--- Comment #9 from Giuseppe Bilotta <> ---
Ok, so now what I find is that running clinfo under heaptrack causes a
segfault. I can get some more information if I run `heaptrack -d clinfo`.
Sadly, the segmentation fault happens inside the proprietary Intel OpenCL
platform. I suspect that one of the LD preloads is interfering. The backtrace

#0  0x00007fffbb08abc1 in ?? () from
#1  0x00007fffbb0082fe in clGetPlatformInfo () from
#2  0x00007ffff79c26d4 in _malloc_clGetPlatformInfo
(plt_info_ptr=plt_info_ptr@entry=0x7fffbb007ef0 <clGetPlatformInfo>,
pid=0x555555ac0d80, cname=cname@entry=2308, sname=<optimized out>)          
    at ocl_icd_loader.c:298
#3  0x00007ffff79c2fc4 in _find_and_check_platforms (num_icds=<optimized out>)
at ocl_icd_loader.c:498                                                         
#4  __initClIcd () at ocl_icd_loader.c:652                                      
#5  _initClIcd_real () at ocl_icd_loader.c:702
#6  0x00007ffff79c36ab in _initClIcd () at ocl_icd_loader.c:724
#7  clGetPlatformIDs (num_entries=0, platforms=0x0,
num_platforms=0x55555576ab58 <num_platforms>) at ocl_icd_loader.c:846           
#8  0x0000555555561ffb in main (argc=1, argv=0x7fffffffe1b8) at

I'm not sure if it's easy to reproduce, but it's probably going to need the
Intel OpenCL ICD and the use of the open-source ocl-icd for libOpenCL.

You are receiving this mail because:
You are watching all bug changes.

Reply via email to