I get reproducable crashes in filesrc when running a lot of files 
(usually happens around 300) through a pipeline of this kind: 
 
   filesrc ! typefind ! fakesink 
 
A new pipeline (and new elements) are created for every file in my case. 
 
 
Valgrind only shows this: 
 
==534== Invalid read of size 1 
==534==    at 0x4394204A: mpeg_video_stream_type_find 
(gsttypefindfunctions.c:688) 
==534==    by 0x4029B97C: gst_type_find_factory_call_function 
(gsttypefind.c:193) 
==534==    by 0x42F70A7C: gst_type_find_element_chain (gsttypefind.c:522) 
==534==    by 0x40285D37: gst_pad_push (gstpad.c:2901) 
==534==  Address 0x9B885834 is 0 bytes after a block of size 2048 alloc'd 
==534==    at 0x4002CBEE: malloc (vg_replace_malloc.c:160) 
==534==    by 0x40507161: g_malloc (gmem.c:136) 
==534==    by 0x40266A2B: gst_buffer_new_and_alloc (gstbuffer.c:224) 
==534==    by 0x42F5F93C: gst_buffer_store_get_buffer 
(gstbufferstore.c:344) 
==534== 
==534== Invalid read of size 4 
==534==    at 0x42F65746: gst_filesrc_get_mmap (gstfilesrc.c:503) 
==534==    by 0x42F2E720: get_group_schedule_function 
(gstoptimalscheduler.c:965) 
==534==    by 0x42F2E1F1: schedule_group (gstoptimalscheduler.c:842) 
==534==    by 0x42F2E2B5: gst_opt_scheduler_schedule_run_queue 
(gstoptimalscheduler.c:881) 
==534==  Address 0x44 is not stack'd, malloc'd or free'd 
==534== 
==534== Process terminating with default action of signal 11 (SIGSEGV): 
dumping core 
==534==  Address not mapped to object at address 0x44 
==534==    at 0x42F65746: gst_filesrc_get_mmap (gstfilesrc.c:503) 
==534==    by 0x42F2E720: get_group_schedule_function 
(gstoptimalscheduler.c:965) 
==534==    by 0x42F2E1F1: schedule_group (gstoptimalscheduler.c:842) 
==534==    by 0x42F2E2B5: gst_opt_scheduler_schedule_run_queue 
(gstoptimalscheduler.c:881) 
==534== 
 
 
I'll attach the code that produces the problem. Compile and let loose on 
a folder with hundreds of mp3s like this: 
 
./detect /some/folde/with/mp3s 
 
 
Cheers 
-Tim