[Bug 46254] New: EnumResourceTypesW should return True if there are no resources to enumerate

https://bugs.winehq.org/show_bug.cgi?id=46254

Bug ID: 46254
           Summary: EnumResourceTypesW should return True if there are no
                    resources to enumerate
           Product: Wine
           Version: 3.17
          Hardware: x86
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: kernel32
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

EnumResourceTypesW -- When there are no resource types to enumerate the
function instead of True will return False (see
https://github.com/wine-mirror/wine/blob/master/dlls/kernel32/resource.c#L25...)

This bug was originally reported in pywin32-ctypes
https://github.com/enthought/pywin32-ctypes/issues/59

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=46254

Nikolay Sivov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEEDINFO
     Ever confirmed|0                           |1

--- Comment #1 from Nikolay Sivov  ---
How did you test this? I tried this:

ret = EnumResourceTypesW(NULL, enum_res_proc, 0);
ok(ret, "Unexpected return value %d, error %d.\n", ret, GetLastError());

in some module without resources, and I get 0 return value, and error 1812,
same on Windows and Wine.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=46254

Fabian Maurer  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #2 from Fabian Maurer  ---
Any news?

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

Janne  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #3 from Janne  ---
Created attachment 79049
  --> http://bugs.winehq.org/attachment.cgi?id=79049
EnumResourceTypesA test code

I made simpple application which calls EnumResourceTypesA.

Application was tested on Wine 10.9 and on Windows 11.

In case callback function returns TRUE. 
EnumResourceTypesA returns TRUE 
In Windows11: GetLastError returns 126.
In Wine: GetLastError returns 126.

If callback function returns FALSE. 
EnumResourceTypesA returns FALSE.
In Windows11: GetLastError returns 15106.
In Wine: GetLastError returns 126.

If callback function pointer is NULL.

In Windows11: 
EnumResourceTypesA returns FALSE. 
GetLastError returns: 998.

Wine Crashes.

Source of my test application attached.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

Maotong Zhang  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #4 from Maotong Zhang  ---
(In reply to Janne from comment #3)
> Created attachment 79049 [details]
> EnumResourceTypesA test code
>
> I made simpple application which calls EnumResourceTypesA.
>
> Application was tested on Wine 10.9 and on Windows 11.
>
> In case callback function returns TRUE. 
> EnumResourceTypesA returns TRUE 
> In Windows11: GetLastError returns 126.
> In Wine: GetLastError returns 126.
>
> If callback function returns FALSE. 
> EnumResourceTypesA returns FALSE.
> In Windows11: GetLastError returns 15106.
> In Wine: GetLastError returns 126.
>
> If callback function pointer is NULL.
>
> In Windows11: 
> EnumResourceTypesA returns FALSE. 
> GetLastError returns: 998.
>
> Wine Crashes.
>
> Source of my test application attached.

In what situations is it necessary to pass NULL to EnumResourceTypes?

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #5 from Janne  ---
(In reply to Maotong Zhang from comment #4)
> (In reply to Janne from comment #3)
> > Created attachment 79049 [details]
> > EnumResourceTypesA test code
> >
> > I made simpple application which calls EnumResourceTypesA.
> >
> > Application was tested on Wine 10.9 and on Windows 11.
> >
> > In case callback function returns TRUE. 
> > EnumResourceTypesA returns TRUE 
> > In Windows11: GetLastError returns 126.
> > In Wine: GetLastError returns 126.
> >
> > If callback function returns FALSE. 
> > EnumResourceTypesA returns FALSE.
> > In Windows11: GetLastError returns 15106.
> > In Wine: GetLastError returns 126.
> >
> > If callback function pointer is NULL.
> >
> > In Windows11: 
> > EnumResourceTypesA returns FALSE. 
> > GetLastError returns: 998.
> >
> > Wine Crashes.
> >
> > Source of my test application attached.
>
> In what situations is it necessary to pass NULL to EnumResourceTypes?

Most likely thee are not situations where NULL is passed to EnumResourceTypes.
But there seems to be a difference in behavior between Wine & Windows.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #6 from Maotong Zhang  ---
I tested your program and it didn’t crash. Try applying the patch from Bug
46255 and test again to see if it still crashes.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #7 from Janne  ---
Created attachment 79055
  --> http://bugs.winehq.org/attachment.cgi?id=79055
console logs of tests

Applied both patches from bug-46255.
Build 32 and 64bit versions of Wine.
Build 32 and 64 bit version of the application.

64bit version of the application seems to behave same way on Wine & Windows.
32bit version has differeneses on output. Also application crashes, but that
might me my build of the Wine causing this.

I attached terminal/console logs.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #8 from Stuart Axon  ---
Hi all,
  It's great to see some progress on this.  There's a bit more context on the
pywin32-ctypes bug

https://github.com/enthought/pywin32-ctypes/issues/59

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #9 from Maotong Zhang  ---
(In reply to Janne from comment #7)
> Created attachment 79055 [details]
> console logs of tests
>
> Applied both patches from bug-46255.
> Build 32 and 64bit versions of Wine.
> Build 32 and 64 bit version of the application.
>
> 64bit version of the application seems to behave same way on Wine & Windows.
> 32bit version has differeneses on output. Also application crashes, but that
> might me my build of the Wine causing this.
>
> I attached terminal/console logs.

I compiled it this way into test program.
32bit: i686-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
enumresource_32.exe

64bit: x86_64-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
enumresource.exe

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #10 from Janne  ---
(In reply to Maotong Zhang from comment #9)
> (In reply to Janne from comment #7)
> > Created attachment 79055 [details]
> > console logs of tests
> >
> > Applied both patches from bug-46255.
> > Build 32 and 64bit versions of Wine.
> > Build 32 and 64 bit version of the application.
> >
> > 64bit version of the application seems to behave same way on Wine & Windows.
> > 32bit version has differeneses on output. Also application crashes, but that
> > might me my build of the Wine causing this.
> >
> > I attached terminal/console logs.
>
> I compiled it this way into test program.
> 32bit: i686-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> enumresource_32.exe
>
> 64bit: x86_64-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> enumresource.exe

I recompiled test application with same parameters.

Now both 32&64bit versions crash on NULL pointer situation.

If application is compiled with clang crash do not happen.
With clang i used these parameters to compile:
clang -m64 enumresource.c -o enumresource_64_clang.exe

I could not spot any changes to function EnumResourceTypesW in patches at
bug-46255.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #11 from Maotong Zhang  ---
(In reply to Janne from comment #10)
> (In reply to Maotong Zhang from comment #9)
> > (In reply to Janne from comment #7)
> > > Created attachment 79055 [details]
> > > console logs of tests
> > >
> > > Applied both patches from bug-46255.
> > > Build 32 and 64bit versions of Wine.
> > > Build 32 and 64 bit version of the application.
> > >
> > > 64bit version of the application seems to behave same way on Wine & Windows.
> > > 32bit version has differeneses on output. Also application crashes, but that
> > > might me my build of the Wine causing this.
> > >
> > > I attached terminal/console logs.
> >
> > I compiled it this way into test program.
> > 32bit: i686-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> > enumresource_32.exe
> >
> > 64bit: x86_64-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> > enumresource.exe
>
> I recompiled test application with same parameters.
>
> Now both 32&64bit versions crash on NULL pointer situation.
>
> If application is compiled with clang crash do not happen.
> With clang i used these parameters to compile:
> clang -m64 enumresource.c -o enumresource_64_clang.exe
>
> I could not spot any changes to function EnumResourceTypesW in patches at
> bug-46255.

I'm a bit confused — the test program I compiled runs under Wine without
crashing, and the results are the same as on Windows 11.
What could be different between our setups?

Could you upload the test program you compiled? I'd like to test it on my side
and see.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #12 from Janne  ---
Created attachment 79058
  --> http://bugs.winehq.org/attachment.cgi?id=79058
64bit test application

Test application 64bit version.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #13 from Janne  ---
I changed my test application slightly.
With this modified application i can define the address of callback function.

Application was compiled with these parameters:
gcc -m32 enumresource_2.c -o test.exe

On Wine result looks like this:

C:\temp>test 3405695742
argv[1]=3405695742 (cafecafe)
wine: Unhandled page fault on execute access to CAFECAFE at address CAFECAFE
(thread 012c), starting debugger.

Code:
tmp=atoi(argv[1]);
printf("argv[1]=%s (%x)\r\n",argv[1],tmp);
ret = EnumResourceTypesW(NULL, (HANDLE) tmp, lval);
printf("lParam:%d Ret val:%d LastError:%d\r\n",lval,ret,GetLastError());

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #14 from Maotong Zhang  ---
(In reply to Janne from comment #13)
> I changed my test application slightly.
> With this modified application i can define the address of callback function.
>
> Application was compiled with these parameters:
> gcc -m32 enumresource_2.c -o test.exe
>
> On Wine result looks like this:
>
> C:\temp>test 3405695742
> argv[1]=3405695742 (cafecafe)
> wine: Unhandled page fault on execute access to CAFECAFE at address CAFECAFE
> (thread 012c), starting debugger.
>
> Code:
> tmp=atoi(argv[1]);
> printf("argv[1]=%s (%x)\r\n",argv[1],tmp);
> ret = EnumResourceTypesW(NULL, (HANDLE) tmp, lval);
> printf("lParam:%d Ret val:%d LastError:%d\r\n",lval,ret,GetLastError());

The test program you compiled does crash on my side.
I'll continue analyzing the cause of the crash.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #15 from Maotong Zhang  ---
(In reply to Maotong Zhang from comment #9)
> (In reply to Janne from comment #7)
> > Created attachment 79055 [details]
> > console logs of tests
> >
> > Applied both patches from bug-46255.
> > Build 32 and 64bit versions of Wine.
> > Build 32 and 64 bit version of the application.
> >
> > 64bit version of the application seems to behave same way on Wine & Windows.
> > 32bit version has differeneses on output. Also application crashes, but that
> > might me my build of the Wine causing this.
> >
> > I attached terminal/console logs.
>
> I compiled it this way into test program.
> 32bit: i686-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> enumresource_32.exe
>
> 64bit: x86_64-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> enumresource.exe

I compiled the test program using MinGW under Linux.
You can give it a try.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #16 from Janne  ---
Created attachment 79059
  --> http://bugs.winehq.org/attachment.cgi?id=79059
patch

I tested to patch EnumResourceTypesW.
After patch test application do not crash anymore with NULL pointer.
Also gives same error 998 as Windows.
patch attached.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #17 from Maotong Zhang  ---
(In reply to Janne from comment #16)
> Created attachment 79059 [details]
> patch
>
> I tested to patch EnumResourceTypesW.
> After patch test application do not crash anymore with NULL pointer.
> Also gives same error 998 as Windows.
> patch attached.

Issue resolved.
I think unsigned int should be changed to ULONG_PTR.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #18 from Janne  ---
(In reply to Maotong Zhang from comment #15)
> (In reply to Maotong Zhang from comment #9)
> > (In reply to Janne from comment #7)
> > > Created attachment 79055 [details]
> > > console logs of tests
> > >
> > > Applied both patches from bug-46255.
> > > Build 32 and 64bit versions of Wine.
> > > Build 32 and 64 bit version of the application.
> > >
> > > 64bit version of the application seems to behave same way on Wine & Windows.
> > > 32bit version has differeneses on output. Also application crashes, but that
> > > might me my build of the Wine causing this.
> > >
> > > I attached terminal/console logs.
> >
> > I compiled it this way into test program.
> > 32bit: i686-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> > enumresource_32.exe
> >
> > 64bit: x86_64-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> > enumresource.exe
>
> I compiled the test program using MinGW under Linux.
> You can give it a try.

I tested compiling test application in Wine.

Command used:
x86_64-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
enumresource_wine.exe

This application do not crash.

So for some reason the application that is compiled on Windows causes the
crash.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #19 from Maotong Zhang  ---
(In reply to Janne from comment #18)
> (In reply to Maotong Zhang from comment #15)
> > (In reply to Maotong Zhang from comment #9)
> > > (In reply to Janne from comment #7)
> > > > Created attachment 79055 [details]
> > > > console logs of tests
> > > >
> > > > Applied both patches from bug-46255.
> > > > Build 32 and 64bit versions of Wine.
> > > > Build 32 and 64 bit version of the application.
> > > >
> > > > 64bit version of the application seems to behave same way on Wine & Windows.
> > > > 32bit version has differeneses on output. Also application crashes, but that
> > > > might me my build of the Wine causing this.
> > > >
> > > > I attached terminal/console logs.
> > >
> > > I compiled it this way into test program.
> > > 32bit: i686-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> > > enumresource_32.exe
> > >
> > > 64bit: x86_64-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> > > enumresource.exe
> >
> > I compiled the test program using MinGW under Linux.
> > You can give it a try.
>
> I tested compiling test application in Wine.
>
> Command used:
> x86_64-w64-mingw32-gcc -Wall -g3 enumresource.c -static -lkernel32 -o
> enumresource_wine.exe
>
> This application do not crash.
>
> So for some reason the application that is compiled on Windows causes the
> crash.

Thanks, I will take another look to see what's different.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #20 from Janne  ---
Hello,
I opened test application built in Wine and in Windows with Ghidra.

When comparing main function generated by Ghidra there was no difference
between the two.

There was difference in mainCRTStartup function between Wine & Windows built
application.

[Bug 46254] EnumResourceTypesW should return True if there are no resources to enumerate

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=46254

--- Comment #21 from Maotong Zhang  ---
Yes, you are correct.
When compiled with Linux + MinGW, the initialization logic of mainCRTStartup is
relatively simple and does not perform checks.
However, the mainCRTStartup compiled on Windows performs function pointer
checks, and it is precisely these checks that cause the crash.