[LLVMdev] How do I disable --version-script when cross compiling clang/LLVM on a Mac?

[Alp Toker]

On 22/12/2013 01:33, Richard Pennington wrote:
> On 12/21/2013 05:21 PM, Alp Toker wrote:
>>
>> On 21/12/2013 22:51, Richard Pennington wrote:
>>> Hi,
>>>
>>> I asked this question a few days ago and didn't get any responses. I 
>>> thought I'd try again with a little more detail.
>>>
>>> I am trying to cross compile my clang/LLVM based ELLCC cross 
>>> development tools project (http://ellcc.org)
>>
>> Hi Richard,
>>
>> Nobody replied because your question lacks immediate relevance to 
>> this mailing list.
>>
>> You may have more luck if you pose your question in a way that 
>> relates to, and can be tried using stock LLVM/clang.
>>
>> If you think there's a bug, you can also file a bug report at 
>> http://llvm.org/bugs/ including the steps needed to reproduce the issue.
>>
>> Feel free to CC me in, hope we can set you in the right direction.
>>
>> Alp.
>>
> Hi Alp,
>
> I hope you don't think I was whining about the lack of response, I 
> wasn't. This list has always been extremely helpful. I understand that 
> my use case is a bit out the norm.
>
> I was just hoping that someone more familiar with the clang/LLVM cross 
> build rules might be able to shed a little insight. I suspect that 
> there aren't too many people using clang/LLVM to cross build for Linux 
> on a Mac. The normal cross build on a Linux box for another Linux 
> target works just fine.

That last bit gave a better idea of what you're aiming for. I think I 
wasn't the only one confused by the original mail as it takes a stretch 
of the imagination for those not using your toolchain(!)

You'll have to test this yourself, but assuming you're on the Makefile 
build system and already set up correctly for a cross-compile, this 
might be the cause:

The variable HAVE_LINK_VERSION_SCRIPT determines how the linker is 
called, and is detected at configure time by autoconf/m4/link_options.m4.

Some of the Makefile.rules files have an additional ifeq 
($(HOST_OS),Darwin) check surrounding use of that flag.

That check may be confusing the host and target platforms when building 
up linker flags, and if so you've found a corner-case bug in the build 
system. I'd suggest hard-coding values in a few of those places to see 
if you can get the right flags passed depending on whether a host or 
target binary is getting built.

Alp.

-- 
http://www.nuanti.com
the browser experts