[Bug c/35579] __attribute__(( warn_unused_result )) warns when it shouldn't, doesn't warn when it should

["manu at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

------- Comment #3 from manu at gcc dot gnu dot org  2008-03-14 09:36 -------
(In reply to comment #2)
> >This is the design of warn_unused_result, you cannot ignore the value, 
> >that is why casting to void does not work.

I agree with the reporter. There should be a way to tell the compiler to avoid
warning for particular cases. Casting to void seems to be the best way to do
that.

> >The warning happens in the front-end so we don't know if you really ignore the
> >value if you don't use the actually variable.
> 
> Is there any way of fixing this?  warn_unused_result is used in a number of
> quite significant pieces of code to enforce safety properties (the Linux kernel
>   springs to mind), having a compiler-enforced safety check that doesn't work
> reliably seems rather dangerous.  If it can't be fixed then could it at least
> be documented that the check isn't always reliable?

Unfortunately, the front-end lacks any data-flow knowledge, so this can't be
fixed. Almost all GCC warnings have the same limitations.

Documenting that warn_unused_result does not detect certain cases seems good.
Care to write a text? I will format it and submit it as a patch if you want.

Thanks.


-- 

manu at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |manu at gcc dot gnu dot org
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2008-03-14 09:36:05
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35579