[GCC BUG] -I switch not functioning correctly.

[GCC BUG] -I switch not functioning correctly.

[Earnie Boyd]

This bug is present in both the Cygwin V1 CD and gcc-2.95.2.  I found this
trying out Mumits JNI examples.

Since I don't know if this is Cygwin specific or not then I haven't forwarded
this to the gcc-bugs list.

-- Output of make --

gcc  -c -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32 -g
 -Wall -O2 -v  -o invoke.o invoke.c
Reading specs from
/install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/spe
cs
gcc version 2.95.2 19991024 (release)
 /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/cpp.exe -lang-c -v 
 -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32 -iprefix /in
stall/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/ -D__GNUC__=2
-D__GNUC_MI
NOR__=95 -Di386 -D_WIN32 -DWINNT -D_X86_=1 -D__STDC__=1
-D__stdcall=__attribute_
_((__stdcall__)) -D__cdecl=__attribute__((__cdecl__))
-D__declspec(x)=__attribut
e__((x)) -D__i386__ -D_WIN32 -D__WINNT__ -D_X86_=1 -D__STDC__=1
-D__stdcall=__at
tribute__((__stdcall__)) -D__cdecl=__attribute__((__cdecl__))
-D__declspec(x)=__
attribute__((x)) -D__i386 -D__WINNT -Asystem(winnt) -Acpu(i386) -Amachine(i386)
-D__OPTIMIZE__ -g -Wall -remap -Acpu(i386) -Amachine(i386) -Di386 -D__i386
-D__i
386__ -Di586 -Dpentium -D__i586 -D__i586__ -D__pentium -D__pentium__
-D__CYGWIN3
2__ -D__CYGWIN__ invoke.c D:\root\Cygwin\/tmp/ccDlpXPm.i
GNU CPP version 2.95.2 19991024 (release) (80386, BSD syntax)
#include "..." search starts here:
#include <...> search starts here:
 .
 /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../include

/install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygw
in/include
 /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/include
 /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
 /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygwin/include
 /usr/include
End of search list.
The following default directories have been omitted from the search path:
 /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include/g++-3
End of omitted list.
invoke.c:1: jni.h: No such file or directory
make: *** [invoke.o] Error 1

-- End output from make --

=====
Earnie Boyd < mailto:earnie_boyd@yahoo.com >
Cygwin Newbies, please visit
< http://www.freeyellow.com/members5/gw32/index.html >
__________________________________________________
Do You Yahoo!?
Thousands of Stores.  Millions of Products.  All in one place.
Yahoo! Shopping: http://shopping.yahoo.com

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Mumit Khan]

On Mon, 6 Dec 1999, Earnie Boyd wrote:

> This bug is present in both the Cygwin V1 CD and gcc-2.95.2.  I found this
> trying out Mumits JNI examples.
> 
> Since I don't know if this is Cygwin specific or not then I haven't forwarded
> this to the gcc-bugs list.
> 
> -- Output of make --
> 
> gcc  -c -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32 -g
>  -Wall -O2 -v  -o invoke.o invoke.c

Ok, let's see what gcc sees. You're telling it to search directories
/install/jdk1.1.8/include and /install/jdk1.1.8/include/win32 for
jdk includes. So far so good.

> Reading specs from
> /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/spe
> cs
[ ... ]
> GNU CPP version 2.95.2 19991024 (release) (80386, BSD syntax)
> #include "..." search starts here:
> #include <...> search starts here:
>  .
>  /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> 
> /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygw
> in/include
>  /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/include
>  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
>  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygwin/include
>  /usr/include

Now note the problem here. There are no signs of either of the jdk
include directories!


> End of search list.
> The following default directories have been omitted from the search path:
>  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include/g++-3

Nor are they here, which would be the case normally if the paths were
inaccessible.

I really don't have a clue, sorry. Can you do a 
  
  $ cat /install/jdk1.1.8/include/<SOME_FILE_NAME>


Regards,
Mumit



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Mumit Khan]

On Mon, 6 Dec 1999, Earnie Boyd wrote:

> This bug is present in both the Cygwin V1 CD and gcc-2.95.2.  I found this
> trying out Mumits JNI examples.
> 
> Since I don't know if this is Cygwin specific or not then I haven't forwarded
> this to the gcc-bugs list.
> 
> -- Output of make --
> 
> gcc  -c -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32 -g
>  -Wall -O2 -v  -o invoke.o invoke.c

Ok, let's see what gcc sees. You're telling it to search directories
/install/jdk1.1.8/include and /install/jdk1.1.8/include/win32 for
jdk includes. So far so good.

> Reading specs from
> /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/spe
> cs
[ ... ]
> GNU CPP version 2.95.2 19991024 (release) (80386, BSD syntax)
> #include "..." search starts here:
> #include <...> search starts here:
>  .
>  /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> 
> /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygw
> in/include
>  /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/include
>  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
>  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygwin/include
>  /usr/include

Now note the problem here. There are no signs of either of the jdk
include directories!


> End of search list.
> The following default directories have been omitted from the search path:
>  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include/g++-3

Nor are they here, which would be the case normally if the paths were
inaccessible.

I really don't have a clue, sorry. Can you do a 
  
  $ cat /install/jdk1.1.8/include/<SOME_FILE_NAME>


Regards,
Mumit



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

[GCC BUG] -I switch not functioning correctly.

[Earnie Boyd]

This bug is present in both the Cygwin V1 CD and gcc-2.95.2.  I found this
trying out Mumits JNI examples.

Since I don't know if this is Cygwin specific or not then I haven't forwarded
this to the gcc-bugs list.

-- Output of make --

gcc  -c -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32 -g
 -Wall -O2 -v  -o invoke.o invoke.c
Reading specs from
/install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/spe
cs
gcc version 2.95.2 19991024 (release)
 /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/cpp.exe -lang-c -v 
 -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32 -iprefix /in
stall/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/ -D__GNUC__=2
-D__GNUC_MI
NOR__=95 -Di386 -D_WIN32 -DWINNT -D_X86_=1 -D__STDC__=1
-D__stdcall=__attribute_
_((__stdcall__)) -D__cdecl=__attribute__((__cdecl__))
-D__declspec(x)=__attribut
e__((x)) -D__i386__ -D_WIN32 -D__WINNT__ -D_X86_=1 -D__STDC__=1
-D__stdcall=__at
tribute__((__stdcall__)) -D__cdecl=__attribute__((__cdecl__))
-D__declspec(x)=__
attribute__((x)) -D__i386 -D__WINNT -Asystem(winnt) -Acpu(i386) -Amachine(i386)
-D__OPTIMIZE__ -g -Wall -remap -Acpu(i386) -Amachine(i386) -Di386 -D__i386
-D__i
386__ -Di586 -Dpentium -D__i586 -D__i586__ -D__pentium -D__pentium__
-D__CYGWIN3
2__ -D__CYGWIN__ invoke.c D:\root\Cygwin\/tmp/ccDlpXPm.i
GNU CPP version 2.95.2 19991024 (release) (80386, BSD syntax)
#include "..." search starts here:
#include <...> search starts here:
 .
 /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../include

/install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygw
in/include
 /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/include
 /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
 /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygwin/include
 /usr/include
End of search list.
The following default directories have been omitted from the search path:
 /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include/g++-3
End of omitted list.
invoke.c:1: jni.h: No such file or directory
make: *** [invoke.o] Error 1

-- End output from make --

=====
Earnie Boyd < mailto:earnie_boyd@yahoo.com >
Cygwin Newbies, please visit
< http://www.freeyellow.com/members5/gw32/index.html >
__________________________________________________
Do You Yahoo!?
Thousands of Stores.  Millions of Products.  All in one place.
Yahoo! Shopping: http://shopping.yahoo.com

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Earnie Boyd]

--- Mumit Khan <khan@NanoTech.Wisc.EDU> wrote:
> On Mon, 6 Dec 1999, Earnie Boyd wrote:
> 
> > This bug is present in both the Cygwin V1 CD and gcc-2.95.2.  I found this
> > trying out Mumits JNI examples.
> > 
> > Since I don't know if this is Cygwin specific or not then I haven't
> forwarded
> > this to the gcc-bugs list.
> > 
> > -- Output of make --
> > 
> > gcc  -c -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32
> -g
> >  -Wall -O2 -v  -o invoke.o invoke.c
> 
> Ok, let's see what gcc sees. You're telling it to search directories
> /install/jdk1.1.8/include and /install/jdk1.1.8/include/win32 for
> jdk includes. So far so good.
> 
> > Reading specs from
> > /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/spe
> > cs
> [ ... ]
> > GNU CPP version 2.95.2 19991024 (release) (80386, BSD syntax)
> > #include "..." search starts here:
> > #include <...> search starts here:
> >  .
> > 
> /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> > 
> >
>
/install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygw
> > in/include
> >  /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/include
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygwin/include
> >  /usr/include
> 
> Now note the problem here. There are no signs of either of the jdk
> include directories!
> 
> 
> > End of search list.
> > The following default directories have been omitted from the search path:
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include/g++-3
> 
> Nor are they here, which would be the case normally if the paths were
> inaccessible.
> 
> I really don't have a clue, sorry. Can you do a 
>   
>   $ cat /install/jdk1.1.8/include/<SOME_FILE_NAME>

/install/jdk-1.1.8/include> cat bool.h
/*
 * @(#)bool.h   1.6 98/07/01
 *
 * Copyright 1995-1998 by Sun Microsystems, Inc.,
 * 901 San Antonio Road, Palo Alto, California, 94303, U.S.A.
 * All rights reserved.
 *
 * This software is the confidential and proprietary information
 * of Sun Microsystems, Inc. ("Confidential Information").  You
 * shall not disclose such Confidential Information and shall use
 * it only in accordance with the terms of the license agreement
 * you entered into with Sun.
 */

#ifndef _BOOL_H_
#define _BOOL_H_

#undef  TRUE
#undef  FALSE

typedef enum {
    FALSE = 0,
    TRUE = 1
} bool_t;

#endif /* !_BOOL_H_ */


=====
Earnie Boyd < mailto:earnie_boyd@yahoo.com >
Cygwin Newbies, please visit
< http://www.freeyellow.com/members5/gw32/index.html >
__________________________________________________
Do You Yahoo!?
Thousands of Stores.  Millions of Products.  All in one place.
Yahoo! Shopping: http://shopping.yahoo.com

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Earnie Boyd]

--- Mumit Khan <khan@NanoTech.Wisc.EDU> wrote:
> On Mon, 6 Dec 1999, Earnie Boyd wrote:
> 
> > This bug is present in both the Cygwin V1 CD and gcc-2.95.2.  I found this
> > trying out Mumits JNI examples.
> > 
> > Since I don't know if this is Cygwin specific or not then I haven't
> forwarded
> > this to the gcc-bugs list.
> > 
> > -- Output of make --
> > 
> > gcc  -c -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32
> -g
> >  -Wall -O2 -v  -o invoke.o invoke.c
> 
> Ok, let's see what gcc sees. You're telling it to search directories
> /install/jdk1.1.8/include and /install/jdk1.1.8/include/win32 for
> jdk includes. So far so good.
> 
> > Reading specs from
> > /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/spe
> > cs
> [ ... ]
> > GNU CPP version 2.95.2 19991024 (release) (80386, BSD syntax)
> > #include "..." search starts here:
> > #include <...> search starts here:
> >  .
> > 
> /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> > 
> >
>
/install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygw
> > in/include
> >  /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/include
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygwin/include
> >  /usr/include
> 
> Now note the problem here. There are no signs of either of the jdk
> include directories!
> 
> 
> > End of search list.
> > The following default directories have been omitted from the search path:
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include/g++-3
> 
> Nor are they here, which would be the case normally if the paths were
> inaccessible.
> 
> I really don't have a clue, sorry. Can you do a 
>   
>   $ cat /install/jdk1.1.8/include/<SOME_FILE_NAME>

/install/jdk-1.1.8/include> cat bool.h
/*
 * @(#)bool.h   1.6 98/07/01
 *
 * Copyright 1995-1998 by Sun Microsystems, Inc.,
 * 901 San Antonio Road, Palo Alto, California, 94303, U.S.A.
 * All rights reserved.
 *
 * This software is the confidential and proprietary information
 * of Sun Microsystems, Inc. ("Confidential Information").  You
 * shall not disclose such Confidential Information and shall use
 * it only in accordance with the terms of the license agreement
 * you entered into with Sun.
 */

#ifndef _BOOL_H_
#define _BOOL_H_

#undef  TRUE
#undef  FALSE

typedef enum {
    FALSE = 0,
    TRUE = 1
} bool_t;

#endif /* !_BOOL_H_ */


=====
Earnie Boyd < mailto:earnie_boyd@yahoo.com >
Cygwin Newbies, please visit
< http://www.freeyellow.com/members5/gw32/index.html >
__________________________________________________
Do You Yahoo!?
Thousands of Stores.  Millions of Products.  All in one place.
Yahoo! Shopping: http://shopping.yahoo.com

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Earnie Boyd]

--- Mumit Khan <khan@NanoTech.Wisc.EDU> wrote:
> On Mon, 6 Dec 1999, Earnie Boyd wrote:
> 
> > This bug is present in both the Cygwin V1 CD and gcc-2.95.2.  I found this
> > trying out Mumits JNI examples.
> > 
> > Since I don't know if this is Cygwin specific or not then I haven't
> forwarded
> > this to the gcc-bugs list.
> > 
> > -- Output of make --
> > 
> > gcc  -c -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32
> -g
> >  -Wall -O2 -v  -o invoke.o invoke.c
> 
> Ok, let's see what gcc sees. You're telling it to search directories
> /install/jdk1.1.8/include and /install/jdk1.1.8/include/win32 for
> jdk includes. So far so good.
> 
> > Reading specs from
> > /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/spe
> > cs
> [ ... ]
> > GNU CPP version 2.95.2 19991024 (release) (80386, BSD syntax)
> > #include "..." search starts here:
> > #include <...> search starts here:
> >  .
> > 
> /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> > 
> >
>
/install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygw
> > in/include
> >  /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/include
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygwin/include
> >  /usr/include
> 
> Now note the problem here. There are no signs of either of the jdk
> include directories!
> 
> 
> > End of search list.
> > The following default directories have been omitted from the search path:
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include/g++-3
> 
> Nor are they here, which would be the case normally if the paths were
> inaccessible.
> 
> I really don't have a clue, sorry. Can you do a 
>   
>   $ cat /install/jdk1.1.8/include/<SOME_FILE_NAME>

Well, I found _MY_ bug.  I had specified the path incorrectly.  It should have
been jdk-1.1.8 instead of jdk1.1.8.

I've another problem though.  In building Mumits JAVA-JNI examples I'm getting
the missing WinMain@16 error from dllwrap which comes from the CD version.

Any clues?

=====
Earnie Boyd < mailto:earnie_boyd@yahoo.com >
Cygwin Newbies, please visit
< http://www.freeyellow.com/members5/gw32/index.html >
__________________________________________________
Do You Yahoo!?
Thousands of Stores.  Millions of Products.  All in one place.
Yahoo! Shopping: http://shopping.yahoo.com

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Mumit Khan]

Earnie Boyd <earnie_boyd@yahoo.com> writes:
> 
> I've another problem though.  In building Mumits JAVA-JNI examples I'm gettin
> g
> the missing WinMain@16 error from dllwrap which comes from the CD version.
> 
> Any clues?

It's an unfortunate bug in dllwrap that checks for cygwin32 in target
name instead just cygwin; I'll get it fixed for next release. 

Here's the solution.

   Linkname: Solution to DLLWRAP problem in Cygwin v1.0 CD
   URL: http://sourceware.cygnus.com/ml/cygwin/1999-11/msg00537.html

Regards,
Mumit


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Chris Faylor]

On Mon, Dec 13, 1999 at 05:55:33PM -0600, Mumit Khan wrote:
>Earnie Boyd <earnie_boyd@yahoo.com> writes:
>> 
>> I've another problem though.  In building Mumits JAVA-JNI examples I'm gettin
>> g
>> the missing WinMain@16 error from dllwrap which comes from the CD version.
>> 
>> Any clues?
>
>It's an unfortunate bug in dllwrap that checks for cygwin32 in target
>name instead just cygwin; I'll get it fixed for next release. 

I don't know if anyone has noticed that you may not need dllwrap as much
in the CD release.  DJ added functionality to ld which should make it
a lot easier to build DLLs.  Essentially you can say something like:

ld -shared -o foo.dll --export-all-symbols foo.o bar.o

cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Mumit Khan]

On Mon, 13 Dec 1999, Chris Faylor wrote:

> I don't know if anyone has noticed that you may not need dllwrap as much
> in the CD release.  DJ added functionality to ld which should make it
> a lot easier to build DLLs.  Essentially you can say something like:
> 
> ld -shared -o foo.dll --export-all-symbols foo.o bar.o

Actually, there's a very good reason why I never brought this up --
it's because the --shared support in v1 ld has certain known problems 
that will generate lots of unnecessary bug reports. These problems 
have been fixed in the sourceware tree, but v1 was cut before that 
point.

Try the following:
  
  /* v1-dll.c */
  __attribute__((dllexport)) int foo;

Now create the DLL.

  $ gcc -c v1-dll.c
  $ ld --shared -o v1-dll.dll v1-dll.o
  Cannot export foo: symbol wrong type (5 vs 3)

ie., can't handle uninitialized commons. 

If you leave out the __attribute__((declspec)):

  /* v1-dll.c */
  int foo;

And use --export-all:

  $ gcc -c v1-dll.c
  $ ld --export --shared -o v1-dll.dll v1-dll.o

Looks ok, but the dll doesn't export foo as expected and you get a 
runtime error.

The problem has already been fixed, but not in v1. Search the binutils
mailing list for "pe-dll". Patch submitted on 1999-06-25 and accepted
on 1999-09-28. I believe there are a few other issues as well that
have been fixed since (something about adding relocs to .exe that
I vaguely remember running into while testing v1 after getting the
CD).

My suggestion is to use either (1) dllwrap, or (2) use ld/dlltool 
multiple times; (2) requires that you also remember to supply the
correct entry point, which is automatically taken care of by
dllwrap. 

Regards,
Mumit



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Chris Faylor]

On Mon, Dec 13, 1999 at 07:06:54PM -0600, Mumit Khan wrote:
>On Mon, 13 Dec 1999, Chris Faylor wrote:
>
>> I don't know if anyone has noticed that you may not need dllwrap as much
>> in the CD release.  DJ added functionality to ld which should make it
>> a lot easier to build DLLs.  Essentially you can say something like:
>> 
>> ld -shared -o foo.dll --export-all-symbols foo.o bar.o
>
>Actually, there's a very good reason why I never brought this up --
>it's because the --shared support in v1 ld has certain known problems 
>that will generate lots of unnecessary bug reports. These problems 
>have been fixed in the sourceware tree, but v1 was cut before that 
>point.

Well, as an FYI, the DLL on the CD was linked using this method as
were all of the DLLs in the perl that was distributed on the CD.

>The problem has already been fixed, but not in v1. Search the binutils
>mailing list for "pe-dll". Patch submitted on 1999-06-25 and accepted
>on 1999-09-28. I believe there are a few other issues as well that   
>have been fixed since (something about adding relocs to .exe that 
>I vaguely remember running into while testing v1 after getting the
>CD).

This falls into the category of problems that could easily have been
rectified had I known they existed.  I guess I'd better start reading
the binutils mailing list.

>My suggestion is to use either (1) dllwrap, or (2) use ld/dlltool
>multiple times; (2) requires that you also remember to supply the
>correct entry point, which is automatically taken care of by
>dllwrap.

Hmm.  Personally, I'd still go with the simple 'ld -shared' method
as it seems that both 1) and 2) have drawbacks as well.

cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Chris Faylor]

On Mon, Dec 13, 1999 at 07:06:54PM -0600, Mumit Khan wrote:
>On Mon, 13 Dec 1999, Chris Faylor wrote:
>
>> I don't know if anyone has noticed that you may not need dllwrap as much
>> in the CD release.  DJ added functionality to ld which should make it
>> a lot easier to build DLLs.  Essentially you can say something like:
>> 
>> ld -shared -o foo.dll --export-all-symbols foo.o bar.o
>
>Actually, there's a very good reason why I never brought this up --
>it's because the --shared support in v1 ld has certain known problems 
>that will generate lots of unnecessary bug reports. These problems 
>have been fixed in the sourceware tree, but v1 was cut before that 
>point.

Well, as an FYI, the DLL on the CD was linked using this method as
were all of the DLLs in the perl that was distributed on the CD.

>The problem has already been fixed, but not in v1. Search the binutils
>mailing list for "pe-dll". Patch submitted on 1999-06-25 and accepted
>on 1999-09-28. I believe there are a few other issues as well that   
>have been fixed since (something about adding relocs to .exe that 
>I vaguely remember running into while testing v1 after getting the
>CD).

This falls into the category of problems that could easily have been
rectified had I known they existed.  I guess I'd better start reading
the binutils mailing list.

>My suggestion is to use either (1) dllwrap, or (2) use ld/dlltool
>multiple times; (2) requires that you also remember to supply the
>correct entry point, which is automatically taken care of by
>dllwrap.

Hmm.  Personally, I'd still go with the simple 'ld -shared' method
as it seems that both 1) and 2) have drawbacks as well.

cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Mumit Khan]

On Mon, 13 Dec 1999, Chris Faylor wrote:

> I don't know if anyone has noticed that you may not need dllwrap as much
> in the CD release.  DJ added functionality to ld which should make it
> a lot easier to build DLLs.  Essentially you can say something like:
> 
> ld -shared -o foo.dll --export-all-symbols foo.o bar.o

Actually, there's a very good reason why I never brought this up --
it's because the --shared support in v1 ld has certain known problems 
that will generate lots of unnecessary bug reports. These problems 
have been fixed in the sourceware tree, but v1 was cut before that 
point.

Try the following:
  
  /* v1-dll.c */
  __attribute__((dllexport)) int foo;

Now create the DLL.

  $ gcc -c v1-dll.c
  $ ld --shared -o v1-dll.dll v1-dll.o
  Cannot export foo: symbol wrong type (5 vs 3)

ie., can't handle uninitialized commons. 

If you leave out the __attribute__((declspec)):

  /* v1-dll.c */
  int foo;

And use --export-all:

  $ gcc -c v1-dll.c
  $ ld --export --shared -o v1-dll.dll v1-dll.o

Looks ok, but the dll doesn't export foo as expected and you get a 
runtime error.

The problem has already been fixed, but not in v1. Search the binutils
mailing list for "pe-dll". Patch submitted on 1999-06-25 and accepted
on 1999-09-28. I believe there are a few other issues as well that
have been fixed since (something about adding relocs to .exe that
I vaguely remember running into while testing v1 after getting the
CD).

My suggestion is to use either (1) dllwrap, or (2) use ld/dlltool 
multiple times; (2) requires that you also remember to supply the
correct entry point, which is automatically taken care of by
dllwrap. 

Regards,
Mumit



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Chris Faylor]

On Mon, Dec 13, 1999 at 05:55:33PM -0600, Mumit Khan wrote:
>Earnie Boyd <earnie_boyd@yahoo.com> writes:
>> 
>> I've another problem though.  In building Mumits JAVA-JNI examples I'm gettin
>> g
>> the missing WinMain@16 error from dllwrap which comes from the CD version.
>> 
>> Any clues?
>
>It's an unfortunate bug in dllwrap that checks for cygwin32 in target
>name instead just cygwin; I'll get it fixed for next release. 

I don't know if anyone has noticed that you may not need dllwrap as much
in the CD release.  DJ added functionality to ld which should make it
a lot easier to build DLLs.  Essentially you can say something like:

ld -shared -o foo.dll --export-all-symbols foo.o bar.o

cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Mumit Khan]

Earnie Boyd <earnie_boyd@yahoo.com> writes:
> 
> I've another problem though.  In building Mumits JAVA-JNI examples I'm gettin
> g
> the missing WinMain@16 error from dllwrap which comes from the CD version.
> 
> Any clues?

It's an unfortunate bug in dllwrap that checks for cygwin32 in target
name instead just cygwin; I'll get it fixed for next release. 

Here's the solution.

   Linkname: Solution to DLLWRAP problem in Cygwin v1.0 CD
   URL: http://sourceware.cygnus.com/ml/cygwin/1999-11/msg00537.html

Regards,
Mumit


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [GCC BUG] -I switch not functioning correctly.

[Earnie Boyd]

--- Mumit Khan <khan@NanoTech.Wisc.EDU> wrote:
> On Mon, 6 Dec 1999, Earnie Boyd wrote:
> 
> > This bug is present in both the Cygwin V1 CD and gcc-2.95.2.  I found this
> > trying out Mumits JNI examples.
> > 
> > Since I don't know if this is Cygwin specific or not then I haven't
> forwarded
> > this to the gcc-bugs list.
> > 
> > -- Output of make --
> > 
> > gcc  -c -I. -I/install/jdk1.1.8/include -I/install/jdk1.1.8/include/win32
> -g
> >  -Wall -O2 -v  -o invoke.o invoke.c
> 
> Ok, let's see what gcc sees. You're telling it to search directories
> /install/jdk1.1.8/include and /install/jdk1.1.8/include/win32 for
> jdk includes. So far so good.
> 
> > Reading specs from
> > /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/spe
> > cs
> [ ... ]
> > GNU CPP version 2.95.2 19991024 (release) (80386, BSD syntax)
> > #include "..." search starts here:
> > #include <...> search starts here:
> >  .
> > 
> /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> > 
> >
>
/install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygw
> > in/include
> >  /install/gcc-2.95.2/bin/../lib/gcc-lib/i686-cygwin/2.95.2/include
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../i686-cygwin/include
> >  /usr/include
> 
> Now note the problem here. There are no signs of either of the jdk
> include directories!
> 
> 
> > End of search list.
> > The following default directories have been omitted from the search path:
> >  /usr/lib/gcc-lib/i686-cygwin/2.95.2/../../../../include/g++-3
> 
> Nor are they here, which would be the case normally if the paths were
> inaccessible.
> 
> I really don't have a clue, sorry. Can you do a 
>   
>   $ cat /install/jdk1.1.8/include/<SOME_FILE_NAME>

Well, I found _MY_ bug.  I had specified the path incorrectly.  It should have
been jdk-1.1.8 instead of jdk1.1.8.

I've another problem though.  In building Mumits JAVA-JNI examples I'm getting
the missing WinMain@16 error from dllwrap which comes from the CD version.

Any clues?

=====
Earnie Boyd < mailto:earnie_boyd@yahoo.com >
Cygwin Newbies, please visit
< http://www.freeyellow.com/members5/gw32/index.html >
__________________________________________________
Do You Yahoo!?
Thousands of Stores.  Millions of Products.  All in one place.
Yahoo! Shopping: http://shopping.yahoo.com

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com