Fix struct.pack with padding bytes

[tannewt]

It used to validate the following arg could fit in a single byte.
Now, it always uses zero to pad.

Previous work on padding bytes in struct was done in #3405

[jepler]

The built-in struct tests are using the "modstruct" version, so either we need to circuitpythonify the "unix" build so it can use the shared-bindings version, or you need to

diff --git a/py/modstruct.c b/py/modstruct.c
index 1c5319608..8f78fa234 100644
--- a/py/modstruct.c
+++ b/py/modstruct.c
@@ -214,9 +214,11 @@ STATIC void struct_pack_into_internal(mp_obj_t fmt_in, byte *p, size_t n_args, c
             p += cnt;
         } else {
             while (cnt--) {
-                mp_binary_set_val(fmt_type, *fmt, args[i], &p);
                 // Pad bytes don't have a corresponding argument.
-                if (*fmt != 'x') {
+                if (*fmt == 'x') {
+                    mp_binary_set_val(fmt_type, *fmt, MP_OBJ_NEW_SMALL_INT(0), &p);
+                } else {
+                    mp_binary_set_val(fmt_type, *fmt, args[i], &p);
                     i++;
                 }
             }

[jepler]

thanks! seems good to go.

[jepler]

As far as I recall, this fix isn't needed upstream, as they truncated values in struct.pack as a code-saving / convenience measure, so packing 0x100 into an x wasn't an issue.