This is the mail archive of the
mailing list for the GCC project.
Re: [trunk<-vta] Re: [vta, 4.4] make df-scan df_collection_rec arrays overflow-safe
Ian Lance Taylor wrote:
> Richard Guenther <firstname.lastname@example.org> writes:
>> On Mon, Jun 1, 2009 at 6:39 PM, Paolo Bonzini <email@example.com> wrote:
>>>> Does anybody think that would be a bad idea?
>>> No, on the contrary it is a good idea.
>> Can we have compile-time checking that you not let such VEC
>> escape somehow?
> The ways that I can see that it could escape would be 1) declare a
> global variable as VEC(type,alloca); 2) assign a pointer to a
> VEC(type,alloca) to a global variable; 3) Return a pointer to a
> VEC(type,alloca) declared in the same function. Case 2 would actually
> be OK as long as you didn't use the global variable after the original
> function returned. I don't see how to warn about these without
> introducing an additional type attribute. Heck, we don't even warn
> about "return alloca (10);".
Could GTY be hacked to scream if it comes across a pointer to a VEC that
shouldn't have escaped? Maybe as a --enable-checking feature?