wrong size of fields for icmp4 timestamp message

16bits are enough for Id/Seq.

The `len` parameter of `NLMSG_LENGTH`  was size of Netlink header instead of RtNetlink header.

Also make it more clear that the parameter of the macro is the payload size, which is RtNetlink message in this case.

…ompiler (#104)

* python/dnet.pyx: fix incompatible-function-pointer-types for modern compiler

which is error now, see https://bugs.gentoo.org/933360,
clang 19 (maybe earlier) has the same problem too

* python/dnet.pyx: fix -Wincompatible-pointer-types reported by gcc14

./dnet.c:8451:52: error: passing argument 2 of ‘PyObject_AsReadBuffer’ from incompatible pointer type [-Wincompatible-pointer-types]
 8451 |   __pyx_t_1 = (PyObject_AsReadBuffer(__pyx_v_pkt, (&__pyx_v_p), (&__pyx_v_n)) == 0);
      |                                                   ~^~~~~~~~~~~
      |                                                    |
      |                                                    char **
/usr/include/python3.12/abstract.h:370:52: note: expected ‘const void **’ but argument is of type ‘char **’
  370 |                                       const void **buffer,
      |                                       ~~~~~~~~~~~~~^~~~~~
./dnet.c:8451:66: error: passing argument 3 of ‘PyObject_AsReadBuffer’ from incompatible pointer type [-Wincompatible-pointer-types]
 8451 |   __pyx_t_1 = (PyObject_AsReadBuffer(__pyx_v_pkt, (&__pyx_v_p), (&__pyx_v_n)) == 0);
      |                                                                 ~^~~~~~~~~~~
      |                                                                  |
      |                                                                  int *
/usr/include/python3.12/abstract.h:371:51: note: expected ‘Py_ssize_t *’ {aka ‘long int *’} but argument is of type ‘int *’
  371 |                                       Py_ssize_t *buffer_len);
      |                                       ~~~~~~~~~~~~^~~~~~~~~~

Signed-off-by: Z. Liu <[email protected]>

* m4/acinclude.m4: write `ETH_P_ALL` check as function body, not function definition

- Fixes #109

- `AC_TRY_COMPILE` macro's second argument is a function body, not a
  complete function definition. Placing a complete function defintion
  here results in m4 generating a malformed test program with a function
  definition inside another function defintion, which would break the
  check to make it always evaluate false whenever Clang is used, since
  Clang rejects programs that use GCC-only extensions like function
  definition inside another function definition.

* tree-wide: regenerate autotools files

- Necessary to apply the `ETH_P_ALL` function body change to the
  `configure` script, in order to fix #109

- `aclocal --force` run

- `autoreconf -fi` run

- `autom4te.cache`, which appeared afterward, added to `.gitignore`