[CMake] Recommended Multilib Build Approach?

[Xavier Besseron]

Hello Greg,


To build i386 binaries on my 64-bit system with multilib, I just do
something like this:

mkdir build-i386
cd build-i386
CFLAGS="-m32" CXXFLAGS="-m32" LDFLAGS="-melf_i386" cmake ../src/

And to build x86-64 binaries on my 64-bit system,  I don't need to
specify anything since it is the default behavior.


Xavier

On Mon, Jul 30, 2012 at 6:43 PM, Gregory Peele ARA/CFD <gpeele at ara.com> wrote:
> From: Alexander Neundorf [mailto:a.neundorf-work at gmx.net]
> Sent: Saturday, July 28, 2012 11:48 AM
> To: cmake at cmake.org
> Cc: Gregory Peele ARA/CFD
> Subject: Re: [CMake] Recommended Multilib Build Approach?
>
> On Thursday 21 June 2012, Gregory Peele ARA/CFD wrote:
>>> Hi all,
>>>
>>> I want to be able to build 32-bit and 64-bit from the same GCC multilib
>>> install (currently for MinGW-w64, though this also applies for Linux/Mac
>>> GCC and LLVM). To clarify, I want to be able to do two completely
>>> separate builds in separate binary dirs - building fat binaries from
>>> multiple architectures in the same binary dir is a separate problem I'm
>>> also interested in for Android, but I'm not worrying about that yet.
>>>
>>> What is the "proper" approach to building the non-default arch in a
>>> multilib setup? Obviously using CMake as-is will build the default arch
>>> just fine.
>> I think there are no recommendations yet. I'm not aware that somebody else already tried to do this.
>> Can the executables built for the non-default architecture be executed on the hopst where you are building ?
>> If so, there should be a way :-)
>> If not, this sounds like somewhat like crosscompiling.
>> Alex
>
> I haven't explored the topic any further since posting.  On a multilib system, all multilib architectures are "native" to the host and can be executed.  Basically the problem I wanted to solve was getting the x86-64 GCC or LLVM compiler to produce 32-bit x86 code in an easy and non-intrusive way.
>
> Right now I'm forcibly adding the "-m32" flags to all relevant CMake variables within my top-level CMake script if CMAKE_SYSTEM_PROCESSOR is set to x86 or i686, and conversely the "-m64" flags if set to x86-64 or AMD64.  It seems like there should be a less intrusive way of doing it, and I'm not sure if this skews some early compiler/platform tests prior to me overriding the flags.  Visual Studio doesn't have this issue since the 32-bit and 64-bit builds are entirely separate CMake generators - a strange decision, in my opinion, but easy enough to work with.
>
> Greg
> --
>
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.cmake.org/mailman/listinfo/cmake