gap-system / gap / 11314

4%

push

travis-ci

kernel: introduce BOOL, TRUE, FALSE

This adds `typedef Int BOOL`; the resulting C type `BOOL` can be used to
indicate that the return type of a function is supposed to be a boolean.
Matching integer constants TRUE=1 and FALSE=0 are also added.

The reason we are not using type `bool` from stdbool.h and the constants
`true` and `false`, all available in C99, is the following: in C99, a pointer
is automatically converted to type bool if necessary. That means that if one
by accident writes `return False` instead of `return false` in a function with
return type `BOOL`, then no compiler warning is issued. Instead, a bug is
introduced, as `False` is a non-NULL pointer, and thus is cast to the `bool`
value `true`.

With this commit, we get warnings similar to the following if we mix up
TRUE/True or FALSE/False; while if using `bool`, we'd only get the latter two
warnings, but not the first one.

    src/objects.c:897:12: error: incompatible pointer to integer conversion
          returning 'Obj' (aka 'unsigned long **') from a function with result
          type 'BOOL' (aka 'unsigned char') [-Werror,-Wint-conversion]
        return False;
               ^~~~~

    src/objects.c:1185:39: error: pointer/integer type mismatch in conditional
          expression ('int' and 'Obj' (aka 'unsigned long **'))
          [-Werror,-Wconditional-type-mismatch]
        return (TNUM_OBJ(obj) == T_COMOBJ ? TRUE : False);
                                          ^ ~~~~   ~~~~~

    src/objects.c:1337:16: error: incompatible integer to pointer conversion
          returning 'int' from a function with result type 'Obj' (aka
          'unsigned long **') [-Werror,-Wint-conversion]
            return TRUE;
                   ^~~~

Note that on the long run, it would be better to switch to something like
`typedef char BOOL`, but that breaks at least one package (NormalizInterface),
so let's postpone that until all affected packages with kernel extensions have
switched to using BOOL.

45 of 90 new or added lines in 38 files covered. (50.0%)

285804 of 337317 relevant lines covered (84.73%)

30026433.33 hits per line