MinGW - Minimalist GNU for Windows

Have you tried looking at the location of the error? It might be that an
extra #define was added to w32api that conflicts with the Tcl source.

Luke

----- Original Message ----- 
From: "Keiichi Takahashi" <bitwalk@...>
To: <mingw-users@...>
Sent: Saturday, September 13, 2003 10:14 AM
Subject: [Mingw-users] [Q] w32api 2.3 vs 2.4


> Hi,
>
> After updating w32api from 2.3 to 2.4 on my system, I tested compiling
> with Tcl 8.4.4 sources, but compiling had been stopped by errors shown
> below.
>
> Then I brought it back to 2.3 and confirmed that the source can be
> certainly compiled without fatal error with w32api-2.3,
>
> My question is what is major difference between w32api 2.3 and 2.4, I
> still need to check compiling with other sources though...
>
> Any help will be appreciated. Thanx.
>
> Keiichi
>
>
> gcc -c -O -Wall -Wconversion  -I"./../generic" -I"." -DHAVE_NO_SEH=1
> -DEXCEPTION_DISPOSITION=int -DHAVE_NO_LPFN_DECLS=1
> -DHAVE_WINNT_IGNORE_VOID=1 -DHAVE_ALLOCA_GCC_INLINE=1
> -DHAVE_CAST_TO_UNION=1 -DTCL_THREADS=1 -DUSE_THREAD_ALLOC=1
> -DBUILD_tcl "./../generic/tclCmdAH.c" -o tclCmdAH.o
> ../generic/tclCmdAH.c: In function `Tcl_FileObjCmd':
> ../generic/tclCmdAH.c:806: error: parse error before numeric constant
> ../generic/tclCmdAH.c:902: error: duplicate case value
> ../generic/tclCmdAH.c:879: error: previously used here
> ../generic/tclCmdAH.c:1376: warning: case value `0' not in enumerated
> type `options'
> ../generic/tclCmdAH.c:1376: warning: case value `1' not in enumerated
> type `options'
> ../generic/tclCmdAH.c:1376: warning: case value `2' not in enumerated
> type `options'
>   :
> (snip)
>   :
> ../generic/tclCmdAH.c:1376: warning: case value `31' not in enumerated
> type `options'
> make: *** [tclCmdAH.o] Error 1
> gcc -c -O -Wall -Wconversion  -I"./../generic" -I"." -DHAVE_NO_SEH=1
> -DEXCEPTION_DISPOSITION=int -DHAVE_NO_LPFN_DECLS=1
> -DHAVE_WINNT_IGNORE_VOID=1 -DHAVE_ALLOCA_GCC_INLINE=1
> -DHAVE_CAST_TO_UNION=1 -DTCL_THREADS=1 -DUSE_THREAD_ALLOC=1
> -DBUILD_tcl "./../generic/tclCmdAH.c" -o tclCmdAH.o
> ../generic/tclCmdAH.c: In function `Tcl_FileObjCmd':
> ../generic/tclCmdAH.c:806: error: parse error before numeric constant
> ../generic/tclCmdAH.c:902: error: duplicate case value
> ../generic/tclCmdAH.c:879: error: previously used here
> ../generic/tclCmdAH.c:1376: warning: case value `0' not in enumerated
> type `options'
> ../generic/tclCmdAH.c:1376: warning: case value `1' not in enumerated
> type `options'
> ../generic/tclCmdAH.c:1376: warning: case value `2' not in enumerated
> type `options'
>   :
> (snip)
>   :
> ../generic/tclCmdAH.c:1376: warning: case value `31' not in enumerated
> type `options'
> make: *** [tclCmdAH.o] Error 1
>
> bitwalk@... ~/tcl8.4.4/win

Hello Luke,

> Have you tried looking at the location of the error?

I'm sorry, not yet so detail.


> It might be that an extra #define was added to w32api
> that conflicts with the Tcl source.

Yes, it could be, as error messages say.

I will look into detail later. Now I have to use w32api-2.3 in order to
compile other sources steadily.

If this error can be observed only when compiling Tcl sources, that's
fine. We can modify Tcl sources to meet new w32api, but it does not make
sense to me, since the Microsoft's compiler can compile them as well as
the MinGW gcc with older w32api on Windows.

Keiichi

-- 
Keiichi Takahashi, bitWalk Co.,Ltd.
mailto:bitwalk@...
http://b2c-1.rocketbeach.com/~bitwalk/

Hi,

I had some problems with my code too: they were due to some changes in =
the
organization of the definitions in the include files. I don't know if =
some
files that used to be included automatically are not included anymore, =
or if
simply some definitions were moved in new/other files.=20

Anyway, it was fixed by just adding a few #include's, which were not =
needed
with 2.3...

Regards,

--=20
Fran=E7ois Ferrand

> -----Original Message-----
> From: mingw-users-admin@...
> [mailto:mingw-users-admin@...] On Behalf Of=20
> Keiichi Takahashi
> Sent: Saturday, September 13, 2003 9:40
> To: mingw-users@...
> Subject: Re: [Mingw-users] [Q] w32api 2.3 vs 2.4
>=20
>=20
> Hello Luke,
>=20
> > Have you tried looking at the location of the error?
>=20
> I'm sorry, not yet so detail.
>=20
>=20
> > It might be that an extra #define was added to w32api
> > that conflicts with the Tcl source.
>=20
> Yes, it could be, as error messages say.
>=20
> I will look into detail later. Now I have to use w32api-2.3=20
> in order to compile other sources steadily.
>=20
> If this error can be observed only when compiling Tcl=20
> sources, that's fine. We can modify Tcl sources to meet new=20
> w32api, but it does not make sense to me, since the=20
> Microsoft's compiler can compile them as well as the MinGW=20
> gcc with older w32api on Windows.
>=20
> Keiichi
>=20
> --=20
> Keiichi Takahashi, bitWalk Co.,Ltd.=20
> mailto:bitwalk@... http://b2c-1.rocketbeach.com/> =
~bitwalk/
>=20
>=20
>=20
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf=20
> _______________________________________________
> MinGW-users mailing list
> MinGW-users@...
>=20
> You may change your MinGW Account Options or unsubscribe at:=20
https://lists.sourceforge.net/lists/listinfo/mingw-users

Hi Keiichi,

Keiichi Takahashi <bitwalk@...> writes:
> ../generic/tclCmdAH.c: In function `Tcl_FileObjCmd':
> ../generic/tclCmdAH.c:806: error: parse error before numeric constant

That's in the declaration of an enum.  I guess one of the enumeration
constants has the same names as new define in one of the winapi
headers (namespace pollution).  I suggest you have a look at the
preprocesser output, see which of the enum values has the problem and
add an #undef in the Tcl code to avoid the Windows definition.

benny

Hi Benjamin,

> That's in the declaration of an enum.  I guess one of the enumeration
> constants has the same names as new define in one of the winapi
> headers (namespace pollution).  I suggest you have a look at the
> preprocesser output, see which of the enum values has the problem and
> add an #undef in the Tcl code to avoid the Windows definition.

Thank you for your helpful comments. I will look into them and try to 
put #undef(s) in the Tcl codes.

Regards,
Keiichi


-- 
Keiichi Takahashi, bitWalk Co.,Ltd.
mailto:bitwalk@...
http://b2c-1.rocketbeach.com/~bitwalk/

Hi,

>> That's in the declaration of an enum.  I guess one of the enumeration
>> constants has the same names as new define in one of the winapi
>> headers (namespace pollution).  I suggest you have a look at the

I've found the reason why compiling Tcl is failed.

winnt.h (w32api-2.4)
#define FILE_EXISTS 0x00000004

This would confilct with the FILE_EXISTS in enum declaration in 
generic/tclCmdAH.c of Tcl8.4.4 sources.

The FILE_EXISTS can not be found in winnt.h of w32api-2.3, but there is 
in w32api-2.4.

At this moment, the following macro is added just before defining enum 
variable in generic/tclCmdAH.c and the rest of compiling is successfully 
completed.

#ifdef __MINGW32__
#undef FILE_EXISTS
#endif

Thanks a lot.

Regards,
Keiichi

-- 
Keiichi Takahashi, bitWalk Co.,Ltd.
mailto:bitwalk@...
http://b2c-1.rocketbeach.com/~bitwalk/

Hi Keiichi,

Keiichi Takahashi <bitwalk@...> writes:
> winnt.h (w32api-2.4)
> #define FILE_EXISTS 0x00000004

> #ifdef __MINGW32__
> #undef FILE_EXISTS
> #endif

You don't need the #ifdef, and of course a comment mentioning the
actual reason (w32api/winnt.h) would be nice.

benny

Benjamin Riefenstahl wrote:
> Hi Keiichi,
> 
> Keiichi Takahashi <bitwalk@...> writes:
> 
>>winnt.h (w32api-2.4)
>>#define FILE_EXISTS 0x00000004
> 
> 
>>#ifdef __MINGW32__
>>#undef FILE_EXISTS
>>#endif
> 
> 
> You don't need the #ifdef, and of course a comment mentioning the
> actual reason (w32api/winnt.h) would be nice.
> 

A better solution is for tcl to prefix its own enumerations with tcl_ or 
something similar to avoid the namespace polution problem.  This is true 
for all packages.

Earnie.

Hi,

> A better solution is for tcl to prefix its own enumerations with tcl_ or 
> something similar to avoid the namespace polution problem.  This is true 
> for all packages.
> 
> Earnie.

This makes sense to me. I am wondering why it has not been discussed 
about same kind of problems so often before, because just I was not 
aware of?

Anyway, I will post this issue to tcl-core@... later. 
Thank you.

Keiichi
-- 
Keiichi Takahashi, bitwalk@...
bitWalk Co., Ltd.
http://b2c-1.rocketbeach.com/~bitwalk/