Boost C++ Libraries: Ticket #10685: mfc_thread_init.hpp does not compile

attachment set

fab@… — Wed, 22 Oct 2014 21:03:32 GMT

attachment → mfc_thread_init.patch

Patch to address issue

description changed

viboes — Sat, 25 Oct 2014 15:27:01 GMT

description modified (diff)

viboes — Sat, 25 Oct 2014 15:38:52 GMT

Could you show the whole compile error?

viboes — Sat, 25 Oct 2014 15:42:09 GMT

Every reference to ExtRawDllMain (https://social.msdn.microsoft.com/Search/en-US?query=ExtRawDllMain&emptyWatermark=true&ac=4) make use of HINSTANCE.

Where do you have find a declaration using HANDLE?

owner changed

viboes — Sat, 25 Oct 2014 15:42:30 GMT

owner changed from Anthony Williams to Niall Douglas

fab@… — Tue, 28 Oct 2014 00:46:29 GMT

Replying to viboes:

Could you show the whole compile error?

That is the whole compile error on the line that I am changing in the patch.

fab@… — Tue, 28 Oct 2014 00:50:36 GMT

Replying to viboes:

Every reference to ExtRawDllMain make use of HINSTANCE.

Where do you have find a declaration using HANDLE?

I don't know why the original author of this file used HANDLE -- you are correct in that it should have used HINSTANCE instead just about everywhere. If the "reverse" change is applied, I guess the change could break some folks who started using the HANDLE version which is currently exposed.

viboes — Tue, 28 Oct 2014 06:47:00 GMT

Replying to fab@…:

Replying to viboes:

Every reference to ExtRawDllMain make use of HINSTANCE.

Where do you have find a declaration using HANDLE?

I don't know why the original author of this file used HANDLE -- you are correct in that it should have used HINSTANCE instead just about everywhere. If the "reverse" change is applied, I guess the change could break some folks who started using the HANDLE version which is currently exposed.

By the patch is doing the opposite. is changing HINSTANCE by HANDLE.

Niall Douglas — Sat, 01 Nov 2014 16:10:40 GMT

DllMain, or RawDllMain, takes a HINSTANCE not a HANDLE. Whatever isn't using HINSTANCE needs to do so.

We actually have a fairly big problem here, this particular file doesn't get compiled unless the project including the header is using MFC which is surely a minority of users. I'll look into writing a new test case which switches on MFC, that should include this file into being tested from now on. I'll also fix everything up to use HINSTANCE as it should have been from the beginning.

Niall Douglas — Sat, 01 Nov 2014 17:00:18 GMT

There is a problem.

The following MFC Extension DLL ought to be close to compiling:

// Purely to test if Boost.Thread compiles with a MFC extension DLL
#ifdef _MSC_VER
#include <windows.h>
#include <stdafx.h>
#include <afxdllx.h>
#include <boost/thread.hpp>
static AFX_EXTENSION_MODULE extensionDLL;
extern "C" int APIENTRY
DllMain(HINSTANCE hInstance, DWORD dwReason, LPVOID)
{
   if (dwReason == DLL_PROCESS_ATTACH)
   {
      // Extension DLL one-time initialization
      if (!AfxInitExtensionModule(
             extensionDLL, hInstance))
         return 0;
   }
   else if (dwReason == DLL_PROCESS_DETACH)
   {
      // Extension DLL per-process termination
      AfxTermExtensionModule(extensionDLL);
   }
   return 1;   // ok
}
#endif

And that would test if Boost.Thread compiles with MFC or not.

My problem is I have the Visual Studio Express free edition. I don't have the commercial one costing thousands of euro, so I don't have MFC.

Can someone else test this for me please?

Thanks, Niall

fab@… — Mon, 03 Nov 2014 16:59:34 GMT

Guys,

Glad this is converging.

I have been under a number of deadlines, and I hope to be able to verify the "reverse" patch this week.

I will retract my patch in the meantime.

viboes — Sat, 06 Dec 2014 09:56:29 GMT

Any news on this issue?

type changed

viboes — Sat, 13 Dec 2014 16:07:57 GMT

type Bugs → Support Requests

Moved to support, until there is more feedback.

status changed; resolution set

viboes — Sat, 04 Apr 2015 21:43:23 GMT

status new → closed
resolution → wontfix

We need tome tester help in order to be able to fix it. Closed until we get it.

Fri, 22 Apr 2016 15:44:25 GMT

mfc_thread_init.hpp does not compile. This still does not compile when used with MFC. The conflict is pretty obvious here as the function declaration and forward declaration/ assignment do not agree. if line 30 is changed from extern "C" declspec(selectany) BOOL (WINAPI * const _pRawDllMainOrig)(HANDLE, DWORD, LPVOID) = &ExtRawDllMain;

to extern "C" declspec(selectany) BOOL (WINAPI * const _pRawDllMainOrig)(HINSTANCE, DWORD, LPVOID) = &ExtRawDllMain;

everything compiles and runs as expected.