[Gc] A zombie using GC 6.2 - Solution

Alec Orr Alec.Orr at wbemsolutions.com
Mon Feb 23 07:56:31 PST 2004


Folks:

Disabling leak detection seems to have solved this problem.

http://www.hpl.hp.com/personal/Hans_Boehm/gc/leak.html
See "Simplified leak detection under Linux" Step 1

Our Linux target was multithreaded.  Answer was there the whole time.

-Alec




Alec Orr wrote:
> RH8 Linux
> GC6.2 --enable-full-debug --enable-threads=posix --enable-static=no 
> --enable-gc-assertions
> 
> Folks:
> 
> I am still looking into this one myself, but would see if anyone else 
> has seen similar behavior.
> 
> On occasion I get a zombie process running a module hooked into the 
> debug build of GC 6.2 (without --enable-full-debug works fine).  The 
> process cannot be recovered (using gdb or otherwise).  Below is a trace 
> from gdb.  In my experience, this is due to heap corruption. (we can 
> switch heap models to DMALLOC - no errors reported).  Anyone else run 
> into this, or can suggest a way to find this bugger (unfortunately this 
> is intermittant and the object address is almost never the same).  We do 
> use pthreads with no hooks into pthread_* functions.
> 
> Thanks for any suggestions,
> Alec Orr
> 
> ------
> 
> Initiating full world-stop collection 73 after 35948 allocd bytes
> --> Marking for collection 73 after 35948 allocd bytes + 2044 wasted bytes
> Collection 72 finished ---> heapsize = 237568 bytes
> World-stopped marking took 10 msecs
> Complete collection took 10 msecs
> Leaked composite object at start: 0x8092400, appr. length: 1024
> Leaked composite object at start: 0x808bf98, appr. length: 8
> Leaked composite object at start: 0x808bfa8, appr. length: 8
> Leaked composite object at start: 0x8069fc0, appr. length: 64
> Leaked composite object at 0x8065c08 (unknown:0, sz=40)
>         Call chain at allocation:
> Cannot find user-level thread for LWP 22920: generic error
> 
> 
> _______________________________________________
> Gc mailing list
> Gc at linux.hpl.hp.com
> http://www.hpl.hp.com/hosted/linux/mail-archives/gc/
> 
> 
> 




More information about the Gc mailing list