Re: [Xen-devel] [PATCH v10 2/6] xen: introduce DEFINE_SYMBOL

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 26.02.19 at 19:43, <sstabellini@xxxxxxxxxx> wrote:
> On Tue, 26 Feb 2019, Ian Jackson wrote:
>> Jan Beulich writes ("Re: [PATCH v10 2/6] xen: introduce DEFINE_SYMBOL"):
>> > On 26.02.19 at 17:46, <ian.jackson@xxxxxxxxxx> wrote:
>> > > I am not aware of a standard C type which could be used instead of
>> > > this struct.  But I think you can use the `packed' attribute to get
>> > > the right behaviour.  The GCC manual says:
>> > > 
>> > >  | Alignment can be decreased by specifying the 'packed' attribute.
>> > >  | See below.
>> ...
>> > Until I've looked at this (again) now, I wasn't even aware that
>> > one can combine packed and aligned attributes in a sensible
>> > way. May I suggest that, because of this being a theoretical
>> > issue only at this point, we limit ourselves to the build time
>> > assertion you suggest?
>> 
>> I am not suggesting combining `packed' and `aligned'.  I am suggesting
>> only `packed' (but based on text which is in the manual section for
>> `aligned').  But I am happy with a build-time assertion if you don't
>> want to add `packed'.  That is just as safe.
> 
> Could you please provide a rough example of the build-time assertion you
> are thinking about? I am happy to add it.

BUILD_BUG_ON(alignof(*s1) != alignof(*s2));

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel