On Fri, Jan 17, 2025 at 8:02 AM Qing Zhao <qing.z...@oracle.com> wrote:
> Joseph, Kees and Bill,
>
> I need your input on this.
> > On Jan 17, 2025, at 10:12, Martin Uecker <uec...@tugraz.at> wrote:
> >
> > Am Donnerstag, dem 16.01.2025 um 21:18 +0000 schrieb Qing Zhao:
> >>
> > ..
> >>
> >> Although in the previous discussion, I agreed with Martin that we should
> >> use the
> >> designator syntax (i.e, counted_by (.n) instead of counted_by (n)) for the
> >> counted_by attribute for pointer fields, after more consideration and
> >> discussion
> >> with Bill Wendling (who is working on the same work for CLANG), we decided
> >> to
> >> keep the current syntax of FAM for pointer fields. And leave the new
> >> syntax (.n)
> >> and more complicate expressions to a later work.
> >>
> > I think this would be a mistake. Once you have established the confusing
> > syntax,
> > it will not easily go away anymore. So I think you should use the
> > unambiguous
> > and future-prove syntax right away. Support for more complicated
> > expressions
> > could be left for later, of course.
>
> Personally I agree with you. -:)
>
> Actually we might need to use such syntax in the very beginning when adding
> counted_by of FAM.
> As I know, linux kernel community asked for the following new feature for
> counted_by of FAM:
>
> struct fs_bulk {
> ...
> int len;
> ...
> }
>
> struct fc_bulk {
> ...
> struct fs_bulk fs_bulk;
> struct fc fcs[] __counted_by(fs_bulk.len);
> };
>
> i.e, the “counted_by” field is in the inner structure of the current
> structure of the FAM.
> With the current syntax, it’s not easy to extend to support this.
>
> But with the designator syntax, it might be much easier to be extended to
> support this.
>
> So, Kees and Bill, what’s your opinion on this? I think that it’s better to
> have a consistent interface between GCC
> and Clang.
>
> Joseph, what’s your opinion on this new syntax? Shall we support the
> designator syntax for counted_by attribute?
>
I've been thinking more about what we discussed regarding this and I
have to agree with Martin that putting off using the new syntax for
both FAMs and pointers will make things worse in the long run. We'll
already have to deprecate the current syntax in compilers for code
using the current syntax, but it may not be too late. Linux is pretty
much the main user of this feature, and we have pretty good control
over how it's used there.
The one wrinkle is that Apple allows for full expressions, not just a
field designator---and they're currently using full expressions.
AFAIK, that code is internal to Apple, but they want to externalize
it. So we'll eventually have to come up with some sort of plan for how
to deal with it probably sooner rather than later. I'll open up a
discussion with them about it.
-bw
