Re: [PATCH 1/3] Introduce gdb::unique_ptr

[Jan Kratochvil <jan dot kratochvil at redhat dot com>]

On Wed, 12 Oct 2016 12:51:16 +0200, Eli Zaretskii wrote:
> > Jan said, in full:
> > 
> > > The discussion is about C++11.  LLVM+LLDB have switched to C++11 in 2014 and
> > > they haven't looked back.
> > > 
> > > I see the C++11 discussion pointless, where is the system which really needs
> > > GDB and which still cannot compile C++11?  Why to waste manyears on bugs which
> > > can no longer exist with C++11?
> > > 
> > > The discussion should be when to switch to C++17 as that removes another tons
> > > of crap like gnulib.
> > 
> > He's saying that we should just require C++11 and be done with it.
> > And then he concluded with a (hopefully tongue-in-cheek) remark about
> > C++17, which (hopefully) is obvious we're not going to be requiring
> > that anytime soon...
> 
> I have no reason to be sure that was tongue-in-cheek.  And I have no
> reason to regard as obvious that no one will be requiring C++17 any
> time soon, not without anyone, nor our coding standards, saying that.
> 
> So from my POV, that was no straw man argument at all.  Jan is one of
> the more influential developers here, so his opinions certainly have a
> significant weight with me.  I regard things that he writes very
> seriously.

I really think the C++11 discussion is pointless, everything everywhere does
support C++11 now.

My C++17 sentence was meant seriously - but literally as it is written - that
I do not expect GDB should really start requiring C++17 right now.  But in
a few years GDB should really start requiring C++17 as it will increase GDB's
developers productivity similarly to how C++11 will/would now.
Please notice C++17 is really offtopic for this mail thread now.


Nowadays software is no longer distributed on floppy disks by snail mail.
I do not see a reason for artificial requirements of "5 years".  Fixes are
being delivered by automated update systems in days and new releases with
integration testing happen in months.


Jan