ASan aborts under ulimit -v (ReserveShadowMemoryRange failed): use an RSS watchdog for memory-capped sanitizer runs

On a shared Linux host where every process must run under a memory cap, a binary built with g++ -fsanitize=address aborts at startup when run under ulimit -v <N>: AddressSanitizer failed to allocate 0xdfff0001000 bytes ... ReserveShadowMemoryRange failed ... Perhaps you are using ulimit -v. ASan reserves ~15 TB of virtual address space for shadow memory, so any virtual-memory limit kills it; the reservation is not resident, so RSS stays small.\n\nWhat worked: drop ulimit -v for ASan builds only and cap resident memory from outside with a small watchdog loop:\n``\n./prog.asan > out 2>&1 & p=$!\nwhile kill -0 $p 2>/dev/null; do r=$(awk "/VmRSS/{print \$2}" /proc/$p/status); [ "${r:-0}" -gt 1500000 ] && kill $p; sleep 0.2; done; wait $p\n`\n(cgroups/systemd-run -p MemoryMax are better if you have them; I did not on that host.)\n\nTwo related gotchas on the same host (GCC 11, RHEL-family):\n- -fsanitize=undefined failed to link: cannot find /usr/lib64/libubsan.so.1.0.0 (runtime not installed). Build with -fsanitize=address alone and build each step separately from the run, so a failed sanitizer link does not get silently skipped by an &&` chain.\n- On Linux, ASan includes LeakSanitizer by default (macOS does not), so a test that never frees a library context (e.g. ZSTD_createDCtx without ZSTD_freeDCtx) exits 1 with a leak report even though every check passed. Free it, or set ASAN_OPTIONS=detect_leaks=0 deliberately.