[PATCH] MINGW64 Driver, TLS and SEH exception

Yaron Keren yaron.keren at gmail.com
Fri Apr 18 03:20:04 PDT 2014


TLS in

  http://reviews.llvm.org/D3421

SEH clang in

  http://reviews.llvm.org/D3419

MinGW toolchain in

  http://reviews.llvm.org/D3420



2014-04-16 1:18 GMT+03:00 Martell Malone <martellmalone at gmail.com>:

> Hi Yaron,
>
> Yes I've update most of this myself without the updated patch from kai.
> That's why is wasn't as clean.
> Thanks for pointing out the one I missed.
> I've just rebuilt and seh looks like it now works properly for mingw32
> x86_64.
>
>
> Thanks
> Martell
>


On Tue, Apr 15, 2014 at 9:11 PM, Yaron Keren <yaron.keren at gmail.com> wrote:

> Hi Martell,
>
> The page I linked is Kai Nacke page, note that the patch there is more
> updated than the 20140201 one in several locations, major changes are
>
> 1) the test for MingW environment is now
>
> if (is64Bit && Triple.isWindowsGNUEnvironment())
>
> the previous test wil cause a silent bug as the OS can't be MinGW anymore
> and the test is always false.
>
> 2) isPrologLabel() was replaced by by isCFIInstruction()
>
> the previous patch will cause a compilation failure against trunk,
> isPrologLabel() does not exist anymore.
>
> 3) Small code changes in X86MCAsmInfoGNUCOFF::X86MCAsmInfoGNUCOFF(const
> Triple &Triple) and more.
>
> The new patch (20140414) should apply to trunk.
>
> Yaron
>
>
>
>
> 2014-04-15 22:11 GMT+03:00 Martell Malone <martellmalone at gmail.com>:
>
> Yaron
>>
>> After looking at the code differences from the updated seh patch you
>> linked me from kai
>> I have noticed that the only change from my patch is that I am using
>>
>> X86RegisterInfo
>> instead of
>> TargetRegisterInfo
>>
>> So his patch looks good to me other than calling that virtual method
>> which would cause a compile error.
>>
>> On a side note doxygen is messing up the X86RegisterInfo class name
>> because of the final word
>> http://llvm.org/doxygen/classllvm_1_1final.html
>>
>>
>>
>>
>>
>
> On Tue, Apr 15, 2014 at 7:59 PM, Martell Malone <martellmalone at gmail.com>wrote:
>
>> Hi Reid,
>>
>> TLS works for me with clang and the mingw64 headers to I assume that it
>> is ready within LLVM.
>> I have just submitted a patch to llvm maillist from kai of redstar.defor the seh that I updated to work on llvm-head.
>>
>> With regard to the _WIN32 ifdefs a sysroot mingw32 compiler should never
>> be generated on windows because that is the native platform.
>> We have all switched from doing that over at the mingw64 camp.
>>
>> In the patch I have removed the Generic_GCC toolchain and replaced it
>> with the MINGW Driver.
>>
>> @@ -1967,10 +1970,7 @@
>>        case llvm::Triple::GNU:
>>          // FIXME: We need a MinGW toolchain.  Use the default Generic_GCC
>>          // toolchain for now as the default case would below otherwise.
>> -        if (Target.isOSBinFormatELF())
>> -          TC = new toolchains::Generic_ELF(*this, Target, Args);
>> -        else
>> -          TC = new toolchains::Generic_GCC(*this, Target, Args);
>> +        TC = new toolchains::MinGW(*this, Target, Args);
>>
>> Many Thanks
>> Martell
>>
>>
>>
>> On Tue, Apr 15, 2014 at 7:51 PM, Reid Kleckner <rnk at google.com> wrote:
>>
>>> For TLS, that seems fine to commit after the LLVM TLS changes land.  Or
>>> is LLVM already ready?
>>>
>>>  Ditto for SEH.  It should land when LLVM is ready.  Otherwise I don't
>>> think it's interesting.
>>>
>>> The toolchain one is interesting, and is closest to ready to land.  I
>>> don't think the _WIN32 ifdefs are correct.  On all platforms, we should use
>>> the sysroot if provided and search relative to the clang binary if not.
>>>
>>> I'm not sure if the Generic_GCC toolchain is helpful here, but there
>>> might be some gcc detection logic there that's worth reusing.
>>>
>>>
>>> On Tue, Apr 15, 2014 at 11:42 AM, Yaron Keren <yaron.keren at gmail.com>wrote:
>>>
>>>> Hi Martell,
>>>>
>>>> For SEH support, the patch by Kai Nacke (
>>>> http://wiki.dlang.org/Building_and_hacking_LDC_on_Windows_using_MSVC )
>>>> should also be required, not?
>>>>
>>>> Yaron
>>>>
>>>>
>>>>
>>>> 2014-04-15 21:32 GMT+03:00 Martell Malone <martellmalone at gmail.com>:
>>>>
>>>>> Hi
>>>>>
>>>>> This is my first time using a mailing list so bare with me please.
>>>>>
>>>>> Based on work done by ruben vb I have created a patch for clang that
>>>>> will find the include directories for the new windows-gnu triplet.
>>>>> I have also enabled TLS which mingw64 supports and the introduction of
>>>>> SEH exceptions on the front end.
>>>>>
>>>>> A patch will also be sent to the llvm mailing list for the backend.
>>>>>
>>>>> Many Thanks
>>>>> Martell
>>>>>
>>>>> _______________________________________________
>>>>> cfe-commits mailing list
>>>>> cfe-commits at cs.uiuc.edu
>>>>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>>>>
>>>>>
>>>>
>>>> _______________________________________________
>>>> cfe-commits mailing list
>>>> cfe-commits at cs.uiuc.edu
>>>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>>>
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140418/ba7ec792/attachment.html>


More information about the cfe-commits mailing list