[lldb-dev] LLDB /w Windows and MinGW64
Zachary Turner via lldb-dev
lldb-dev at lists.llvm.org
Mon Mar 28 11:28:55 PDT 2016
Almost, there's one more step. Click Create a New Revision on that screen,
then give it a title and a description. For reviewers put zturner, and for
subscribers put lldb-commits
On Mon, Mar 28, 2016 at 11:20 AM Eran Ifrah <eran.ifrah at gmail.com> wrote:
> Is this what you meant:
> http://reviews.llvm.org/differential/diff/51809/
>
> Thanks
>
> On Mon, Mar 28, 2016 at 8:58 PM, Zachary Turner <zturner at google.com>
> wrote:
>
>> For the patch, can you create an account on reviews.llvm.org, and upload
>> your patch there? This makes interactive reviewing / commenting much
>> easier. Let me know if you need help getting that set up.
>>
>> On Mon, Mar 28, 2016 at 10:58 AM Zachary Turner <zturner at google.com>
>> wrote:
>>
>>> If you compile with MSVC or Clang-cl it wouldn't ask for lldb-server.
>>> So most likely there is some code that is using #if defined(_MSC_VER) when
>>> it should be using #if defined(LLVM_ON_WINDOWS).
>>>
>>> You'll have to hunt that down, but a good starting point might be to put
>>> a breakpoint in ProcessWindowsLive::CreateInstance and then work backwards
>>> to see why that isn't getting called (assuming it's not).
>>>
>>> On Mon, Mar 28, 2016 at 9:05 AM Eran Ifrah <eran.ifrah at gmail.com> wrote:
>>>
>>>> On Mon, Mar 28, 2016 at 6:56 PM, Zachary Turner <zturner at google.com>
>>>> wrote:
>>>>
>>>>> Patches welcome. If you can split it into independent pieces that
>>>>> would be helpful, but it's not always possible.
>>>>>
>>>>> Patch is attached, I think you will find it quite straight forward -
>>>> feel free to comment and send it back for revise
>>>>
>>>>
>>>>
>>>>> The NativeProcessProtocol error, that's the interface that converts
>>>>> debugging events that occur on the inferior into packets that can be sent
>>>>> to the server, and vice versa. Since Windows doesn't currently use lldb
>>>>> server, this piece has never been written for Windows
>>>>
>>>> So this raises the question: how come lldb asks for it? (see my first
>>>> emai inl this conversation)
>>>> I would have build LLDB in debug mode, but it seems that MinGW as.exe
>>>> fails to write some of the files "File too big"
>>>>
>>>>
>>>>
>>>>>
>>>>> On Mon, Mar 28, 2016 at 8:41 AM Eran Ifrah <eran.ifrah at gmail.com>
>>>>> wrote:
>>>>>
>>>>>> On Mon, Mar 28, 2016 at 6:10 PM, Zachary Turner <zturner at google.com>
>>>>>> wrote:
>>>>>>
>>>>>>> I'm the main Windows maintainer,
>>>>>>
>>>>>> Hi
>>>>>>
>>>>>>
>>>>>>
>>>>>>> and while We've gotten things working pretty well on Windows, our
>>>>>>> effort has been 100% on building with msvc and/or clang-cl. Building with
>>>>>>> mingw has a different set of pre processor defines and some other subtle
>>>>>>> differences, so it doesn't surprise me that things don't work quite right.
>>>>>>>
>>>>>>> I got it to compile (I have a big patch that I can send you if you
>>>>>> are interested)
>>>>>> mainly involves blocking code under __MINGW32__ and some updates to
>>>>>> the various CMakeLists.txt and AddLLDB.cmake module files
>>>>>> Some functions are missing in MinGW implementations (like gets_s and
>>>>>> others :/)
>>>>>>
>>>>>>
>>>>>>
>>>>>>> You can try getting lldb-server to build and run under Windows, or
>>>>>>> you can try to get it to use the non lldb server codepath on MinGW, but you
>>>>>>> may still run into some
>>>>>>
>>>>>> I got lldb-server to compile and run on Windows, however, it crashes
>>>>>> immediately and the backtrace shows this:
>>>>>>
>>>>>> 0 0x0000000000724615
>>>>>> lldb_private::NativeProcessProtocol::Launch(lldb_private::ProcessLaunchInfo&,
>>>>>> lldb_private::NativeProcessProtocol::NativeDelegate&,
>>>>>> lldb_private::MainLoopBase&,
>>>>>> std::shared_ptr<lldb_private::NativeProcessProtocol>&)
>>>>>> 1 0x00000000005f815d
>>>>>> lldb_private::process_gdb_remote::GDBRemoteCommunicationServerLLGS::LaunchProcess()
>>>>>>
>>>>>> 2 0x00000000004020e5
>>>>>> handle_launch(lldb_private::process_gdb_remote::GDBRemoteCommunicationServerLLGS&,
>>>>>> int, char const* const*)
>>>>>> 3 0x000000000040335e main_gdbserver(int, char**)
>>>>>> 4 0x000000000188329c main
>>>>>>
>>>>>>
>>>>>> I had to #ifndef __MINGW32__ around lldb-platform.cpp as it contains
>>>>>> too much Linux code that can not be compiled under Windows (fork, exec*)
>>>>>> and basically lldb-server calls the main_gdbserver instead of
>>>>>> main_platform function
>>>>>>
>>>>>> Looking at the function that crashes, I see this:
>>>>>>
>>>>>> llvm_unreachable("Platform has no NativeProcessProtocol support");
>>>>>>
>>>>>> Any ideas?
>>>>>> Thanks!
>>>>>>
>>>>>> issues after that as well, since you're the first person afaik to try
>>>>>>> building with MinGW
>>>>>>>
>>>>>>
>>>>>>
>>>>>>> On Sun, Mar 27, 2016 at 10:29 PM Eran Ifrah via lldb-dev <
>>>>>>> lldb-dev at lists.llvm.org> wrote:
>>>>>>>
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> I have built LLDB on Windows 7 using MinGW64/4.9.2 (took some
>>>>>>>> effort to get the job done...)
>>>>>>>> When I tried to debug a simple hello world executable, I get this
>>>>>>>> output:
>>>>>>>>
>>>>>>>> D:\software\msys-for-clang\1.0\home\PC\build-release-64-lldb\bin>lldb.exe
>>>>>>>> HelloWorld.exe
>>>>>>>> (lldb) target create "HelloWorld.exe"
>>>>>>>> Current executable set to 'HelloWorld.exe' (x86_64).
>>>>>>>> (lldb) b main.cpp:7
>>>>>>>> Breakpoint 1: where = HelloWorld.exe`main + 26 at main.cpp:7,
>>>>>>>> address = 0x000000000040154a
>>>>>>>> (lldb) r
>>>>>>>> error: process launch failed: unable to locate lldb-server
>>>>>>>> (lldb)
>>>>>>>>
>>>>>>>> I can't seem to locate lldb-server anywhere, and according
>>>>>>>> to LLDBConfig.cmake, this target should not get built on Windows:
>>>>>>>>
>>>>>>>> # Figure out if lldb could use lldb-server. If so, then we'll
>>>>>>>> # ensure we build lldb-server when an lldb target is being built.
>>>>>>>> if ((CMAKE_SYSTEM_NAME MATCHES "Darwin") OR
>>>>>>>> (CMAKE_SYSTEM_NAME MATCHES "FreeBSD") OR
>>>>>>>> (CMAKE_SYSTEM_NAME MATCHES "Linux") OR
>>>>>>>> (CMAKE_SYSTEM_NAME MATCHES "NetBSD"))
>>>>>>>> set(LLDB_CAN_USE_LLDB_SERVER 1)
>>>>>>>> else()
>>>>>>>> set(LLDB_CAN_USE_LLDB_SERVER 0)
>>>>>>>> endif()
>>>>>>>>
>>>>>>>> and in the tools/CMakeLists.txt file we have this:
>>>>>>>>
>>>>>>>> if (LLDB_CAN_USE_LLDB_SERVER)
>>>>>>>> add_subdirectory(lldb-server)
>>>>>>>> endif()
>>>>>>>>
>>>>>>>>
>>>>>>>> Any ideas?
>>>>>>>>
>>>>>>>> Thanks,
>>>>>>>>
>>>>>>>> --
>>>>>>>> Eran Ifrah,
>>>>>>>> Author of
>>>>>>>> CodeLite
>>>>>>>> , a cross platform open source C/C++ IDE: http://www.codelite.org
>>>>>>>> CodeLite IDE Blog: http://codeliteide.blogspot.com/
>>>>>>>> _______________________________________________
>>>>>>>> lldb-dev mailing list
>>>>>>>> lldb-dev at lists.llvm.org
>>>>>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Eran Ifrah,
>>>>>> Author of codelite, a cross platform open source C/C++ IDE:
>>>>>> http://www.codelite.org
>>>>>> CodeLite IDE Blog: http://codeliteide.blogspot.com/
>>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Eran Ifrah,
>>>> Author of codelite, a cross platform open source C/C++ IDE:
>>>> http://www.codelite.org
>>>> CodeLite IDE Blog: http://codeliteide.blogspot.com/
>>>>
>>>
>
>
> --
> Eran Ifrah,
> Author of codelite, a cross platform open source C/C++ IDE:
> http://www.codelite.org
> CodeLite IDE Blog: http://codeliteide.blogspot.com/
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/lldb-dev/attachments/20160328/4d85b55c/attachment.html>
More information about the lldb-dev
mailing list