Eval bug: fail to load model in MacOS

[Animaxx]

Name and Version

5931c1f (ggml : add support for dynamic loading of backends) this commit causing model loading error in MacOS

in llama.cpp, make_cpu_buft_list function

   // add extra buffer types
   auto * cpu_dev = ggml_backend_dev_by_type(GGML_BACKEND_DEVICE_TYPE_CPU);
   auto * cpu_reg = ggml_backend_dev_backend_reg(cpu_dev);

cpu_dev is NULL and ggml_backend_dev_backend_reg(cpu_dev) give EXC_BAD_ACCESS

Operating systems

Mac

GGML backends

Metal

Hardware

Mac M1

Models

No response

Problem description & steps to reproduce

Load this commit, build in MacOS and load any model.

First Bad Commit

#10469

Relevant log output

"exception" : {"codes":"0x0000000000000001, 0x0000000000000078","rawCodes":[1,120],"type":"EXC_BAD_ACCESS","signal":"SIGSEGV","subtype":"KERN_INVALID_ADDRESS at 0x0000000000000078"},
  "termination" : {"flags":0,"code":11,"namespace":"SIGNAL","indicator":"Segmentation fault: 11","byProc":"exc handler","byPid":3547},
  "vmregioninfo" : "0x78 is not in any region.  Bytes before following region: 4308057992\n      REGION TYPE                    START - END         [ VSIZE] PRT\/MAX SHRMOD  REGION DETAIL\n      UNUSED SPACE AT START\n--->

[slaren]

That's not enough information to reproduce the issue.

[Animaxx]

hmm let me create a demo project

[Animaxx]

the issue related to my project cache, thank you for replying @slaren. I am going to close this one