I wasn't able to reproduce your crash, but there is definitely a problem - when using --enable-shared and --with-jemalloc together, the Ruby that gets built still uses libc's malloc and ignores jemalloc. This is because we pass -ljemalloc to the link line for libruby.so, but we don't pass it to ruby. This means that the built Ruby isn't marked as needing libjemalloc.so:
root@jammy-189dc9d584290f1a:/var/ruby# readelf --dynamic ruby | grep NEEDED 0x0000000000000001 (NEEDED) Shared library: [libruby.so.3.3] 0x0000000000000001 (NEEDED) Shared library: [libc.so.6]
And because the dynamic linker (at least the glibc one) links libraries in breadth-first order, that means that libc.so.6 is linked before libjemalloc.so.2:
root@jammy-189dc9d584290f1a:/var/ruby# ldd ruby linux-vdso.so.1 (0x00007ffe873fb000) libruby.so.3.3 => /usr/local/lib/libruby.so.3.3 (0x00007f8870000000) libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f886fc00000) libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007f887054c000) libjemalloc.so.2 => /lib/x86_64-linux-gnu/libjemalloc.so.2 (0x00007f886f800000) libcrypt.so.1 => /lib/x86_64-linux-gnu/libcrypt.so.1 (0x00007f8870512000) libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f886ff19000) /lib64/ld-linux-x86-64.so.2 (0x00007f8870572000) libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f886f400000) libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f886fef9000)
We need to pass -ljemalloc to the linker command line for the final Ruby executable. I'm playing around trying to find the right Autoconf magic spells for this now.
Updated by 
Updated by 
Updated by 
Updated by 
Updated by