Problems building TCL for insight on cygwin & win2k

Problems building TCL for insight on cygwin & win2k

[Andy Hare]

I have been trying to build insight under cygwin on a windows 2k box. The
source was obtained from the cvs server last night. The build fails whilst
building TCL. The configure line and make line are given below :

 cd /build/arm ; \
 rm -r * ; \
 CC="gcc -mwin32" /GNU-Source-Code/src/configure -v \
   --prefix=/xgcc-arm/gnu \
   --exec-prefix=/xgcc-arm/gnu \
   --with-headers=/Gnu-Source-Code/src/newlib/libc/include \
   --host=i586-cygwin32 --target=arm-elf >$(HOME)/configure.out 2>&1


 cd /build/arm ; \
 make -w all-gdb install-gdb CC='gcc -mwin32' >$(HOME)/make-gdb.out 2>&1


The error occours as below :

gcc -mwin32 -DPACKAGE=\"libtcl_cygwin\" -DVERSION=\"8.3\" -DHAVE_GETCWD=1 -D
HAVE_OPENDIR=1 -DHAVE_STRSTR=1 -DHAVE_STRTOL=1 -DHAVE_TMPNAM=1 -DHAVE_WAITPI
D=1 -DNO_VALUES_H=1 -DHAVE_UNISTD_H=1 -DUSE_TERMIOS=1 -DHAVE_SYS_TIME_H=1 -D
TIME_WITH_SYS_TIME=1 -DHAVE_ST_BLKSIZE=1 -DSTDC_HEADERS=1 -DNO_UNION_WAIT=1 
-DNEED_MATHERR=1 -DRETSIGTYPE=void -DHAVE_SIGNED_CHAR=1 -DHAVE_SYS_IOCTL_H=1
  -I. -I/GNU-Source-Code/src/tcl/cygwin  -I/GNU-Source-Code/src/tcl/cygwin/.
./generic   -D__TCL_UNIX_VARIANT -DTCL_LIBRARY='"/xgcc-arm/gnu/share/tcl8.3"
' -DTCL_PACKAGE_PATH='"/xgcc-arm/gnu/lib"' -g  -c
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c
In file included from
/GNU-Source-Code/src/tcl/cygwin/../generic/../win/tclWinPort.h:68,
                 from
/GNU-Source-Code/src/tcl/cygwin/../generic/tclPort.h:22,
                 from
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:17:
/usr/include/w32api/winsock2.h:92: warning: #warning "fd_set and associated
macros have been defined in sys/types.      This may cause runtime problems
with W32 sockets"
In file included from /usr/include/termios.h:4,
                 from
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:47:
/usr/include/sys/termios.h:46: warning: `FIONBIO' redefined
/usr/include/w32api/winsock2.h:128: warning: this is the location of the
previous definition
In file included from /usr/include/w32api/windef.h:145,
                 from /usr/include/w32api/windows.h:98,
                 from /usr/include/w32api/winsock2.h:18,
                 from
/GNU-Source-Code/src/tcl/cygwin/../generic/../win/tclWinPort.h:68,
                 from
/GNU-Source-Code/src/tcl/cygwin/../generic/tclPort.h:22,
                 from
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:17:
/usr/include/w32api/winnt.h:45: warning: redefinition of `CHAR'
/GNU-Source-Code/src/tcl/cygwin/../generic/tcl.h:318: warning: `CHAR'
previously declared here
/usr/include/w32api/winnt.h:46: warning: redefinition of `SHORT'
/GNU-Source-Code/src/tcl/cygwin/../generic/tcl.h:319: warning: `SHORT'
previously declared here
/usr/include/w32api/winnt.h:47: warning: redefinition of `LONG'
/GNU-Source-Code/src/tcl/cygwin/../generic/tcl.h:320: warning: `LONG'
previously declared here
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:235: warning:
initialization from incompatible pointer type
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:254: warning:
initialization from incompatible pointer type
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:273: warning:
initialization from incompatible pointer type
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c: In function
`TclUnixWaitForFile':
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:2626: `MASK_SIZE'
undeclared (first use in this function)
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:2626: (Each undeclared
identifier is reported only once
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:2626: for each
function it appears in.)
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:2626: size of array
`readyMasks' has non-integer type
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:2701: `SELECT_MASK'
undeclared (first use in this function)
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:2701: parse error
before `)'
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:2702: parse error
before `)'
/GNU-Source-Code/src/tcl/cygwin/../unix/tclUnixChan.c:2703: parse error
before `)'
make[3]: *** [tclUnixChan.o] Error 1

Checking the make files generated by the configure shows that
tcl/cygwin/makefile is using files from the tcl/unix sub directory but the
include files are comming from the tcl/win sub-directory due to the
CC=gc -mwin32 part.

Can any one tell me if the TCl should use the win or unix directories for
its build files and how I can get the right set of files to build. Before
putting the -mwin32 part into the config and make lines the tcl build failed
with some macros WIFEXIST etc having the wrong types defined.

I hope you can help as I have been trying to get insight built for the last
week.

Andy Hare
www.ahare.btinternet.co.uk

Re: Problems building TCL for insight on cygwin & win2k

[Ian Roxborough]

Andy Hare wrote:
> 
> I have been trying to build insight under cygwin on a windows 2k box. The
> source was obtained from the cvs server last night. The build fails whilst
> building TCL. The configure line and make line are given below :
> 
>  cd /build/arm ; \
>  rm -r * ; \
>  CC="gcc -mwin32" /GNU-Source-Code/src/configure -v \

The "-mwin32" is confusing the tcl/cygwin build.  Try again but
without the -mwin32.

However, the current Win32 CVS build is suffering from strange crashing
problems just now.  But it should build.

Ian.

Re: Problems building TCL for insight on cygwin & win2k

[Keith Seitz]

On Thu, 25 Oct 2001, Andy Hare wrote:

> I have been trying to build insight under cygwin on a windows 2k box. The
> source was obtained from the cvs server last night. The build fails whilst
> building TCL. The configure line and make line are given below :
>
>  cd /build/arm ; \
>  rm -r * ; \
>  CC="gcc -mwin32" /GNU-Source-Code/src/configure -v \
>    --prefix=/xgcc-arm/gnu \
>    --exec-prefix=/xgcc-arm/gnu \
>    --with-headers=/Gnu-Source-Code/src/newlib/libc/include \
>    --host=i586-cygwin32 --target=arm-elf >$(HOME)/configure.out 2>&1

Hmmm. What version of cygwin are you using? (uname -a) Right now, you can
ONLY use a current net release of cygwin to build the public repository.
(This should be the case for everyone except Red Hat GNUPro customers.)

I know I just built this thing yesterday without incident.

I'm updating my sources now, just in case, but I can see some obvious
differences between the way you have configured your toolchain and the way
I do it. Of course, this may not be the problem, but who knows? It's
worked once this week already. ;-)

Actually, I'm gonna say almost the same thing that I told the last person
with configure problems. When configuring, either specify "--target" alone
OR specify "--target", "--host", and "--build". Never use just two of
them. This is empirical only. YMMV.

I think that if you did this, your problem would go away. I don't think
that "i586-cygwin32" is the appropriate host triple nowadays. I use
"i686-pc-cgywin".

I also never use the 'CC="gcc -mwin32"' thing. I know this was needed some
time in the past, but I've never needed it at all in the last few months.

Finally, the last thing that causes me some concern.. Why do you need to
use "--with-headers"? I've never used that before except when doing a
cross build for another OS.

So, try using:

  /GNU-Source-Code/src/configure -v \
    --prefix=/xgcc-arm/gnu \
    --exec-prefix=/xgcc-arm/gnu \
    --target=arm-elf >$(HOME)/configure.out 2>&1

Let us know if it helps.

> I hope you can help as I have been trying to get insight built for the last
> week.

You should have spoken up sooner...

Keith

Re: Problems building TCL for insight on cygwin & win2k

[Christopher Faylor]

On Thu, Oct 25, 2001 at 12:30:00PM -0700, Ian Roxborough wrote:
>Andy Hare wrote:
>> 
>> I have been trying to build insight under cygwin on a windows 2k box. The
>> source was obtained from the cvs server last night. The build fails whilst
>> building TCL. The configure line and make line are given below :
>> 
>>  cd /build/arm ; \
>>  rm -r * ; \
>>  CC="gcc -mwin32" /GNU-Source-Code/src/configure -v \
>
>The "-mwin32" is confusing the tcl/cygwin build.  Try again but
>without the -mwin32.

Um.  It's not confusing *my* build.  That's why I added it.
You can't build a native windows build without -mwin32 if you're
using a recent compiler.

Just removing it is not going to be a long term solution.

cgf

Re: Problems building TCL for insight on cygwin & win2k

[Christopher Faylor]

On Thu, Oct 25, 2001 at 03:29:19PM -0400, Christopher Faylor wrote:
>On Thu, Oct 25, 2001 at 12:30:00PM -0700, Ian Roxborough wrote:
>>Andy Hare wrote:
>>> 
>>> I have been trying to build insight under cygwin on a windows 2k box. The
>>> source was obtained from the cvs server last night. The build fails whilst
>>> building TCL. The configure line and make line are given below :
>>> 
>>>  cd /build/arm ; \
>>>  rm -r * ; \
>>>  CC="gcc -mwin32" /GNU-Source-Code/src/configure -v \
>>
>>The "-mwin32" is confusing the tcl/cygwin build.  Try again but
>>without the -mwin32.
>
>Um.  It's not confusing *my* build.  That's why I added it.
>You can't build a native windows build without -mwin32 if you're
>using a recent compiler.
>
>Just removing it is not going to be a long term solution.

Just to clarify -- the -mwin32 should *already* be added to the CFLAGS
of a build of a cygwin-hosted toolchain.

cgf

Re: Problems building TCL for insight on cygwin & win2k

[Ian Roxborough]

Christopher Faylor wrote:
> 
> On Thu, Oct 25, 2001 at 12:30:00PM -0700, Ian Roxborough wrote:
> >>  CC="gcc -mwin32" /GNU-Source-Code/src/configure -v \
> >
> >The "-mwin32" is confusing the tcl/cygwin build.  Try again but
> >without the -mwin32.
> 
> Um.  It's not confusing *my* build.  That's why I added it.
> You can't build a native windows build without -mwin32 if you're
> using a recent compiler.
> 
> Just removing it is not going to be a long term solution.

True.  I thought that the tcl/cygwin build would override that
with -mno-win32 anyway.  I guess the problem is that the -mno-win32
is missing.  I'm not sure why either.

Ian.

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

Keith,

Thanks for your input I am trying the configure again without the host=
line. The reason for the with-headers is that I am using newlib and the
compiler is to be used on an embedded ARM micro, the configure actually
configs the whole toolchain, binutils, gcc and insight.

uname -a gives
CYGWIN_NT-5.0 DUAL-BEAST 1.3.3(0.46/3/2) 2001-09-12 23:54 i686 unknown

Will let you know what happens with the new build when it finishes.

Andy Hare
www.ahare.btinternet.co.uk

----- Original Message -----
From: "Keith Seitz" <keiths@cygnus.com>
>
> Hmmm. What version of cygwin are you using? (uname -a) Right now, you can
> ONLY use a current net release of cygwin to build the public repository.
> (This should be the case for everyone except Red Hat GNUPro customers.)
>
> I know I just built this thing yesterday without incident.
>
> I'm updating my sources now, just in case, but I can see some obvious
> differences between the way you have configured your toolchain and the way
> I do it. Of course, this may not be the problem, but who knows? It's
> worked once this week already. ;-)
>
> Actually, I'm gonna say almost the same thing that I told the last person
> with configure problems. When configuring, either specify "--target" alone
> OR specify "--target", "--host", and "--build". Never use just two of
> them. This is empirical only. YMMV.
>
> I think that if you did this, your problem would go away. I don't think
> that "i586-cygwin32" is the appropriate host triple nowadays. I use
> "i686-pc-cgywin".
>
> I also never use the 'CC="gcc -mwin32"' thing. I know this was needed some
> time in the past, but I've never needed it at all in the last few months.
>
> Finally, the last thing that causes me some concern.. Why do you need to
> use "--with-headers"? I've never used that before except when doing a
> cross build for another OS.
>
> So, try using:
>
>   /GNU-Source-Code/src/configure -v \
>     --prefix=/xgcc-arm/gnu \
>     --exec-prefix=/xgcc-arm/gnu \
>     --target=arm-elf >$(HOME)/configure.out 2>&1
>
> Let us know if it helps.
>
> > I hope you can help as I have been trying to get insight built for the
last
> > week.
>
> You should have spoken up sooner...
>
> Keith
>
>
>

Re: Problems building TCL for insight on cygwin & win2k

[Keith Seitz]

On Thu, 25 Oct 2001, Andy Hare wrote:

> Thanks for your input I am trying the configure again without the host=
> line. The reason for the with-headers is that I am using newlib and the
> compiler is to be used on an embedded ARM micro, the configure actually
> configs the whole toolchain, binutils, gcc and insight.

Ah. Okay. That makes sense.

> uname -a gives
> CYGWIN_NT-5.0 DUAL-BEAST 1.3.3(0.46/3/2) 2001-09-12 23:54 i686 unknown

I just updated and rebuilt this using the same cygwin and gcc (current net
release). It worked fine.

> Will let you know what happens with the new build when it finishes.

Please remember that the trunk is experiencing some windows problems. You
might consider using the 5.1 branch instead, unless you'd like to help us
track some of these problems down. :-)

Keith

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

Keith,

----- Original Message -----
From: "Keith Seitz" <keiths@cygnus.com>
To: "Andy Hare" <ahare@btinternet.com>
Cc: "Insight Mail list" <insight@sourceware.cygnus.com>
Sent: Thursday, October 25, 2001 10:00 PM
Subject: Re: Problems building TCL for insight on cygwin & win2k


> On Thu, 25 Oct 2001, Andy Hare wrote:
>
> > Thanks for your input I am trying the configure again without the host=
> > line. The reason for the with-headers is that I am using newlib and the
> > compiler is to be used on an embedded ARM micro, the configure actually
> > configs the whole toolchain, binutils, gcc and insight.
>
> Ah. Okay. That makes sense.
>
> > uname -a gives
> > CYGWIN_NT-5.0 DUAL-BEAST 1.3.3(0.46/3/2) 2001-09-12 23:54 i686 unknown
>
> I just updated and rebuilt this using the same cygwin and gcc (current net
> release). It worked fine.

The makefile that is in (my case) /build/arm/tcl/cygwin has tclunix*
filenames within it, is this as you have built or did your make file use the
tclwin* files. i.e. the line:

libtcl_cygwin_a_SOURCES =   regcomp.c  regexec.c         regfree.c
regerror.c  tclAlloc.c  tclAsync.c  tclBasic.c  tclBinary.c  tclCkalloc.c
tclClock.c  tclCmdAH.c  tclCmdIL.c  tclCmdMZ.c         tclCompCmds.c
tclCompExpr.c  tclCompile.c  tclDate.c         tclEncoding.c  tclEnv.c
tclEvent.c  tclExecute.c  tclFCmd.c  tclFileName.c  tclGet.c  tclHash.c
tclHistory.c  tclIndexObj.c  tclInterp.c  tclIO.c  tclIOCmd.c
tclIOGT.c  tclIOSock.c  tclIOUtil.c  tclLink.c  tclListObj.c  tclLiteral.c
tclLoad.c  tclLoadNone.c  tclMain.c  tclNamesp.c  tclNotify.c  tclObj.c
tclParse.c  tclPanic.c  tclParseExpr.c  tclPipe.c  tclPkg.c  tclPosixStr.c
tclPreserve.c  tclProc.c  tclRegexp.c  tclTestProcBodyObj.c  tclResolve.c
tclResolve.c         tclResult.c  tclScan.c  tclStringObj.c
tclStubInit.c         tclStubLib.c         tclThread.c  tclTimer.c  tclUtf.c
tclUtil.c  tclVar.c  tclAppInit.c  tclUnixChan.c  tclUnixEvent.c
tclUnixFCmd.c  tclUnixFile.c  tclUnixNotfy.c  tclUnixPipe.c  tclUnixSock.c
tclUnixTest.c         tclUnixThrd.c  tclUnixTime.c  tclUnixInit.c

 I ask because the tclwin*.h files are then used in these files and there
are some missing #defines in the tclwin*.h files that exist in the
tclunix*.h files. Ian Roxborough mentioned a missing -mno-win32, that
certainly is missing from my makefiles and would ensure that the tclunix*.h
file were included.

>
> > Will let you know what happens with the new build when it finishes.
>
> Please remember that the trunk is experiencing some windows problems. You
> might consider using the 5.1 branch instead, unless you'd like to help us
> track some of these problems down. :-)
>

I am quite happy to help track the problems if I can. At the begining of the
year I was doing weekly builds of insight and the GNU toolchain but since
our first son was born and the MSc thesis I have been writing, life seems to
run out of time. I am now trying to keep up with the CVS and build the
toolset each week and then report any problems I find.

Andy Hare
www.ahare.btinternet.co.uk

Re: Problems building TCL for insight on cygwin & win2k

[Keith Seitz]

On Thu, 25 Oct 2001, Andy Hare wrote:

> The makefile that is in (my case) /build/arm/tcl/cygwin has tclunix*
> filenames within it, is this as you have built or did your make file use the
> tclwin* files. i.e. the line:
>
> libtcl_cygwin_a_SOURCES =   regcomp.c  regexec.c         regfree.c
> regerror.c  tclAlloc.c  tclAsync.c  tclBasic.c  tclBinary.c  tclCkalloc.c
> tclClock.c  tclCmdAH.c  tclCmdIL.c  tclCmdMZ.c         tclCompCmds.c
> tclCompExpr.c  tclCompile.c  tclDate.c         tclEncoding.c  tclEnv.c
> tclEvent.c  tclExecute.c  tclFCmd.c  tclFileName.c  tclGet.c  tclHash.c
> tclHistory.c  tclIndexObj.c  tclInterp.c  tclIO.c  tclIOCmd.c
> tclIOGT.c  tclIOSock.c  tclIOUtil.c  tclLink.c  tclListObj.c  tclLiteral.c
> tclLoad.c  tclLoadNone.c  tclMain.c  tclNamesp.c  tclNotify.c  tclObj.c
> tclParse.c  tclPanic.c  tclParseExpr.c  tclPipe.c  tclPkg.c  tclPosixStr.c
> tclPreserve.c  tclProc.c  tclRegexp.c  tclTestProcBodyObj.c  tclResolve.c
> tclResolve.c         tclResult.c  tclScan.c  tclStringObj.c
> tclStubInit.c         tclStubLib.c         tclThread.c  tclTimer.c  tclUtf.c
> tclUtil.c  tclVar.c  tclAppInit.c  tclUnixChan.c  tclUnixEvent.c
> tclUnixFCmd.c  tclUnixFile.c  tclUnixNotfy.c  tclUnixPipe.c  tclUnixSock.c
> tclUnixTest.c         tclUnixThrd.c  tclUnixTime.c  tclUnixInit.c

My build/tcl/cygwin/Makefile uses what you have above.

>  I ask because the tclwin*.h files are then used in these files and there
> are some missing #defines in the tclwin*.h files that exist in the
> tclunix*.h files. Ian Roxborough mentioned a missing -mno-win32, that
> certainly is missing from my makefiles and would ensure that the tclunix*.h
> file were included.

The sources should build for you as-is. Is this a clean build and
configure? (You're not reconfiguring an existing directory, for example.)

Can you send a snippet of the build procedure which shows what errors you
are seeing?

Keith

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

Have now re-built the config and the run the make in the /build/tcl
directory. Problem still exists with the missing #defines for MASK_SIZE and
SELECT_MASK. Tried Ians suggestion of -mno-win32 and now the tcl stuff
compiles without a problem. So I will look at the way tcl is configured to
see if I can find the problem and the missing -mno-win32.

Thanks for all your help, I will go back modify the make files in the tcl
and then continue.

Andy Hare
www.ahare.btinternet.co.uk

----- Original Message -----
From: "Keith Seitz" <keiths@cygnus.com>
To: "Andy Hare" <ahare@btinternet.com>
Cc: "Insight Mail list" <insight@sourceware.cygnus.com>
Sent: Thursday, October 25, 2001 10:00 PM
Subject: Re: Problems building TCL for insight on cygwin & win2k


> On Thu, 25 Oct 2001, Andy Hare wrote:
>
> > Thanks for your input I am trying the configure again without the host=
> > line. The reason for the with-headers is that I am using newlib and the
> > compiler is to be used on an embedded ARM micro, the configure actually
> > configs the whole toolchain, binutils, gcc and insight.
>
> Ah. Okay. That makes sense.
>
> > uname -a gives
> > CYGWIN_NT-5.0 DUAL-BEAST 1.3.3(0.46/3/2) 2001-09-12 23:54 i686 unknown
>
> I just updated and rebuilt this using the same cygwin and gcc (current net
> release). It worked fine.
>
> > Will let you know what happens with the new build when it finishes.
>
> Please remember that the trunk is experiencing some windows problems. You
> might consider using the 5.1 branch instead, unless you'd like to help us
> track some of these problems down. :-)
>
> Keith
>
>

Re: Problems building TCL for insight on cygwin & win2k

[Keith Seitz]

On Thu, 25 Oct 2001, Andy Hare wrote:

> Have now re-built the config and the run the make in the /build/tcl
> directory. Problem still exists with the missing #defines for MASK_SIZE and
> SELECT_MASK. Tried Ians suggestion of -mno-win32 and now the tcl stuff
> compiles without a problem. So I will look at the way tcl is configured to
> see if I can find the problem and the missing -mno-win32.

Hmm. I don't understand why this doesn't work for you, but it works for
me. What is gcc -v?

> Thanks for all your help, I will go back modify the make files in the tcl
> and then continue.

This should not be necessary. I am really confused about why it is.

Does the repository build for anyone else out there? I can do a checkout
and build it as-is with the latest cygwin (1.3.3) and the gcc distributed
with it (gcc-2.95.3-5 cygwin special).

I am confused.

Keith

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

>
> Hmm. I don't understand why this doesn't work for you, but it works for
> me. What is gcc -v?

$ gcc -v
Reading specs from /usr/lib/gcc-lib/i686-pc-cygwin/2.95.3-5/specs
gcc version 2.95.3-5 (cygwin special)


> > Thanks for all your help, I will go back modify the make files in the
tcl
> > and then continue.
>
> This should not be necessary. I am really confused about why it is.
>
> Does the repository build for anyone else out there? I can do a checkout
> and build it as-is with the latest cygwin (1.3.3) and the gcc distributed
> with it (gcc-2.95.3-5 cygwin special).
>
> I am confused.
>

I am still using the CC=gcc -mwin-32 config maybe this is the problem with
the build, I will try tomorrow to re-config without this instructin and see
how the build goes.

Andy Hare
www.ahare.btinternet.co.uk

Re: Problems building TCL for insight on cygwin & win2k

[Christopher Faylor]

On Thu, Oct 25, 2001 at 11:44:26PM +0100, Andy Hare wrote:
>>Hmm.  I don't understand why this doesn't work for you, but it works
>>for me.  What is gcc -v?
>
>$ gcc -v
>Reading specs from /usr/lib/gcc-lib/i686-pc-cygwin/2.95.3-5/specs
>gcc version 2.95.3-5 (cygwin special)
>
>>>Thanks for all your help, I will go back modify the make files in the
>>>tcl and then continue.
>>
>>This should not be necessary.  I am really confused about why it is.
>>
>>Does the repository build for anyone else out there?  I can do a
>>checkout and build it as-is with the latest cygwin (1.3.3) and the gcc
>>distributed with it (gcc-2.95.3-5 cygwin special).
>>
>>I am confused.
>
>I am still using the CC=gcc -mwin-32 config maybe this is the problem
>with the build, I will try tomorrow to re-config without this
>instructin and see how the build goes.

Actually, now I understand.  This *would* be a problem.  Ian's
suggestion was correct.  If you use this option in the cygwin directory
it will probably screw things up since the cygwin directory needs to use
the -mno-win32 option. So, you'll end up doing a:

gcc -mwwin32 -mno-win32 ...

who knows what that will do?

Sorry for confusing things by suggesting that it was incorrect to do
what Ian suggested.

cgf

Re: Problems building TCL for insight on cygwin & win2k

[Ian Roxborough]

Christopher Faylor wrote:
> Actually, now I understand.  This *would* be a problem.  Ian's
> suggestion was correct.  If you use this option in the cygwin directory
> it will probably screw things up since the cygwin directory needs to use
> the -mno-win32 option. So, you'll end up doing a:
> 
> gcc -mwwin32 -mno-win32 ...
> 
> who knows what that will do?

-mno-win32 will override -mwin32, I'm not sure why (maybe just flag order).
 
> Sorry for confusing things by suggesting that it was incorrect to do
> what Ian suggested.

I'm not sure my suggestion is correct.  This could be a bug in
tcl/cygwin/configure.in where some how -mno-win32 isn't being
set in the tcl/cygwin build.  I think you are right, setting
CC="gcc -mwin32" shouldn't break the build.

Ian.

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

Keith,

    I have now re-configed into an empty directory and re-built without
the -mwin32 flags. I now get the errors shown below:

gcc -c -g -O2 -Wall -Wconversion  -I"/GNU-Source-Code/src/tcl/win/../generic
" -I"/GNU-Source-Code/src/tcl/win" -DHAVE_SYS_TIME_H=1 -DTIME_WITH_SYS_TIME=
1 -DHAVE_TZNAME=1   -DBUILD_tcl
"/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c" -o tclPipe.o
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c: In function
`TclCleanupChildren':
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:313: invalid operands to
binary &
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:313: invalid operands to
binary >>
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:318: invalid operands to
binary &
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:320: invalid operands to
binary >>
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:325: invalid operands to
binary &
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:325: invalid operands to
binary &
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:329: invalid operands to
binary &
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:331: invalid operands to
binary &
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:336: invalid operands to
binary &
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:340: invalid operands to
binary >>
/GNU-Source-Code/src/tcl/win/../generic/tclPipe.c:342: invalid operands to
binary >>
make[3]: *** [tclPipe.o] Error 1

This is the problem I started with before adding the win32 flags. The
problem is that the lines use a set of macros defined in tclunixport.h that
assume that the value passed is an int, but the value passed isn't an int.
Here is the snippet of code fromthe .h file:


#ifdef _AIX
#   define WAIT_STATUS_TYPE pid_t
#else
#ifndef NO_UNION_WAIT
#   define WAIT_STATUS_TYPE union wait
#else
#   define WAIT_STATUS_TYPE int
#endif
#endif

/*
 * Supply definitions for macros to query wait status, if not already
 * defined in header files above.
 */

#ifndef WIFEXITED
#   define WIFEXITED(stat)  (((*((int *) &(stat))) & 0xff) == 0)
#endif

The value passed in the line that give erros in tclpipe.c use this WIFEXIT
but the value passed is of type WAIT_STATUS_TYPE and I guesss is of type
union wait rather than the int. So should there be a definition for
NO_UNION_WAIT or should the defines be within the #ifndef NO_UNION_WAIT so
they take account of the differeing types.

Andy Hare
www.ahare.btinternet.co.uk

----- Original Message -----
From: "Keith Seitz" <keiths@cygnus.com>
To: "Andy Hare" <ahare@btinternet.com>
Cc: "Insight Mail list" <insight@sourceware.cygnus.com>
Sent: Thursday, October 25, 2001 11:31 PM
Subject: Re: Problems building TCL for insight on cygwin & win2k


> On Thu, 25 Oct 2001, Andy Hare wrote:
>
> > Have now re-built the config and the run the make in the /build/tcl
> > directory. Problem still exists with the missing #defines for MASK_SIZE
and
> > SELECT_MASK. Tried Ians suggestion of -mno-win32 and now the tcl stuff
> > compiles without a problem. So I will look at the way tcl is configured
to
> > see if I can find the problem and the missing -mno-win32.
>
> Hmm. I don't understand why this doesn't work for you, but it works for
> me. What is gcc -v?
>
> > Thanks for all your help, I will go back modify the make files in the
tcl
> > and then continue.
>
> This should not be necessary. I am really confused about why it is.
>
> Does the repository build for anyone else out there? I can do a checkout
> and build it as-is with the latest cygwin (1.3.3) and the gcc distributed
> with it (gcc-2.95.3-5 cygwin special).
>
> I am confused.
>
> Keith
>
>

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

Ian,

----- Original Message -----
From: "Ian Roxborough" <irox@redhat.com>
> I'm not sure my suggestion is correct.  This could be a bug in
> tcl/cygwin/configure.in where some how -mno-win32 isn't being
> set in the tcl/cygwin build.  I think you are right, setting
> CC="gcc -mwin32" shouldn't break the build.

The makefile generated in tcl/cygwin is definitely not producing a flag
of -mno-win32, however removing the CC="gcc -mwin32" from both the config
generation line and the make line ensures that the tcl/cygwin/makefile uses
the ../unix includes and not the ../win includes. I would look at the config
scripts for the tcl/cygwin but I do not understand these scripts and would
not know what I am looking for, sorry.

Andy Hare
www.ahare.btinternet.co.uk

Re: Problems building TCL for insight on cygwin & win2k

[Ian Roxborough]

Hi Andy,

this is going to sound strange, but you're missing the -mwin32 flag
which should be automatically added to the command line.

Your problem doesn't seem to be with setting CC="gcc -mwin32", but
some how the -mwin32 and -mno-win32 flags are not being passed at
the right time.

Something is going wrong at configure time and the EXTRA_CFLAGS
variable used to pass -m*win32 isn't being set or isn't being passed.

They should be present in <src-dir>/tcl/cygwin/configure.in (grep
for mno-win32) and <src-dir>/tcl/win/configure.in (grep for mwin32).
Also check the value of EXTRA_CFLAGS in <build-dir>/tcl/unix/tclConfig.sh
and <build-dir>/tcl/win/tclConfig.sh.

These might yield some clues to the problem.

Ian.


Andy Hare wrote:
> 
> Keith,
> 
>     I have now re-configed into an empty directory and re-built without
> the -mwin32 flags. I now get the errors shown below:
> 
> gcc -c -g -O2 -Wall -Wconversion  -I"/GNU-Source-Code/src/tcl/win/../generic
> " -I"/GNU-Source-Code/src/tcl/win" -DHAVE_SYS_TIME_H=1 -DTIME_WITH_SYS_TIME=
> 1 -DHAVE_TZNAME=1   -DBUILD_tcl

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

Keith,

Have been trying some different things with the makefies within the tcl
directory. Biggest change was to do a make clean from within the tcl
directory and suddenly the build inside cygwin completes without a problem.
The difference is that no the compile lines have the define
'-DNO_UNION_WAIT=1' and that ensures that the tclunixport.h gets the right
definition of WAIT_STATUS_TYPE.

Another issue is that now doing a make clean in tcl, does a make all in the
cygwin subdirectory. It also does a build in the win subdirectory and I had
to add -mwin32 to the subdir makefile in order that the win32 header files
get included otherwise the build fails in these directories.

I am now getting an update of the cvs and will try a config again in a blank
build directory and then compare the makefiles that work currently to those
that get generated by the config.

Hope some if this helps others to get a better idea of what is going on.
Will report back when I have the new config done.

Andy Hare
www.ahare.btinternet.co.uk

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

Ian,
----- Original Message -----
From: "Ian Roxborough" <irox@redhat.com>
To: "Andy Hare" <ahare@btinternet.com>
Cc: "Keith Seitz" <keiths@cygnus.com>; "Insight Mail list"
<insight@sourceware.cygnus.com>
Sent: Friday, October 26, 2001 10:04 PM
Subject: Re: Problems building TCL for insight on cygwin & win2k


>
> Hi Andy,
>
> this is going to sound strange, but you're missing the -mwin32 flag
> which should be automatically added to the command line.
>
> Your problem doesn't seem to be with setting CC="gcc -mwin32", but
> some how the -mwin32 and -mno-win32 flags are not being passed at
> the right time.
>
> Something is going wrong at configure time and the EXTRA_CFLAGS
> variable used to pass -m*win32 isn't being set or isn't being passed.
>
> They should be present in <src-dir>/tcl/cygwin/configure.in (grep
> for mno-win32) and <src-dir>/tcl/win/configure.in (grep for mwin32).
> Also check the value of EXTRA_CFLAGS in <build-dir>/tcl/unix/tclConfig.sh
> and <build-dir>/tcl/win/tclConfig.sh.
>

Have been through the config scripts in both tcl/cygwin and tcl/win and
found the following. The scripts check for the cygwin build and then check
against the target name thus:
case "${target}" in
*-*-cygwin*)
 touch ac$$.c
 if ${CC} -c -mno-win32 ac$$.c >/dev/null 2>&1; then
     case "$EXTRA_CFLAGS" in
  *-mno-win32*) ;;
  *) EXTRA_CFLAGS="-mno-win32 $EXTRA_CFLAGS" ;;
     esac
 fi
 rm -f ac$$.o ac$$.c
 ;;
esac

but as I am building a cross version, i.e. hosted on cygwin but targetting
ARM it does not build the correct version. Modifying it to the following:

case "${host}" in
*cygwin*)
 touch ac$$.c
        if ${CC} -c -mwin32 ac$$.c >/dev/null 2>&1; then
            case "$EXTRA_CFLAGS" in
                *-mwin32*) ;;
                *) EXTRA_CFLAGS="-mwin32 $EXTRA_CFLAGS" ;;
            esac
        fi
        rm -f ac$$.o ac$$.c
        ;;
esac

allows the tcl stuff to build and the CC=gcc -mwin32 is then not needed.

I would produce a patch to this effect but do not know how to create the
patch in the correct format or where to post it to so hopefully someone with
better knowledge of the system than me can take this info and produce the
required.

I will check the rest of the configure scripts to see if there are any more
such issues as now I have reached another problem within tk.

Andy Hare
www.ahare.btinternet.co.uk

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

In one of my previous messages I stated that make clean in the tcl directory
caused the cygwin sub directory to do a make all. The faults is in
makefile.in in the tcl directory.

maintainer-clean-recursive:
 @cd $(CONFIGDIR) && $(MAKE) `echo $@ | sed 's/-recursive//'`
 @test x"$(CONFIGDIR2)" = x"" || (cd "$(CONFIGDIR2)" && $(MAKE) `echo $@ |
sed 's/-recursive'`)

Surely the second line should end sed 's/-recursive//' as does the first
line ?

Andy Hare
www.ahare.btinternet.co.uk

Re: Problems building TCL for insight on cygwin & win2k

[Christopher Faylor]

[NOTE: Reply-To is set to insight@sources.redhat.com]

On Sun, Oct 28, 2001 at 10:05:44AM -0000, Andy Hare wrote:
>I would produce a patch to this effect but do not know how to create the
>patch in the correct format or where to post it to so hopefully someone with
>better knowledge of the system than me can take this info and produce the
>required.

I appreciate the analysis and have checked in patches to rectify this behavior.

In the future, you can refer to the gdb/CONTRIBUTE document for information on
how to submit a patch.  It should provide information that will allow you
to contribute your own patches in the future.

cgf

Re: Problems building TCL for insight on cygwin & win2k

[Andy Hare]

Christopher,

    I have greped through the code tree and found that tk/win also appears
to have the same problems. Can you confirm this is the case.

Andy Hare
www.ahare.btinternet.co.uk

----- Original Message -----
From: "Christopher Faylor" <cgf@redhat.com>
To: "Andy Hare" <ahare@btinternet.com>
Cc: "Insight Mail list" <insight@sources.redhat.com>
Sent: Sunday, October 28, 2001 4:31 PM
Subject: Re: Problems building TCL for insight on cygwin & win2k


> [NOTE: Reply-To is set to insight@sources.redhat.com]
>
> On Sun, Oct 28, 2001 at 10:05:44AM -0000, Andy Hare wrote:
> >I would produce a patch to this effect but do not know how to create the
> >patch in the correct format or where to post it to so hopefully someone
with
> >better knowledge of the system than me can take this info and produce the
> >required.
>
> I appreciate the analysis and have checked in patches to rectify this
behavior.
>
> In the future, you can refer to the gdb/CONTRIBUTE document for
information on
> how to submit a patch.  It should provide information that will allow you
> to contribute your own patches in the future.
>
> cgf

Re: Problems building TCL for insight on cygwin & win2k

[Christopher Faylor]

On Sun, Oct 28, 2001 at 08:10:56PM -0000, Andy Hare wrote:
>I have greped through the code tree and found that tk/win also appears
>to have the same problems.  Can you confirm this is the case.

Um.  Have you done a checkout recently?

cgf