attr_offset_array_get + 144
4 com.cycling74.MaxMSP 0x000ba5fa object_method + 396
5 com.cycling74.MaxAPI 0x0452830a object_method + 156
6 com.cycling74.JitterAPI 0x1ee2da7c jit_object_method + 136
7 com.jmp.cv.jit.shift 0x20f72e08 max_cv_jit_shift_bang(_max_cv_jit_shift*) + 196
8 com.cycling74.JitterAPI 0x1ee6417f max_jit_mop_jit_matrix + 731
ok, i’m pretty sure i know the reason for the crashing now – but not necessarily how to fix it. i have a very bad memory leak that is eating up all of my 4GB of ram until max dumps. i also seeing higher cpu levels than i would expect. could this indicate a scheduling or queuing problem?
i am not using mxj or js, just standard objects, cv.jit and a number of jit.lcd objects for compositing. eventually i will convert the compositing to gl, but i don’t think that is practical now…unless the multiple jit.lcds are the problem. can anyone offer advice?
Yes the crash report looks like a memory leak problem. I ran tests with cv.jit.shift but it doesn’t seem to leak. If you don’t mind sending me a patch that’s causing your problems I can try to see where the problem is.
hi jean-marc, thanks for getting back. i stripped my patch all the way down and i think i may have identified the problem. it is in cv.jit.touches. even running the help patch, you can see that it is eating memory at a rate between .1 and .2 MB a second. since my patch is for an overhead tracking installation you can see why this is a problem. i get maybe a couple hours before a crash, depending on what else the computer is doing.
searching the forums for memory leaks points to issues with the rapid creation of symbols. is that a factor here?