[llvm-dev] Building SVN head with CMake - shared libraries?

[Daniel Sanders via llvm-dev]

> -----Original Message-----
> From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of Joerg
> Sonnenberger via llvm-dev
> Sent: 14 January 2016 21:02
> To: llvm-dev at lists.llvm.org
> Subject: Re: [llvm-dev] Building SVN head with CMake - shared libraries?
> 
> On Thu, Jan 14, 2016 at 11:31:55AM -0800, Chris Bieneman via llvm-dev
> wrote:
> >
> > > On Jan 14, 2016, at 11:22 AM, Mehdi Amini <mehdi.amini at apple.com>
> wrote:
> > >
> > >>
> > >> On Jan 14, 2016, at 9:38 AM, Chris Bieneman via llvm-dev <llvm-
> dev at lists.llvm.org> wrote:
> > >>
> > >>
> > >>> On Jan 14, 2016, at 5:18 AM, Dan Liew <dan at su-root.co.uk> wrote:
> > >>>
> > >>> On 14 January 2016 at 11:24, David Jones via llvm-dev
> > >>> <llvm-dev at lists.llvm.org> wrote:
> > >>>> Thanks - I'll try this tonight.
> > >>>>
> > >>>> Assuming it works, should these variables be added to the docs at
> > >>>> http://llvm.org/docs/CMake.html ?
> > >>>
> > >>> Yes.
> > >>
> > >> It probably makes sense to add LLVM_BUILD_LLVM_DYLIB, but I would
> discourage adding BUILD_SHARED_LIBS to that document.
> BUILD_SHARED_LIBS is rarely what people actually want, and a lot of people
> gravitate to it because the wording is similar to the autoconf —enable-shared
> flag.
> > >
> > > Considering that BUILD_SHARED_LIBS is rarely what people want,
> wouldn’t it make sense to document it to explicitly discourage people from
> using it and avoid them surprising behavior?
> >
> > Honestly, I’m not sure *anyone* actually wants that behavior, and I’d kinda
> like to remove the option entirely.
> 
> It is exactly what I want to have for debug builds. It means linking
> moderately chunked blocks of code together and can even parallelize that
> easily. I agree that it is not an end user option, but for developers it
> is helpful exactly as it is.
> 
> Joerg

+1. I use it because it makes a huge difference to build time on incremental debug builds. I generally find that the build time saving is worth a lot more than the run-time cost.