[LLVMdev] Native build clarification

罗勇刚(Yonggang Luo) luoyonggang at gmail.com
Sun May 15 19:37:31 PDT 2011


It's looks the progress is very slow, is there any problem here?


2011/5/15 Hans Stimer <hans.stimer at gmail.com>

> It looks like it is kind of stalled:
>
> ~/extern/llvm.head/lib/Object $ svn log
>
>
>
> On Sat, May 14, 2011 at 11:07 AM, Hans Stimer <hans.stimer at gmail.com>wrote:
>
>> It looks like someone got started on this project:
>>
>> http://www.llvm.org/devmtg/2010-11/Spencer-ObjectFiles.pdf
>>
>> Does anyone know what kind of progress has been made?
>>
>>
>> On Sat, May 14, 2011 at 9:10 AM, Reed Kotler <rkotler at mips.com> wrote:
>>
>>>  It's not so easy to create a linker that has all the bells and whistles,
>>> especially on windows.
>>> You have to look at the totality of options, linker files, shared
>>> objects, etc.
>>> It's not a weekend project.
>>>
>>> If you want to read the elf manual and gnu ld and then write a new one
>>> you could do that. Clang can easily be fixed to incorporate it either as a
>>> default or as an optional linker. The GNU ld for mingw must know a lot about
>>> windows linking but probably not all.
>>>
>>> In some worlds Coff format is still relevant.
>>>
>>> Microsoft has their own form of Coff that they use even today.
>>>
>>> Back in the early days of Windows NT I wrote one for that OS because
>>> Microsoft stopped distributing one with the OS and spent a lot of time
>>> reverse engineering Microsofts until I could really do all that it could do.
>>> Traditionally the OS manufacturer wrote the linker and did not document
>>> things well enough for someone else to.
>>>
>>> Clang is not just for linux; people are going to build embedded bare
>>> metal apps and the OS itself (linux for example) and need many obscure
>>> features.
>>>
>>> I think writing a new elf linker for linux and bare metal is possible but
>>> I would forget about supporting COFF and Microsoft format and other OS's
>>> like BSD or at least leave that open for someone else to add that can really
>>> add all the bells and whistles for that OS.
>>>
>>> reed
>>>
>>>
>>> On 05/14/2011 07:35 AM, Justin Holewinski wrote:
>>>
>>> On Fri, May 13, 2011 at 2:20 PM, Hans Stimer <hans.stimer at gmail.com>wrote:
>>>
>>>> It appears that when you link bitcode files into a native executable
>>>> using llvm-ld, that gcc is used for the final step.
>>>>
>>>>  I would like to create a tool that creates native executables without
>>>> requiring gcc/clang. Is that currently supported with llvm?
>>>>
>>>
>>>  To create executables, LLVM relies on the system linker (gcc, visual c,
>>> etc.).  Even Clang uses the system linker for the final link step.
>>>
>>>
>>>>
>>>> _______________________________________________
>>>> LLVM Developers mailing list
>>>> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>>>>
>>>>
>>>
>>>
>>> --
>>>
>>> Thanks,
>>>
>>>  Justin Holewinski
>>>
>>>
>>> _______________________________________________
>>> LLVM Developers mailing listLLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.eduhttp://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>>>
>>>
>>>
>>> _______________________________________________
>>> LLVM Developers mailing list
>>> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>>>
>>>
>>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
>


-- 
         此致
礼
罗勇刚
Yours
    sincerely,
Yonggang Luo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20110516/0e935b24/attachment.html>


More information about the llvm-dev mailing list