Required VS compiler

[Thomas Friedrichsmeier]

Hi,

sorry for the late reply.

On Sat, 26 Mar 2016 16:07:03 +0100
Dominik Haumann <dhaumann at kde.org> wrote:
> On Wed, Mar 23, 2016 at 7:51 AM, Thomas Friedrichsmeier
> <thomas.friedrichsmeier at ruhr-uni-bochum.de> wrote:
> > On Mon, 21 Mar 2016 22:07:57 +0100
> > Dominik Haumann <dhaumann at kde.org> wrote:
> [...]
> >> Therefore, I would like to propose v2015 as required compiler
> >> and Qt5.6 as required Qt version on Windows, or are there
> >> reasons against it?
> >
> > I'll put it the other way around. Annoying as it is, does that
> > QStringLiteral nuisance justify upping the requirements? Factually
> > that _is_ going to happen as the MSVC 2013 (and 2012) builds are
> > going to get less testing. But I don't think that is a state we
> > should actively rush into. Increased requirements always come with
> > an increased cost.
> 
> True, and yes, the Windows version already now has very few testers, I
> don't think we can maintain that in any sane way.
> 
> But true, we can wait a bit more.

I think what we need / want is probably a graded level of support for
compilers / libs. Something like:

- Recommended: Compiler / lib version that is expected to "just
  work" (for the most part). This will be what developers and users
  will be encouraged to use, unless and until they have specific
  reasons not to. Developers are highly encouraged to test their work
  against this version.
- Supported: Compiler / lib version that will work with a bit of luck
  and effort. This is going to be less tested and not
  generally recommended, but:
  - Compilation or runtime errors with this version shall be treated as
    true bugs.
  - Patches to make this work shall not be rejected for purely
    aesthetical reasons.
  - Developers shall not _knowingly_ commit code that will break with
    this version, or remove code supporting this version.
- Unsupported: We don't mind if it works, but developers will not be
  expected to do anything to make it work. Developers are encouraged to
  drop patches targeting this version as part of their annual spring
  cleaning.

I read your initial mail as suggesting to shove MSVC 2013 straight
into "unsupported", and I think that would not be a good idea (not even
in six months time). No objections to saying MSVC 2015 / Qt 5.6 is
recommended, though.  (Yet, I would actually like to see some
version of MinGW inside the "recommended" set, too.)

Regards
Thomas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
URL: <http://mail.kde.org/pipermail/kde-core-devel/attachments/20160402/545a3307/attachment.sig>