comp.lang.ada
 help / color / mirror / Atom feed
From: "Dan'l Miller" <optikos@verizon.net>
Subject: Re: GNAT LLVM
Date: Sat, 21 Jul 2018 18:36:58 -0700 (PDT)
Date: 2018-07-21T18:36:58-07:00	[thread overview]
Message-ID: <1f2d1a97-7aa9-4d74-9e9b-4231e5a489b8@googlegroups.com> (raw)
In-Reply-To: <6d8d45a1-2262-4108-9db3-c958b7f6f92d@googlegroups.com>

On Saturday, July 21, 2018 at 8:17:05 PM UTC-5, Roger wrote:
> On Sunday, July 22, 2018 at 10:53:35 AM UTC+10, Luke A. Guest wrote:
> > Roger wrote:
> > > On Saturday, July 21, 2018 at 11:53:11 PM UTC+10, Simon Wright wrote:
> > >> An interesting hint in the gprbuild repo:
> > >> 
> > >> https://github.com/AdaCore/gprbuild/commit/eb94de5bfcc5c5ddd36f701cfd43e542a96c118f
> > > 
> > > If using gprbuild on mac, make sure to use the latest repo.
> > > See:
> > > https://github.com/AdaCore/gprbuild/issues/44
> > > 
> > 
> > I don’t know what you’re responding to, but it’s not really this.
> 
> I was responding to Simon’s comment which concerns Adacore’s gprbuild tool.
> The version of gprbuild shipped with GNAT 2012 fails on Mac OSX as it does not fully support LLVM.
> My comment concerns gprbuild with respect to this problem.
> An "An interesting hint " is that if you want to use gprbuild with LLVM you need to use the latest distro of
> gprbuild.
> See, the last few items of https://github.com/AdaCore/gprbuild/issues/44n which is titled "
> C++ code fails under GNAT 2018”.  The reason for the failure turned out to be a gprbuild bug involving
> gprconfig which did not recognise the LLVM C++ compiler.
> An "An interesting hint " is 
> The latest gprbuild bistro does not have this bug (Simon's hint I think).

Thank you very much, Roger.  Your reply was actually very much more on-topic & a-propos than anyone (including myself) realized at first.  It seems that Simon's hint evidently is actually about interworking GIMPLE/RTL-backended GNAT's generated code with LLVM-backended Clang's generated code.  It seems that Simon's hint is evidently not pursuant to a full-fledged LLVM-backended GNAT.

(Can anyone imagine the reaction from RMS if the official FSF GNAT maintainers at AdaCore were to check in the bulk of LLVM into FSF repositories for an LLVM-backended GNAT to comply with the intermediate-representation prohibitions/restrictions in GPLv3's Runtime Library Exception?  I think that some wild bulls in rodeo rings might be more gentle & tame.)

  reply	other threads:[~2018-07-22  1:36 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-07-21 13:53 GNAT LLVM Simon Wright
2018-07-22  0:35 ` Roger
2018-07-22  0:53   ` Luke A. Guest
2018-07-22  1:17     ` Roger
2018-07-22  1:36       ` Dan'l Miller [this message]
2018-07-22  2:52 ` olivermkellogg
replies disabled

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox