Martin Sebor msebor@redhat.com
Tue May 19 17:34:00 GMT 2015

On 05/19/2015 10:04 AM, Marc Glisse wrote:
> On Mon, 18 May 2015, Martin Sebor wrote:
>> On 05/16/2015 08:58 AM, Eduardo Piombino wrote:
>>> Hi,
>>> Is there any way to get gcc to generate a warning upon some
>>> compile-time condition?
>>> I'd be aiming at something like _Static_assert, but raising a warning
>>> instead of an error.
>>> Nowadays, to achieve something similar, I have to run 2 builds, one
>>> with the assert enabled just to get notified of such conditions, and
>>> then a another one, without the asserts, to actually compile them.
>> I don't think there's anything like what you're looking for.
>> Short of hacking the compiler and implementing it, a similar
>> effect can be achieved by [ab]using an existing warning. For
>> example, like so:
>> $ cat t.c && gcc -Wall -c -o/dev/null t.c
>> extern int printf (const char*, ...) __attribute__ ((nonnull (1)));
>> #define Static_Warning(x, txt) \
>>        (void)((x) ? 0 : printf ("%s%s\n", txt, (char*)x))
>> void foo (void)
>> {
>>    Static_Warning (0, "foo");
>>    Static_Warning (1, "bar");
>> }
>> t.c: In function ‘foo’:
>> t.c:8:5: warning: reading through null pointer (argument 3) [-Wformat=]
>>     Static_Warning (0, "foo");
>>     ^
> Does the "warning" function attribute not work? (or maybe "deprecated")

I hadn't tried attribute warning. It does work and makes for
a cleaner solution than the workarounds I suggested. Thanks
for pointing it out! (Unfortunately, there doesn't appear to
be a way to use the warning message for the static warning,
but that's only a minor wrinkle.)

Attribute deprecated is not suitable for this purpose because
it's not subject to dead code elimination and so the warning
is issued regardless of the value of the constant expression.


More information about the Gcc-help mailing list