Re: [9front] [PATCH] Fix assert macro to not break on commas

[Dan Cross <crossd@gmail.com>]

On Mon, Nov 27, 2023 at 10:40 AM <ori@eigenstate.org> wrote:
> Quoth Blue-Maned_Hawk <bluemanedhawk@gmail.com>:
> > On 11/26/23 19:00, ori@eigenstate.org wrote:
> > > Quoth Blue-Maned_Hawk <bluemanedhawk@gmail.com>:
> > >> On 11/26/23 17:52, ori@eigenstate.org wrote:
> > >>> Quoth Blue-Maned_Hawk <bluemanedhawk@gmail.com>:
> > >>>> What reason have y'all for avoiding macros?
> > >>>
> > >>> Well, the fact that you need hacks like the one that started this
> > >>> thread, for one;
> > >>
> > >> I don't see how it's a hack, but maybe i've just become numb to macro
> > >> shenaniganry.
> > >>
> > >>>                   what if you had  macro that took *two* arguments?
> > >>>
> > >>>   #define assert_msg(cond, msg) \
> > >>>           ...what here?
> > >>>
> > >>> I don't see how you can avoid just saying "it's a macro, it's going
> > >>> to be weird, don't even try".
> > >>>
> > >>
> > >> Don't even try writing the macro or don't even try making it not weird?
> > >> Because i can thing of a way to define such a thing:
> > >>
> > >>    #define assert_msg(cond, msg) ((cond) ? (void)0 : (print(msg), exits(msg)))
> > >>
> > >> and i don't think that that's a particularly weird way to do so.
> > >
> > >
> > > Wouldn't that fall apart on the same case that you showed above?
> > >
> > >     assert_msg((Thing){0, 1}.a, "foo");
> > >
> >
> > Aye, correct you are.  In that case, it could be rearranged to
> >
> >       #define assert_msg(msg, ...) ((__VA_ARGS__) ? (void)0 : (print(msg),
> > exits(msg)))
> >
> > and it would be able to tolerate unparenthesized commas in the checked
> > expression.
>
> assert_msg((Thing){0, "foo"}.a, (Thing){1, "bar"}.b);

It seems like one could keep going with more complex macros and every
expanded use on those macros to demonstrate that the way program text
is parsed in the preprocessor and the language proper is different.
But this is getting a bit far away from the original change
suggestion, which was substituting three symbols in a single macro for
a different three symbols. Whether that's worth it just to avoid a set
of parentheses when using the comma operator or a struct literal in an
`assert` is debatable, but the difference in complexity is pretty much
negligible.

Regardless, I'd suggest that almost no one implements `assert()` the
suggested way, and it would be just another gratuitous difference
between the C used here and the rest of the world; that alone is
enough to reject the patch (to me anyway; you all do you).

        - Dan C.