[Bug default/26646] unexpected declaration-only types

["gprocida at google dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=26646

--- Comment #30 from gprocida at google dot com ---
Hi Dodji.

On Mon, 28 Feb 2022 at 10:01, dodji at redhat dot com
<sourceware-bugzilla@sourceware.org> wrote:
>
> https://sourceware.org/bugzilla/show_bug.cgi?id=26646
>
> --- Comment #28 from dodji at redhat dot com ---
> (In reply to gprocida from comment #26)
>
> > > If the regression disappears, I guess we can say that the current state is
> > > an improvement.  If so, I'll clean-up and post the patches on the list, if
> > > you agree.  Further investigation will continue after that.
> >
> > Agreed. Thanks!
>
> I have posted the cleaned-up patch for review at
> https://sourceware.org/pipermail/libabigail/2022q1/004179.html.
>

I've thought a little bit more about this one.

The current check is "local", recursion prevention at *this* DIE
prevents it from being canonicalised, but it could still depend on
child DIEs that are actually also parent DIEs and whose processing has
not yet been completed.

Would it be safer (more precise) to inhibit on-the-fly
canonicalisation whenever the set (stack) of aggregates being compared
is non-empty?

Giuliano.

> Thanks.
>
> --
> You are receiving this mail because:
> You are on the CC list for the bug.

-- 
You are receiving this mail because:
You are on the CC list for the bug.