[3.4-RC] Macro problem

[PowerWyrm]

Just a little reminder for C devs: be careful when using macros evaluating variables more than once, because unwanted effects could occur. I've screwed myself more than once with such macros...

Here's a good example from mon-spell.c:

inven_damage(p_ptr, re_ptr->gf, MIN(dam * randcalc(re_ptr->dam, 0, RANDOMISE), 300));

The MIN macro is defined as:

#define MIN(a,b)    (((a) > (b))? (b)   : (a))

So writing something like MIN(random, fixed) is actually (random > fixed? fixed: random).

In the previous example, nothing can ensure you that the result will be lower than 300, since randcalc will be called twice, once to compare the result to 300 and once to return the result!

The fix is simple:

int calcdam = dam * randcalc(re_ptr->dam, 0, RANDOMISE);
inven_damage(p_ptr, re_ptr->gf, MIN(calcdam, 300));

[Derakon]

As noted in another thread, it'd probably be a good idea to write macro names in ALL_CAPS or some other distinguishing method. It won't guarantee that you won't make this mistake, but it should help remind you that you aren't making a normal function call.

And of course, where possible macros should be replaced by function calls. I don't think even Angdroid needs the minor performance improvements macros get you.

[Patashu]

Won't good compilers inline the kinds of functions you'd make macros in the first place?

[Magnate]

Looks like this ought to be my job, since PowerWyrm's example (and all the many ones like it) are my errors. It might explain quite a few things!

Opened as #1668.