[llvmlinux] Test IAS patch should compile kernel (take 2)

Tinti viniciustinti at gmail.com
Sun Feb 23 22:11:42 UTC 2014


On Sun, Feb 23, 2014 at 1:33 PM, Saleem Abdulrasool
<compnerd at compnerd.org>wrote:

> On Sun, Feb 23, 2014 at 4:53 AM, Tinti <viniciustinti at gmail.com> wrote:
>
>> On Sun, Feb 23, 2014 at 2:58 AM, Saleem Abdulrasool <
>> compnerd at compnerd.org> wrote:
>>
>>> On Sat, Feb 22, 2014 at 6:01 PM, Tinti <viniciustinti at gmail.com> wrote:
>>>
>>>> On Sat, Feb 22, 2014 at 7:21 PM, Renato Golin <renato.golin at linaro.org>wrote:
>>>>
>>>>> Folks,
>>>>>
>>>>> It looks like Saleem has fixed the last bug in the IAS to make it
>>>>> compiler the kernel. It breaks some odd driver end on MIPS, but we
>>>>> should be able to test compiling the ARM kernel regardless.
>>>>>
>>>>> Vinicius, can you apply patch svn 201952 on current tree (or checkout
>>>>> trunk now, if Saleem didn't revert it because the failures), and test?
>>>>>
>>>>> He tested on QEMU, so it should boot the vexpress kernel. Would also
>>>>> be good to know if it boots on the Nexus 7 before we announce to the
>>>>> world.
>>>>>
>>>>> cheers,
>>>>> --renato
>>>>>
>>>>
>>>> Hi Renato and Saleem,
>>>> Unfortunately it did not build in my case in vexpress and I have not
>>>> tried N7 yet.
>>>> I am using svn 201963. Do you want me to use svn 201952?
>>>>
>>>
>>> 201963 is perfect.  It requires an additional patch which I have
>>> attached to this message.  The generation of the assembly offsets generates
>>> invalid assembly, which LLVM will flag.
>>>
>>> From the logs it seems to be the missing piece (and explains why it
>>> built for me).
>>>
>>> Thanks for testing this!
>>>
>>> Please find attached the log.
>>>>
>>>> ==
>>>> cmake version 2.8.11.2
>>>> gcc (Ubuntu/Linaro 4.8.1-10ubuntu9) 4.8.1
>>>> git version 1.8.3.2
>>>> GNU Make 3.81
>>>> quilt version 0.60
>>>> LLVM version 3.5svn commit c7ee77714925ef2f5d5ce7c2cbb0f39171abc9ae
>>>> LLVM version 3.5svn
>>>> clang version 3.5 commit 8c702d1f447aa6f33370756b7e53b8f1529cef58
>>>> clang version 3.5
>>>> src/linux version 3.14.0-rc3 commit
>>>> 0f0ca14386e0431fbedaae5efc550d46cf93b9cf
>>>> src/linux-gcc version 3.14.0-rc3 commit
>>>> 0f0ca14386e0431fbedaae5efc550d46cf93b9cf
>>>> buildroot version 2014.02-rc2 commit
>>>> 70ee9fcdfcf6fc7cb214e454afe55cbffec84621
>>>> LTP version 20120614 (from sourceforge)
>>>> QEMU version 1.6.2 commit e82ee0845c3240541e79b9bf21779b3f8743f1b4
>>>> arm-none-linux-gnueabi-gcc (Sourcery CodeBench Lite 2013.05-24) 4.7.3
>>>> ==
>>>>
>>>> Regards,
>>>> Vinicius
>>>>
>>>> --
>>>> Simplicity is the ultimate sophistication
>>>>
>>>
>>> --
>>> Saleem Abdulrasool
>>> compnerd (at) compnerd (dot) org
>>>
>>
>> Still some missing parts. I am a bit busy today but I am free to dig it
>> more at night.
>> I saw some issues related to flags that were already discussed on this
>> ML. And some other related for example to the "BUG()" macro.
>> Please let me know if I can help in something else.
>>
>> Please find attached the log of the building with your patch.
>>
>
> Oh, bleh.  It seems that there is a patch that disables the IAS by
> default.  Ill look into some of the failures.
>
> The BUG() macro is fixed as of SVN r201968.
>
> Im not sure what to do about the VFP macros.  I believe that the mrc
> equivalents are invalid in UAL, so they need to be swapped to UAL syntax.
>  I do have equivalents for it, that I can provide a patch for.
>
>
>> Regards,
>> Vinicius
>>
>> --
>> Simplicity is the ultimate sophistication
>>
>
>
>
> --
> Saleem Abdulrasool
> compnerd (at) compnerd (dot) org
>

Yes, please. As soon as you have something to test mail me.

Thanks for the work!

-- 
Simplicity is the ultimate sophistication
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfoundation.org/pipermail/llvmlinux/attachments/20140223/1df21076/attachment.html>


More information about the LLVMLinux mailing list