Skip to main content

FFI utilities (/std/ffi_util)

A pure-LPC helper library for code that calls native functions through package_ffi. It packages the recurring idioms at the C boundary — NUL-terminated strings, returned char* pointers, out-parameters, argv arrays, struct fields — so binding code doesn't re-derive them. It lives in testsuite/std/ffi_util.lpc and is portable to any FluffOS mudlib (copy the file plus include/ffi.h).

inherit "/std/ffi_util";
Inherit it into the privileged object — don't simul-efun it

c_string()/c_string_enc() call ffi_peek, which the master gates via valid_ffi("peek", addr, caller) — and the caller the master sees is the object this library is inherited into. Exposed as a simul-efun, every peek in the mud would run with the simul_efun object as its security principal. Inherit the library directly into the object(s) allowed to do FFI work instead.

C strings in: cstr(), cstr_enc()​

buffer cstr(string s);
buffer cstr_enc(string s, string encoding);

cstr() returns the string's raw UTF-8 bytes plus a NUL terminator — exactly what a const char* parameter (an FFI_POINTER argument) wants. cstr_enc() converts through string_encode first, for C functions that expect Latin-1 or another byte-oriented encoding. (The terminator is a single NUL byte; wide encodings like UTF-16 need a wider terminator — build those by hand.)

int sl = ffi_prepare(lib, "strlen", FFI_UINT64, ({ FFI_POINTER }));
ffi_call(sl, ({ cstr("café") }));                  // 5: UTF-8 bytes
ffi_call(sl, ({ cstr_enc("café", "latin-1") }));   // 4: é is one byte

C strings out: c_string(), c_string_enc()​

string c_string(int addr);
string c_string_enc(int addr, string encoding);

Read the NUL-terminated bytes at a raw foreign address (a returned char*) and decode them into an LPC string. A NULL (0) address yields 0 instead of an error, matching the C convention, so returned pointers can be passed straight through:

int ge = ffi_prepare(lib, "getenv", FFI_POINTER, ({ FFI_POINTER }));
string path = c_string(ffi_call(ge, ({ cstr("PATH") })));   // 0 if unset

Out-parameters: c_out()​

buffer c_out(int type_code);

A zeroed block sized for one scalar of the given FFI_* type code — pass it where C wants T *out, then ffi_read(out, 0, code):

int fx = ffi_prepare(lib, "frexp", FFI_DOUBLE, ({ FFI_DOUBLE, FFI_POINTER }));
buffer exp = c_out(FFI_INT);
float mant = ffi_call(fx, ({ 8.0, exp }));   // 0.5
int e = ffi_read(exp, 0, FFI_INT);           // 4

c_out(FFI_POINTER) works for pointer out-parameters too — e.g. strtol's char **endptr, read back with ffi_read and c_string().

Argument vectors: c_argv()​

mixed *c_argv(string *args);

Builds a C char *argv[]: a NULL-terminated pointer array whose slots hold the addresses of one cstr() buffer per string. Returns ({ pointer_array, string_buffers }) — pass element 0 as the FFI_POINTER argument, and keep the returned array referenced for as long as C may use the pointers: the addresses in the pointer array die when the buffers in element 1 are collected.

mixed *av = c_argv(({ "ls", "-l" }));
ffi_call(f, ({ cstr("/bin/ls"), av[0] }));   // av stays in scope

Struct fields: c_field(), c_field_set()​

mixed c_field(buffer data, mixed *layout, int idx, int type_code);
void  c_field_set(buffer data, mixed *layout, int idx, int type_code, mixed value);

Read/write field idx of a C struct held in data, using a layout from ffi_struct_layout (the offset is layout[1][idx]). With the symbolic indexes the tools/ffi generator emits, field access reads by name:

mixed *lay = ffi_struct_layout(STRUCT_TM_TYPES);
int year = 1900 + c_field(tm, lay, STRUCT_TM_tm_year, FFI_INT);

See also​