lldb on sparc64

On Tue, Nov 27, 2018 at 12:45:36PM +0100, Jeremie Courreges-Anglas wrote:
> On Tue, Nov 06 2018, Stuart Henderson <stu@spacehopper.org> wrote:
> > On 2018/11/06 11:11, John Gould wrote:
> >> Hello, I am trying to build parts of xfce4 and some kde4 applications on
> >> 6.4 current on sparc64.
> >> Although these applications worked or were available from packages on 6.3 I
> >> am having no
> >> luck on 6.4. I have several sparc machines here all doing nothing! Can
> >> someone please help me with this or am I wasting my time? I've included a
> >> dmesg below and some of the output of a recent build.
> >>
> >> Kind regards John
> >>
> >> /usr/ports/pobj/llvm-6.0.1/llvm-6.0.1.src/tools/lldb/include/lldb/Host/Editline.h:49:19:
> >> fatal error: codecvt: No such file or directory
> >>
> >> #include <codecvt>
> >>
> >> ^
> >
> > This is exactly the reason why these are not available in packages on
> > 6.4. The patch below for devel/llvm should get things unblocked though I
> > don't know if it will help get you as far as the ports you're really
> > interested in.
>
> Thanks. Diff refreshed for -current, successfully tested on sparc64
> (and amd64).
>
> IMO this has been broken since too long already. I'd like us to fix
> what we can now instead of waiting for a switch to gcc 6 in the upcoming
> weeks/months/releases. Stuart, if you want to commit this, ok jca@

Brad, you're still listed as the maintainer of devel/llvm, and you proxy
commits to it via sthen@ and ajacoutot@, so i understand that you still
work on it. devel/llvm has been broken on sparc64 (and maybe macppc,
we'll see) since you rushed the 'enable lldb' bits before 6.4, which
resulted in the low amount of sparc64 release packages for 6.4, and that
affects actual users of the platform.

I've loudly complained about it, sthen@ proposed a variation of the diff
below that i've successfully tested in one of the past bulks, and i
think you know it.

Now, i know that "switching to gcc6" also fixes it, but it requires more
work before being a short-time goal, so to unblock the situation, will
you step up and give your opinion on jca@'s diff so that we can move
forward ?

I'm starting a sparc64 bulk without any diff. As a result, it wont have
llvm & any of the packages depending on it, which i think is better:
If llvm is available, all c++ ports end up depending on gcc (for
libstdc++) *and* llvm because many (340-so) ports have
COMPILER=base-clang ports-clang ports-gcc which is deeply wrong, but
this has to be fixed separately from the lldb issue. For which as
MAINTAINER, you have to step up.

Thanks.

Landry