This bug is marked as fixed but its happening to us again. Clicking “Deep Profile” soft locks unity for about 10 seconds without displaying the “Unity thinking wheel”. Then when you enter play crash.
Its from 2015 was 100% repro then and still seams to be for some people. It came out of no where though ( I was deep profiling fine before) maybe another library issue? Also we have auto-refresh disabled as with large projects it causes unity to hang way too much when working between code and editor to be useable. Disabling auto refresh causes a ton of new crashes though (but we are needing to pick crashes over a complete inability to get things done) Maybe this is related to auto-refresh?
Also as a note, turning deep profile on then off before entering play does not cause the crash. Turning it back on does though.
Thanks.
Happens to me also. Was working fine, then one day it just crashes after a second when running a scene. It always crashes on that scene (a very complex one) but I’ve tried it on a empty scene and it works there.
I get the debug the crash dialog from windows, where I can see exception thrown is AccessViolation from KernelBase.dll.
KernelBase.dll!RaiseException() Unknown
mono.dll!g_logv(const char * log_domain, GLogLevelFlags log_level, const char * format, char * args) Line 201 C
mono.dll!g_log(const char * log_domain, GLogLevelFlags log_level, const char * format, …) Line 211 C
mono.dll!mono_codegen(MonoCompile * cfg) Line 3196 C
mono.dll!mini_method_compile(_MonoMethod * method, unsigned int opts, _MonoDomain * domain, int run_cctors, int compile_aot, int parts) Line 4016 C
mono.dll!mono_jit_compile_method_inner(_MonoMethod * method, _MonoDomain * target_domain, int opt, _MonoException * * jit_ex) Line 4344 C
mono.dll!mono_jit_compile_method_with_opt(_MonoMethod * method, unsigned int opt, _MonoException * * ex) Line 4556 C
mono.dll!mono_jit_compile_method(_MonoMethod * method) Line 4580 C
mono.dll!mono_magic_trampoline(__int64 * regs, unsigned char * code, void * arg, unsigned char * tramp) Line 477 C
I think I have found a work around for this. If you start deep profiling in a simple scene (In my case a menu scene) and then use that scene to load the one you want to profile then it prevents the crash