Re: GDB 8.2 release 2018-08-26 status update

[Keith Seitz <keiths at redhat dot com>]

On 08/26/2018 03:11 AM, Joel Brobecker wrote:
> As far as I can tell, there is only PR still left pending:
> 
>   * [KeithS] PR symtab/23010
>     Regression: Assertion: SYMBOL_LANGUAGE (sym) == DICT_LANGUAGE (dict)->la_language
>     https://sourceware.org/bugzilla/show_bug.cgi?id=23010
> 
> 	This regression was seemingly fixed, and then makes a come back.
>         I'm still not sure whether it's the same issue, or rather
>         a different one that lead to the same symptom, though.
> 
> At this point, I think we should only wait so long for a fix to this
> one, and consider the idea of releasing what we have at the moment,
> without PR symtab/23010 completely fixed if necessary. The release
> is getting very late, and this problem was both present in 8.0 and
> I believe mitigated in the 8.2 branch. So we can afford to defer
> the rest of the fix in 8.3.

Yes (re: mitigation).

As far as I know, the only reported examples of the bug still are:
1) non-DWARF systems
2) LTO executables
3) Possibly a Fedora bug demonstrating another path (or could be LTO
   again)

I am working on a fix for #2, but I don't think that is going to come very
timely, but it is generally well-known (I'm told) that debugging with LTO
is problematic. [Not that we shouldn't make it work, but it is right now
already a difficult proposition.]

#1 will undoubtedly involve audting the code and rolling some dice.
Fortunately we have some access to such systems in the compile farm
which should make any auditing much easier and deterministic.

I would urge maintainers to consider a point release containing any fix
should the problem prove so prevalent/widespread over holding up the
current release.

My $0.002,
Keith