[Mono-dev] Memleak in Array.CreateInstanceImpl?

[Paweł Sakowski]

I'm investigating an apparent memory leak occurring in my Mono 
application. I still haven't concluded the investigation, but I wanted 
to share one suspicious entry in Valgrind's output:

==27640== 116,000 (115,164 direct, 836 indirect) bytes in 14,004 blocks 
are definitely lost in loss record 17 of 19
==27640==    at 0x4005932: calloc (vg_replace_malloc.c:279)
==27640==    by 0x5BABDF: g_malloc0 (in /usr/lib/libglib-2.0.so.0.400.7)
==27640==    by 0x40BC74E: mono_class_bind_generic_parameters 
(reflection.c:8917)
==27640==    by 0x4136C3F: mono_bounded_array_class_get (class.c:3564)
==27640==    by 0x40CAE6A: ves_icall_System_Array_CreateInstanceImpl 
(icall.c:512)
==27640==    by 0x5D1D357: ???
==27640==    by 0x47755BE: ???
==27640==    by 0x4775513: ???
==27640==    by 0x5F8E9C5: ???
==27640==    by 0x5F8FBBD: ???
==27640==    by 0x5F8EB4A: ???
==27640==    by 0x5F8E798: ???

Independent from what's going on in my code, this doesn't look good. 
Does it ring any bells?

Some background details: Mono 1.2.2.1, Linux, ix86, the process was 
actively running for about 20 minutes, then stopped with SIGTERM. No 
native or pinvoked code is involved. I never experienced the 
aforementioned leak running the same code under mono 1.1.13.6.

-- 
MINDBREEZE®

KNOWLEDGE IS A MATTER OF SECONDS.
______________________________________________________________

mgr inż.
PAWEŁ SAKOWSKI, Product Development
pawel.sakowski at mindbreeze.com

Mindbreeze Software GmbH | Honauerstr. 4 | 4020 Linz | Austria
T +43. 732. 606162 - 0
F +43. 732. 606162 - 609
www.mindbreeze.com