[Bug exp/26875] Incorrect value printed for address of first element of zero-length array

["simark at simark dot ca" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=26875

--- Comment #12 from Simon Marchi <simark at simark dot ca> ---
(In reply to Tom de Vries from comment #11)
> (In reply to Simon Marchi from comment #7)
> > I am not able to reproduce.  In my version of the compiled test case, both
> > bounds are const:
> > 
> > (top-gdb) p type->bounds ()->low.kind ()
> > $4 = PROP_CONST
> > (top-gdb) p type->bounds ()->high.kind ()
> > $5 = PROP_CONST
> > 
> > What compiler do you use to build the test case?
> 
> Gcc 7.5.0

Thanks, I see the failure with gcc 7.

> > So I presume GDB translates that to two constant bounds with value 0.  What
> > does it look like in yours?
> > 
> > I stumbled on a similar problem and filed this bug here, they are likely
> > related:
> > 
> > https://sourceware.org/bugzilla/show_bug.cgi?id=26901
> > 
> > I have this patch series in the pipeline that I made to address 26901:
> > 
> > https://review.lttng.org/c/binutils-gdb/+/4399/4
> > 
> 
> Ah, thanks for the pointer, I'll take a look.

The patches are not commented at all now, but the idea is to be able to fetch
the low and high bounds separately, not all or nothing like it is now.  Here,
we only care about the low bound, so we only get that one, and ignore the high
bound (which is unknown).

> 
> > If/when I am able to reproduce this (26875) bug, I can check if that helps
> > with it too.
> 
> That would be great.

The failure goes away with these patches applied.

The only thing that bugs me is that some part of the code assumes that a bound
of "-1" means an unknown bound.  It seems to me that if we use gdb::optional,
we don't need that: an empty optional means an unknown bound.  So we would need
to go through all the users to see we if need the "set to -1" vs "empty
optional" disctinction or not.

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