Segfault on Linux when executable is started with a non-path argv[0]

[ruth-o7]

When the Pharo executable is started with an argv[0] that is not a relative or absolute path, the VM will segfault.

PoC:

wrapper.sh

#!/usr/bin/env bash
exec -a "pharo" "$(which pharo)"

$ ./wrapper.sh 
Segmentation fault (core dumped)

stack trace

              #0  0x00007f775f908c5d __strlen_avx2 (libc.so.6 + 0x15bc5d)
              #1  0x00007f775f8c9af9 __strcpy_chk (libc.so.6 + 0x11caf9)
              #2  0x00007f775fab0371 setVMPath (libPharoVMCore.so + 0x22371)
              #3  0x00007f775fab2945 vm_parameters_parse (libPharoVMCore.so + 0x24945)
              #4  0x00007f775fab17a3 vm_main (libPharoVMCore.so + 0x237a3)
              #5  0x00007f775f7d4fce __libc_start_call_main (libc.so.6 + 0x27fce)
              #6  0x00007f775f7d5089 __libc_start_main@@GLIBC_2.34 (libc.so.6 + 0x28089)
              #7  0x0000000000401075 _start (.pharo-wrapped + 0x1075)
              ELF object binary architecture: AMD x86-64

I believe this is because the realpath call in src/utils.c:getFullPath returns a NULL if the provided path does not resolve to a real file (https://man7.org/linux/man-pages/man3/realpath.3.html), which is then fed later on to setVMPath. realpath also does not look at the PATH when searching for a file.

We should log the errno at the very least, and handle this correctly.