Re: compiling python under cygwin

Re: compiling python under cygwin

[Jason Tishler]

David,

On Wed, Oct 04, 2000 at 11:42:16AM -0400, Robinow, David wrote:
>    > I will try to get the Python source patched as above.
>    Please don't.

I won't -- less work for me.

> Cygwin has already been patched to solve this problem.

Thanks for the edification.

Jason

-- 
Jason Tishler
Director, Software Engineering       Phone: +1 (732) 264-8770 x235
Dot Hill Systems Corporation         Fax:   +1 (732) 264-8798
82 Bethany Road, Suite 7             Email: Jason.Tishler@dothill.com
Hazlet, NJ 07730 USA                 WWW:   http://www.dothill.com

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

Re: compiling python under cygwin

[Jason Tishler]

David,

On Wed, Oct 04, 2000 at 01:17:36PM -0400, Jason Tishler wrote:
> On Wed, Oct 04, 2000 at 11:42:16AM -0400, Robinow, David wrote:
> >    > I will try to get the Python source patched as above.
> >    Please don't.
> 
> I won't -- less work for me.

I didn't but it happened anyway:

    http://sourceforge.net/patch/download.php?id=101802

Jason

-- 
Jason Tishler
Director, Software Engineering       Phone: +1 (732) 264-8770 x235
Dot Hill Systems Corporation         Fax:   +1 (732) 264-8798
82 Bethany Road, Suite 7             Email: Jason.Tishler@dothill.com
Hazlet, NJ 07730 USA                 WWW:   http://www.dothill.com

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

Re: compiling python under cygwin

[Jason Tishler]

On Wed, Oct 04, 2000 at 11:42:16AM -0400, Robinow, David wrote:
>    The following was added as netinet/tcp.h on 9/5/2000:
>    
>    #ifndef _NETINET_TCP_H
>    #define _NETINET_TCP_H
>    
>    #include <winsock.h>
>    
>    #endif

Unfortunately after further investigation, I have to report that the
above change to Cygwin causes Python 2.0b2's Modules/socketmodule.c to
fail compilation.

BTW, this change also causes problems with PostgreSQL 7.0.2's
src/backend/libpq/pqcomm.c and src/interfaces/libpq/fe-connect.c too.

I guess I prefer that Cygwin has an empty netinet/tcp.h.  Is there any
chance of that happening?

Thanks,
Jason

-- 
Jason Tishler
Director, Software Engineering       Phone: +1 (732) 264-8770 x235
Dot Hill Systems Corporation         Fax:   +1 (732) 264-8798
82 Bethany Road, Suite 7             Email: Jason.Tishler@dothill.com
Hazlet, NJ 07730 USA                 WWW:   http://www.dothill.com

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

Re: compiling python under cygwin

[Chris Faylor]

On Wed, Oct 04, 2000 at 08:51:00PM -0400, Jason Tishler wrote:
>On Wed, Oct 04, 2000 at 11:42:16AM -0400, Robinow, David wrote:
>>    The following was added as netinet/tcp.h on 9/5/2000:
>>    
>>    #ifndef _NETINET_TCP_H
>>    #define _NETINET_TCP_H
>>    
>>    #include <winsock.h>
>>    
>>    #endif
>
>Unfortunately after further investigation, I have to report that the
>above change to Cygwin causes Python 2.0b2's Modules/socketmodule.c to
>fail compilation.
>
>BTW, this change also causes problems with PostgreSQL 7.0.2's
>src/backend/libpq/pqcomm.c and src/interfaces/libpq/fe-connect.c too.
>
>I guess I prefer that Cygwin has an empty netinet/tcp.h.  Is there any
>chance of that happening?

How about a few more details other than "fail compilation" before any
decisions are made?

cgf

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

Re: compiling python under cygwin

[Jason Tishler]

Chris,

On Wed, Oct 04, 2000 at 08:56:40PM -0400, Chris Faylor wrote:
> How about a few more details other than "fail compilation" before any
> decisions are made?

See attached for the compilation errors from compiling Python 2.0b2's
Modules/socketmodule.c after replacing an empty /usr/include/netinet/tcp.h 
with the one from the latest snapshot.  The PostgreSQL compilation
errors are similar.

Thanks,
Jason

-- 
Jason Tishler
Director, Software Engineering       Phone: +1 (732) 264-8770 x235
Dot Hill Systems Corporation         Fax:   +1 (732) 264-8798
82 Bethany Road, Suite 7             Email: Jason.Tishler@dothill.com
Hazlet, NJ 07730 USA                 WWW:   http://www.dothill.com

Re: compiling python under cygwin

[Chris Faylor]

On Wed, Oct 04, 2000 at 11:05:24PM -0400, Jason Tishler wrote:
>On Wed, Oct 04, 2000 at 08:56:40PM -0400, Chris Faylor wrote:
>>How about a few more details other than "fail compilation" before any
>>decisions are made?
>
>See attached for the compilation errors from compiling Python 2.0b2's
>Modules/socketmodule.c after replacing an empty
>/usr/include/netinet/tcp.h with the one from the latest snapshot.  The
>PostgreSQL compilation errors are similar.

Ok.  You've convinced me.  tcp.h is conflicting with sys/socket.h among other
things.  I'll remove the winsock.h reference.

cgf

>gcc  -g -O2 -I./../Include -I.. -DHAVE_CONFIG_H -c ./socketmodule.c
>In file included from /usr/lib/gcc-lib/i686-pc-cygwin/2.95.2/../../../../include/netinet/tcp.h:4,
>                 from ./socketmodule.c:126:
>/usr/lib/gcc-lib/i686-pc-cygwin/2.95.2/../../../../include/winsock.h:104: warning: `_IO' redefined
>/usr/lib/gcc-lib/i686-pc-cygwin/2.95.2/../../../../include/asm/socket.h:11: warning: this is the location of the previous definition
>/usr/lib/gcc-lib/i686-pc-cygwin/2.95.2/../../../../include/winsock.h:105: warning: `_IOR' redefined
>/usr/lib/gcc-lib/i686-pc-cygwin/2.95.2/../../../../include/asm/socket.h:12: warning: this is the location of the previous definition

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

RE: compiling python under cygwin

[Norman Vine]

[-- Attachment #1: Type: text/plain, Size: 1318 bytes --]

Chris Faylor writes:
>
>On Wed, Oct 04, 2000 at 08:51:00PM -0400, Jason Tishler wrote:
>>On Wed, Oct 04, 2000 at 11:42:16AM -0400, Robinow, David wrote:
>>>    The following was added as netinet/tcp.h on 9/5/2000:
>>>    
>>>    #ifndef _NETINET_TCP_H
>>>    #define _NETINET_TCP_H
>>>    
>>>    #include <winsock.h>
>>>    
>>>    #endif
>>
>>Unfortunately after further investigation, I have to report that the
>>above change to Cygwin causes Python 2.0b2's Modules/socketmodule.c to
>>fail compilation.
>>
>>BTW, this change also causes problems with PostgreSQL 7.0.2's
>>src/backend/libpq/pqcomm.c and src/interfaces/libpq/fe-connect.c too.
>>
>>I guess I prefer that Cygwin has an empty netinet/tcp.h.  Is there any
>>chance of that happening?
>
>How about a few more details other than "fail compilation" before any
>decisions are made?

It is my understanding that the 'reason' for Cygwin is so one can
write Posix code that will run on a Windows platform.  By including
<winsock.h> any code written to use the Posix headers 
 <netdb.h>  <sys/socket.h>  <netinet/in.h>
will have numerous namespace collisions.

Perhaps it would be advantageous performance wise to use
the native Windows sockets rather then the Cygwin Posix sockets
but that IMHO rather defeats the purpose of Cygwin 

Cheers

Norman Vine






[-- Attachment #2: err.log.gz --]
[-- Type: application/x-gzip, Size: 1293 bytes --]

Re: compiling python under cygwin

[Chris Faylor]

On Wed, Oct 04, 2000 at 11:32:53PM -0400, Norman Vine wrote:
>>How about a few more details other than "fail compilation" before any
>>decisions are made?
>
>It is my understanding that the 'reason' for Cygwin is so one can
>write Posix code that will run on a Windows platform.  By including
><winsock.h> any code written to use the Posix headers 
> <netdb.h>  <sys/socket.h>  <netinet/in.h>
>will have numerous namespace collisions.
>
>Perhaps it would be advantageous performance wise to use
>the native Windows sockets rather then the Cygwin Posix sockets
>but that IMHO rather defeats the purpose of Cygwin 

You realize that you posted this insightful observation about an hour
after Jason helpfully posted the details that I asked for and about
fifty minutes after I sent this:

>Ok.  You've convinced me.  tcp.h is conflicting with sys/socket.h among
>other things.  I'll remove the winsock.h reference.

Right?  I've waited a little while to see if that realization has sunk
in but apparently it hasn't.

In case it isn't obvious, I'd like people to think about posting
*details* when they report problems.  Posting "fail to compile" (no
offense, Jason) is not helpful to anyone else reading the list.

I was pretty sure when I saw the email how I'd screwed up by including
winsock.h but a terse exchange of:

"It failed to compile.  Please make tcp.h empty."

"Ok, I will."

is not useful reading for anyone.

And, of course, if I did make the change, I would only be ass*u*ming
that I understood the reason for removing winsock.h.

Now, however, we have some nice details in the archives for people to
ignore at a later date when someone wonders why tcp.h is empty.

I have this possibly fruitless hope that if I keep asking for details,
eventually people will provide them.  I hate to be making this point in
any way around email from Jason, since he is usually very good at
providing said details, and has been a valuable Cygwin contributor, but
since you seemed to think that I needed to be educated about the reason
for Cygwin, I've returned the favor to educate you in the proper way to
report problems.

I'm sure you appreciated the education as much as I did.

cgf

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

RE: compiling python under cygwin

[Norman Vine]

Chris Faylor writes:
>
>You realize that you posted this insightful observation about an hour
>after Jason helpfully posted the details that I asked for and about
>fifty minutes after I sent this:

My apologies for the postings crossing in the ether
My head is bent in shame

I also apologize for taking 20 minutes to create some data
during which time others did the same,  and my subsequent 
error of posting of the requested data before rechecking my mail.

A mistake I will not make again. 
< when on a dial up not rechecking mail before posting reply >

FYI
Jason and I have been wrestling with the Python source,
the Cygwin DLL creation methods and the Python make system
to have a full featured dynamically linked Python for Cygwin users.  
With the exception of threading support, I believe that we will have
it SOON. :-)

All the best and THANK YOU 
for your fine stewardship of Cygwin.

It has come along ways since I started using B-18

Norman



FWIW below is my message to Jason that spawned this.

===============
Subject: FW: Python 2.0b2 Regression Test Results

Importance: High



-----Original Message-----
From: Norman Vine [ mailto:nhv@cape.com ] 
Sent: Wednesday, October 04, 2000 5:39 PM
To: 'Jason Tishler'
Cc: 'drobinow@yahoo.com'
Subject: RE: Python 2.0b2 Regression Test Results
Importance: High


Jason Tishler writes:

> Norman Vine wrote:
>
>> Be sure to mention the missing tcp.h thing as well as 
>> the two fixes I submitted to the Python Bug Collector.
>
>The missing tcp.h has already been corrected:
>
>On Wed, Oct 04, 2000 at 11:42:16AM -0400, Robinow, David wrote:
>>   The following was added as netinet/tcp.h on 9/5/2000
>>   See a recent snapshot.
>>
>>   #ifndef _NETINET_TCP_H
>>   #define _NETINET_TCP_H
>>
>>   #include <winsock.h>
>>
>>   #endif
>> ...

ARGGGGHHHH !!!!!!!

I complained when they wanted to do this
I guess not loudly enough   thread @
http://sources.redhat.com/ml/cygwin/2000-09/msg00010.html


THIS BREAKS socketmodule.c

Try it for youself 

$PY_SRC / Modules / socketmodule.c  ~line 125
#if !(defined(__BEOS__)) // || defined(__CYGWIN__))
//#include <netinet/tcp.h>
#include <winsock.h>
#endif

Grrr... !@$#

Maybe if this breaks for you you could post a 
request to the cygwin list.  

I do not seem to have much luck there :-(

Cheers

Norman



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

RE: compiling python under cygwin

[Laurence F. Wood]

I have had problems porting a large system because of the winsock.h include.
For one thing, I was having a problem with WIN32/_WIN32 being suddenly
defined (by something that winsock.h includes) in conditional code.  It took
me a while to track down what was going on.

Laurence F. Wood
Chief Science Officer
Sunyata Systems Corporation
Tel: 732-701-9746
Fax: 732-701-9748
www.sunyatasystems.com

"Pioneers in High Fidelity Computational Biochemistry
for Lead Optimization and Compound Synthesis"



-----Original Message-----
From: cygwin-owner@sources.redhat.com
[ mailto:cygwin-owner@sources.redhat.com]On Behalf Of Chris Faylor
Sent: Thursday, October 05, 2000 12:25 AM
To: cygwin@sources.redhat.com
Subject: Re: compiling python under cygwin


On Wed, Oct 04, 2000 at 11:32:53PM -0400, Norman Vine wrote:
>>How about a few more details other than "fail compilation" before any
>>decisions are made?
>
>It is my understanding that the 'reason' for Cygwin is so one can
>write Posix code that will run on a Windows platform.  By including
><winsock.h> any code written to use the Posix headers
> <netdb.h>  <sys/socket.h>  <netinet/in.h>
>will have numerous namespace collisions.
>
>Perhaps it would be advantageous performance wise to use
>the native Windows sockets rather then the Cygwin Posix sockets
>but that IMHO rather defeats the purpose of Cygwin

You realize that you posted this insightful observation about an hour
after Jason helpfully posted the details that I asked for and about
fifty minutes after I sent this:

>Ok.  You've convinced me.  tcp.h is conflicting with sys/socket.h among
>other things.  I'll remove the winsock.h reference.

Right?  I've waited a little while to see if that realization has sunk
in but apparently it hasn't.

In case it isn't obvious, I'd like people to think about posting
*details* when they report problems.  Posting "fail to compile" (no
offense, Jason) is not helpful to anyone else reading the list.

I was pretty sure when I saw the email how I'd screwed up by including
winsock.h but a terse exchange of:

"It failed to compile.  Please make tcp.h empty."

"Ok, I will."

is not useful reading for anyone.

And, of course, if I did make the change, I would only be ass*u*ming
that I understood the reason for removing winsock.h.

Now, however, we have some nice details in the archives for people to
ignore at a later date when someone wonders why tcp.h is empty.

I have this possibly fruitless hope that if I keep asking for details,
eventually people will provide them.  I hate to be making this point in
any way around email from Jason, since he is usually very good at
providing said details, and has been a valuable Cygwin contributor, but
since you seemed to think that I needed to be educated about the reason
for Cygwin, I've returned the favor to educate you in the proper way to
report problems.

I'm sure you appreciated the education as much as I did.

cgf

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



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

Re: compiling python under cygwin

[Corinna Vinschen]

"Laurence F. Wood" wrote:
> 
> I have had problems porting a large system because of the winsock.h include.
> For one thing, I was having a problem with WIN32/_WIN32 being suddenly
> defined (by something that winsock.h includes) in conditional code.  It took
> me a while to track down what was going on.

Yup, the WIN32 define sometimes results in, erm, *interesting* problems
while porting.

However, the winsock include in tcp.h is fixed in the winsup cvs.

Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Developer                        mailto:cygwin@sources.redhat.com
Red Hat, Inc.
mailto:vinschen@cygnus.com

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

Re: compiling python under cygwin

[Earnie Boyd]

--- Jason Tishler <Jason.Tishler@dothill.com> wrote:
>
>  I still need to track
> down why the failures are occurring especially test_poll.
> 

This would be nice of you.  Based on the ChangeLog the poll function was
implemented 2000.07.04 in case you need to know.

Cheers,

=====
Earnie Boyd
mailto:earnie_boyd@yahoo.com

---         < http://earniesystems.safeshopper.com >         ---
--- Cygwin: POSIX on Windows < http://gw32.freeyellow.com/ > ---
---   Minimalist GNU for Windows < http://www.mingw.org/ >   ---

__________________________________________________
Do You Yahoo!?
Yahoo! Photos - 35mm Quality Prints, Now Get 15 Free!
http://photos.yahoo.com/

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

compiling python under cygwin

[Lockwood Nick L CONT PHCA]

Hello all:
	I want to compile the python interpreter on Windows NT 4, using the
gcc compiler under cygwin.  I have installed cygwin and have obtained the
source code to python.  The python interpreter, like most Windows software
is currently available compiled using Microsoft Visual C++.  Makefiles for
various platforms are available, but not for gcc.  I am experienced in C,
C++, Unix, make, and various other things, since I have been a programmer
for a long time.  I work on linux at home, but I am forced to use Windows at
work when on the PC.  Since I have converted many programs from one platform
to another, I am aware there may be some difficulties in my proposed
project, but I would rather face them than obtain Visual C++.  I consider it
desirable for python to be compiled on a free compiler for the Windows
platform.
	Can anyone give me any hints to help me get going?  For instance,
which existing makefile would be closest to working under cygwin?  Or am I
stuck with figuring this out for myself or starting from scratch?  Any help
or suggestions would be appreciated.
	
	Nick Lockwood

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

Re: compiling python under cygwin

[Jason Tishler]

Nick,

On Fri, Sep 29, 2000 at 04:34:18PM -0700, Lockwood Nick L CONT PHCA wrote:
> 	Can anyone give me any hints to help me get going?  For instance,
> which existing makefile would be closest to working under cygwin?  Or am I
> stuck with figuring this out for myself or starting from scratch?  Any help
> or suggestions would be appreciated.

Python (at least 1.5.2 and 2.0b2) builds OOTB on Cygwin.  I just
built Python 2.0b2 under Cygwin 1.1.4 using the following:

    $ tar -xzf Python-2.0b2.tar.gz
    $ cd Python-2.0b2
    $ configure --with-threads=no --with-suffix=.exe --with-libm=
    $ make
    $ make install

Jason

-- 
Jason Tishler
Director, Software Engineering       Phone: +1 (732) 264-8770 x235
Dot Hill Systems Corporation         Fax:   +1 (732) 264-8798
82 Bethany Road, Suite 7             Email: Jason.Tishler@dothill.com
Hazlet, NJ 07730 USA                 WWW:   http://www.dothill.com

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

Re: compiling python under cygwin

[Jason Tishler]

Erik,

On Tue, Oct 03, 2000 at 03:23:59PM -0600, Erik Nolte wrote:
> I must be missing something...  I have cygwin 1.1.4, all the latest cygwin
> packages, and Python 2.0b2 source.  Configure runs as expected, but make
> dies in Modules/socketmodule.c because it can't find netinet/tcp.h.  I
> searched the cygwin packages and none of them contain a tcp.h.  I conclude
> that tcp.h shouldn't be part of the cygwin release and that's ok because the
> #include is wrapped with an #ifndef MS_WINDOWS.  Unfortunately configure
> doesn't set an MS_WINDOWS symbol and neither do any of the include files
> (python or /usr/include).  If I #define MS_WINDOWS I get a slew of winsock.h
> conflict-type errors.  I was finally able to build python by commenting out
> '#include <netinet/tcp.h>' in socketmodule.c.  I'm not sure the resulting
> executable is fully functional.

Due to building PostgreSQL 7.0.2, my Cygwin installation already had an
(empty) /usr/include/netinet/tcp.h (due to following their Cygwin README).
Hence, I was unaware of the above problem.

Your workaround is fine and your executable should be as "fully
functional" as mine.  If you run the regression tests, you should find
that it fails test_fcntl, test_poll (which should be excluded due to
hanging), test_pty, test_strftime, and test_time.  I still need to track
down why the failures are occurring especially test_poll.

Alternatively, you can change the relevant lines in Modules/socketmodule.c,
from:

    #ifndef __BEOS__
    #include <netinet/tcp.h>
    #endif

to:

    #if !define(__BEOS__) && !define(__CYGWIN__)
    #include <netinet/tcp.h>
    #endif

I will try to get the Python source patched as above.

Jason

-- 
Jason Tishler
Director, Software Engineering       Phone: +1 (732) 264-8770 x235
Dot Hill Systems Corporation         Fax:   +1 (732) 264-8798
82 Bethany Road, Suite 7             Email: Jason.Tishler@dothill.com
Hazlet, NJ 07730 USA                 WWW:   http://www.dothill.com

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