[llvm-dev] Retiring VS2015 Support

Simon Pilgrim via llvm-dev llvm-dev at lists.llvm.org
Wed Jun 19 15:26:30 PDT 2019

On 19/06/2019 22:05, De Azevedo Piovezan, Felipe via llvm-dev wrote:
> Isn't this already covered by the compiler upgrade plan that was under discussion a while ago? The plan covers all compilers, not only VS.
Yes, but I see no problem with bumping the VS requirements earlier - 
there's nothing binding GCC + VS minimum requirements together. And as I 
pointed out this will let us start removing some VS compatibility hacks 
in time for the 9.0 release cycle, plus we no longer seem to have 
little/no build+test coverage for VS2015.
> If I understand correctly, VS2015 is already in the "soft error" range, and it will trigger a hard CMake error next time we bump the minimum requirements. Last I heard, we were waiting for Google to finish some internal work.

That's what I'd vaguely meant by the C++14 migration - I might have been 
mistaken but I thought that was due to Google's dependency on older GCC 
versions not VS?

> -----Original Message-----
> From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of Aaron Ballman via llvm-dev
> Sent: Wednesday, June 19, 2019 3:38 PM
> To: Simon Pilgrim <llvm-dev at redking.me.uk>
> Cc: llvm-dev <llvm-dev at lists.llvm.org>
> Subject: Re: [llvm-dev] Retiring VS2015 Support
> On Tue, Jun 18, 2019 at 4:11 PM Simon Pilgrim via llvm-dev <llvm-dev at lists.llvm.org> wrote:
>> AFAICT we traditionally only support building llvm/clang with the 2
>> most recent releases of Visual Studio.
>> With the release of VS2019, what do people think of retiring VS2015
>> support before the next major llvm/clang release?
> +1 from me.
> ~Aaron
>> As well as simplifying C++14 migration (admittedly only a little),
>> this should also let us remove a number of hacks (e.g. the
>> alignas/AlignedCharArray hack in AlignOf.h), as well as re-enabling a
>> number of warnings in HandleLLVMOptions.cmake.
>> I haven't noticed any build bots still referring to VS2015 (at least
>> not by name).
>> Simon.

More information about the llvm-dev mailing list