This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: PATCH RFA: C++ frontend: Don't warn about shifts which will not be run


"Joseph S. Myers" <joseph@codesourcery.com> writes:

> On Fri, 12 Jun 2009, Ian Lance Taylor wrote:
>
>> -/* Nonzero means the expression being parsed will never be evaluated.
>> -   This is a count, since unevaluated expressions can nest.  */
>> +/* Nonzero means the expression being parsed will never be
>> +   evaluated.  */
>
> Pre-existing condition, but this is not accurate.  skip_evaluation is a 
> *heuristic* to avoid spurious warnings; it's set inside sizeof even though 
> we don't know until after the whole expression/type inside sizeof has been 
> parsed whether it's actually sizeof(VLA) which is evaluated.
>
> I'm not convinced the distinction you are making here between 
> unexecuted/unevaluated is a very well-defined one for C, although such 
> concepts do appear in the standard to some extent (6.9 considers only 
> expressions inside sizeof-non-VLA while 6.6#3 has a less precisely defined 
> "not evaluated" for constant expressions that includes at least the 
> example 2 || 1 / 0).  The implementations of these two distinct concepts 
> in GCC are quite complicated, and separate from the heuristic 
> skip_evaluation.

Thanks for the comments but I'm not clear on what the recommendation is.
Are you saying that my patch should be modified?  Or are you saying that
I should restrict the change to the C++ frontend?  Or are you saying
something else?

What I'm trying to do here is fundamentally a near-trivial modification
to the C++ frontend.  I'd like to do this in the simplest possible
fashion.

Ian


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]