https://gitlab.synchro.net/main/sbbs/-/commit/a1720ea9f6d2f2fa0b0969f2
Added Files:
src/xpdev/SNPRINTF_AUDIT.md
Log Message:
xpdev: scope the snprintf() return-value divergence for a future fix

genwrap.h redefines snprintf() to safe_snprintf() unless USE_SNPRINTF is defined, and safe_snprintf() clamps its return to size-1 rather than
returning C99's would-be length. USE_SNPRINTF is set for darwin and freebsd only (Common.gmake), and sbbs3 inherits those flags -- so the return value means the would-be length on macOS/FreeBSD and the truncated length on Linux and Windows. The same source line behaves differently per platform, with no diagnostic either way.

That is not a theoretical problem: main.cpp's unix-domain spy-socket path
check (`snprintf(...) >= sizeof(sun_path)`) is correct on macOS/FreeBSD and UNREACHABLE on Linux, where an over-long temp_dir silently binds a truncated socket path instead of logging the error and skipping the node.

The obvious fix -- make safe_snprintf() conformant -- is wrong as a first
move: ~32 callers accumulate with `d += snprintf(...)` and depend on the
clamp to stay inside the buffer, so removing it trades one silent truncation for a set of overruns.

No code changed. This records the platform matrix, the confirmed bug, why
each of the four available fixes is or isn't safe, and a recommended inventory-first path, so the work can be picked up without re-deriving any
of it.

---
þ Synchronet þ Vertrauen þ Home of Synchronet þ [vert/cvs/bbs].synchro.net