On 2/18/21 11:03 AM, Jakub Jelinek wrote:
On Thu, Feb 18, 2021 at 07:00:52PM +0100, Jakub Jelinek wrote:
The size of the VLA is zero regardless of its bound and accessing
it is invalid so the warning is expected.
Yes, some warning, but not the one you are giving, that is nonsensical.
Array subscript 0 is not outside of array bounds of struct S[n], a[anything]
will still be zero sized and will not be problematic.
The warning is designed for ordinary arrays of nonzero size. There's
no point in putting an effort into coming up with a special warning
just for those because they serve no purpose in these contexts (as
complete objects).
Scalar objects with zero size will always have that zero size,
similarly arrays thereof (constant or variable sized).
So the warning should be simply if eltsize == 0,
check if the access is before or after the object and complain
that a memory access is done before or after a zero sized object %qD.
Jakub
No, I don't think making this exception would be helpful. Zero length
arrays are a non-standard extension meant to be used as struct members,
before flexible array members were added to C. In other contexts, they
are almost certainly unintended and so likely bugs. There's no valid
use case for such arrays, and diagnosing accesses to them helps find
such bugs.
That said, I also don't think a fix for the ICE should be held up
because we disagree on this vanishingly unimportant corner case.
The ICE effectively prevents using such arrays (VLAs) and since no
bug reports have been raised for it since it was introduced in GCC
9 it's unlikely that any code relies on it. (I suspect the bug
itself was the result of fuzzing.)
Martin
