Re: [PATCH] Fix fold_offsetof_1 (PR middle-end/41935)

[Paolo Bonzini <bonzini at gnu dot org>]

> Also, IMHO we shouldn't warn with -Warray-bounds for offsetof with
> one past the last entry, as taking address of that is allowed in address
> arithmetics and offsetof doesn't dereference that in any way.

We also shouldn't warn about array bounds for the last entry of a 
struct, like

typedef struct _VPD_IDENTIFICATION_DESCRIPTOR {
    UCHAR   CodeSet         : 4;
    UCHAR   Reserved        : 4;
    UCHAR   IdentifierType  : 4;
    UCHAR   Association     : 2;
    UCHAR   Reserved2       : 2;
    UCHAR   Reserved3;
    UCHAR   IdentifierLength;
    UCHAR   Identifier[1];
} VPD_IDENTIFICATION_DESCRIPTOR;
typedef VPD_IDENTIFICATION_DESCRIPTOR   *PVPD_IDENTIFICATION_DESCRIPTOR;

...

return offsetof (VPD_IDENTIFICATION_DESCRIPTOR, Identifier[4]);

(Saw that today in real-world code).

Paolo