From: Arnd Bergmann > Sent: 14 September 2021 15:54 > > On Tue, Sep 14, 2021 at 4:24 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote: > > On 9/14/21 7:17 AM, David Laight wrote: > > >> Sorry, I didn't realize that a field of size 0 increases the structure size > > >> on sparc. I had checked the size of the old and the new structure with gcc > > >> on x86_64 and didn't see a field size increase. > > > > > > clang output doesn't change: > > > > > > https://godbolt.org/z/bTeeq19j1 > > > > > > gcc ought to generate the same size. > > > > > > It ought to be 'char data[];' though. > > > > > > > I am never sure if [] or [0] is "correct". Anyway, is there agreement that this > > is an acceptable solution ? I'll be happy to resend if that is the case. > > Yes, looks good to me, in the [] version. I think the [0] version can be > interpreted as a zero-length array that may not be accessed, while the > [] flexible array syntax clearly means that extra data follows, and it's > part of the C standard now, while [0] is a gcc extension. More problematic is where is the correct place for the 'char data[]'. It follows the header rather than being part of it. So the: data = (void *)(hdr + 1); construct (I've lost the original patch) is absolutely descriptive. gcc is getting to be a real PITA for system coding. For this particular check 'size 0' ought to be 'size unknown' and always valid. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
