SIGSEGV when running rdm with LLVM-12

[fperies]

Description
Hello,
I'm getting systematically a SIGSEV when I run rdm built with LLVM-12.

To Reproduce
I simply build and run rdm.
During rdm build I had the following output from CMake:

...
-- Detecting CXX compile features - done
-- llvm-config executable found: /usr/local/llvm/bin/llvm-config
-- Using Clang version 12.0.1 from /usr/local/llvm/lib with CXXFLAGS -I/usr/local/llvm/include -D_GNU_SOURCE -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS
-- Found LibClang: /usr/local/llvm/lib/libclang.so  
...

Expected behavior
Server should run without crashing.

Backtrace
With GDB:

(gdb) r
+r
Starting program: /ebs/frperies/repo/rtags/build/bin/rdm 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".

Program received signal SIGSEGV, Segmentation fault.
std::call_once<void (&)()>(std::once_flag&, void (&)())::{lambda()#1}::operator()() const (this=0x0) at /usr/local/gcc-10.3.0/include/c++/10.3.0/mutex:717
717		  std::__invoke(std::forward<_Callable>(__f),
(gdb) bt
#0  std::call_once<void (&)()>(std::once_flag&, void (&)())::{lambda()#1}::operator()() const (this=0x0) at /usr/local/gcc-10.3.0/include/c++/10.3.0/mutex:717
#1  std::call_once<void (&)()>(std::once_flag&, void (&)())::{lambda()#2}::operator()() const (this=0x0) at /usr/local/gcc-10.3.0/include/c++/10.3.0/mutex:722
#2  std::call_once<void (&)()>(std::once_flag&, void (&)())::{lambda()#2}::_FUN() () at /usr/local/gcc-10.3.0/include/c++/10.3.0/mutex:722
#3  0x00007ffff5dd520b in __pthread_once_slow () from /lib64/libpthread.so.0
#4  0x00007fffef1c8c2b in llvm::ManagedStaticBase::RegisterManagedStatic(void* (*)(), void (*)(void*)) const () from /usr/local/llvm/lib/../lib/libLLVM-12.so
#5  0x00007fffef16c512 in llvm::cl::OptionCategory::registerCategory() () from /usr/local/llvm/lib/../lib/libLLVM-12.so
#6  0x00007fffef05345c in __static_initialization_and_destruction_0(int, int) [clone .constprop.0] () from /usr/local/llvm/lib/../lib/libLLVM-12.so
#7  0x00007ffff7dea9c3 in _dl_init_internal () from /lib64/ld-linux-x86-64.so.2
#8  0x00007ffff7ddc17a in _dl_start_user () from /lib64/ld-linux-x86-64.so.2
#9  0x0000000000000001 in ?? ()
#10 0x00007fffffffcb55 in ?? ()
#11 0x0000000000000000 in ?? ()

As suggested from another similar issue, I did a run with valgrind:

$ valgrind --tool=memcheck --leak-check=yes --show-reachable=yes --num-callers=20 --track-fds=yes ~/bin/rdm
==27316== Memcheck, a memory error detector
==27316== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==27316== Using Valgrind-3.18.1 and LibVEX; rerun with -h for copyright info
==27316== Command: /ebs/frperies/bin/rdm
==27316== 
==27316== Invalid read of size 8
==27316==    at 0x6025C2: std::call_once<void (&)()>(std::once_flag&, void (&)())::{lambda()#2}::_FUN() (invoke.h:60)
==27316==    by 0x69F020A: __pthread_once_slow (in /usr/lib64/libpthread-2.17.so)
==27316==    by 0x89DCC2A: llvm::ManagedStaticBase::RegisterManagedStatic(void* (*)(), void (*)(void*)) const (in /usr/local/llvm/lib/libLLVM-12.so)
==27316==    by 0x8980511: llvm::cl::OptionCategory::registerCategory() (in /usr/local/llvm/lib/libLLVM-12.so)
==27316==    by 0x886745B: __static_initialization_and_destruction_0(int, int) [clone .constprop.0] (in /usr/local/llvm/lib/libLLVM-12.so)
==27316==    by 0x400F9C2: _dl_init (in /usr/lib64/ld-2.17.so)
==27316==    by 0x4001179: ??? (in /usr/lib64/ld-2.17.so)
==27316==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
==27316== 
==27316== 
==27316== Process terminating with default action of signal 11 (SIGSEGV): dumping core
==27316==  Access not within mapped region at address 0x0
==27316==    at 0x6025C2: std::call_once<void (&)()>(std::once_flag&, void (&)())::{lambda()#2}::_FUN() (invoke.h:60)
==27316==    by 0x69F020A: __pthread_once_slow (in /usr/lib64/libpthread-2.17.so)
==27316==    by 0x89DCC2A: llvm::ManagedStaticBase::RegisterManagedStatic(void* (*)(), void (*)(void*)) const (in /usr/local/llvm/lib/libLLVM-12.so)
==27316==    by 0x8980511: llvm::cl::OptionCategory::registerCategory() (in /usr/local/llvm/lib/libLLVM-12.so)
==27316==    by 0x886745B: __static_initialization_and_destruction_0(int, int) [clone .constprop.0] (in /usr/local/llvm/lib/libLLVM-12.so)
==27316==    by 0x400F9C2: _dl_init (in /usr/lib64/ld-2.17.so)
==27316==    by 0x4001179: ??? (in /usr/lib64/ld-2.17.so)
==27316==  If you believe this happened as a result of a stack
==27316==  overflow in your program's main thread (unlikely but
==27316==  possible), you can try to increase the size of the
==27316==  main thread stack using the --main-stacksize= flag.
==27316==  The main thread stack size used in this run was 8388608.
==27316== 
==27316== FILE DESCRIPTORS: 3 open (3 std) at exit.
==27316== 
==27316== HEAP SUMMARY:
==27316==     in use at exit: 0 bytes in 0 blocks
==27316==   total heap usage: 0 allocs, 0 frees, 0 bytes allocated
==27316== 
==27316== All heap blocks were freed -- no leaks are possible
==27316== 
==27316== For lists of detected and suppressed errors, rerun with: -s
==27316== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
Segmentation fault

It looks a read of size 8 is done at adress 0x0.

Environment

• OS/Version: CentOs
• RTags(rdm,rc): 2.33
• Emacs: 28.1
• LLVM/Clang: 12.01

Thank you for your support.

[aschmitz-crnc]

I think this is an ongoing issue as rtags seems to be currently mostly unmaintained. @Andersbakken is there any chance to get rtags back on track for more recent LLVM releases? It is a massively useful tool and in my experience much faster and leaner than LSP. So it would be nice to have it work correctly again. I tried to have a look but am not experienced enough with LLVM / clang to understand what's going on.