SWIG is a tool that simplifies the integration of C/C++ with common scripting languages

Marcelo Matus | 6 Dec 2005 20:43

Picon

Swig/Python thread support

Thread support was added to swig/python, based in the Joseph's proposal.

Please test your interfaces using

     swig -threads -python ....

and send us a mail if you find a problem, or have suggestions, comments, 
etc.

Ah, you will need python 2.3 or later, see CHANGES.current for  more 
details.

If you need thread support for older python versions, and know how to 
safely use
threads with them, please check the  pythreads.swg file and send  us an 
implementation
patch.

Marcelo

-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems?  Stop!  Download the new AJAX search engine that makes
searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click

Marcelo Matus | 7 Dec 2005 08:51

Picon

Re: [Swig-user] Swig/Python thread support

More info/docs:

  CHANGES.current
  Lib/python/pythreads.swg

the original proposal:

  
http://sourceforge.net/tracker/index.php?func=detail&aid=1364704&group_id=1645&atid=101645

but mainly preserved in spirit.

Marcelo

Robert Jones wrote:

> Where can I find Joseph's proposal and/or any documentation?
>
> Rob.
>
> On 06/12/05, *Marcelo Matus* <mmatus <at> acms.arizona.edu 
> <mailto:mmatus <at> acms.arizona.edu>> wrote:
>
>     Thread support was added to swig/python, based in the Joseph's
>     proposal.
>
>     Please test your interfaces using
>
>          swig -threads -python ....
>

(Continue reading)

Ames Andreas | 7 Dec 2005 12:49

RE: Swig/Python thread support

> -----Original Message-----
> From: swig-devel-admin <at> lists.sourceforge.net 
> [mailto:swig-devel-admin <at> lists.sourceforge.net] On Behalf Of 
> Marcelo Matus
> Sent: Tuesday, December 06, 2005 8:43 PM
> Cc: swig-user <at> lists.sourceforge.net; swig-devel <at> lists.sourceforge.net
> Subject: [Swig-devel] Swig/Python thread support
> 
> Thread support was added to swig/python, based in the 
> Joseph's proposal.
> 
> Please test your interfaces using
> 
>      swig -threads -python ....
> 
> and send us a mail if you find a problem, or have 
> suggestions, comments, 
> etc.

Great stuff, thanks.

I've got a small suggestion though.  Please replace

#      define SWIG_PYTHON_THREAD_BEGIN_BLOCK   SWIG_Python_Thread_Block _swig_thread_block
#      define SWIG_PYTHON_THREAD_END_BLOCK     _swig_thread_block.end()
#      define SWIG_PYTHON_THREAD_BEGIN_ALLOW   SWIG_Python_Thread_Allow _swig_thread_allow
#      define SWIG_PYTHON_THREAD_END_ALLOW     _swig_thread_allow.end()

with

(Continue reading)

Marcelo Matus | 7 Dec 2005 19:34

Picon

Re: Swig/Python thread support

Ames Andreas wrote:

>>-----Original Message-----
>>From: swig-devel-admin <at> lists.sourceforge.net 
>>[mailto:swig-devel-admin <at> lists.sourceforge.net] On Behalf Of 
>>Marcelo Matus
>>Sent: Tuesday, December 06, 2005 8:43 PM
>>Cc: swig-user <at> lists.sourceforge.net; swig-devel <at> lists.sourceforge.net
>>Subject: [Swig-devel] Swig/Python thread support
>>
>>Thread support was added to swig/python, based in the 
>>Joseph's proposal.
>>
>>Please test your interfaces using
>>
>>     swig -threads -python ....
>>
>>and send us a mail if you find a problem, or have 
>>suggestions, comments, 
>>etc.
>>    
>>
>
>Great stuff, thanks.
>
>I've got a small suggestion though.  Please replace
>
>#      define SWIG_PYTHON_THREAD_BEGIN_BLOCK   SWIG_Python_Thread_Block _swig_thread_block
>#      define SWIG_PYTHON_THREAD_END_BLOCK     _swig_thread_block.end()
>#      define SWIG_PYTHON_THREAD_BEGIN_ALLOW   SWIG_Python_Thread_Allow _swig_thread_allow

(Continue reading)

Marcelo Matus | 7 Dec 2005 20:21

Picon

Re: [Swig-user] RE: Swig/Python thread support

Ok, I remembered the main reason why we can't do this:

#      define SWIG_PYTHON_THREAD_BEGIN_BLOCK   { SWIG_Python_Thread_Block _swig_thread_block
#      define SWIG_PYTHON_THREAD_END_BLOCK     }
#      define SWIG_PYTHON_THREAD_BEGIN_ALLOW   { SWIG_Python_Thread_Allow _swig_thread_allow
#      define SWIG_PYTHON_THREAD_END_ALLOW     }

the code should work in plain C, and we have some gotos inside there, ie:

{
  SWIG_PYTHON_THREAD_BEGIN_BLOCK;
  ...
     if (error) goto fail;
  ...
  SWIG_PYTHON_THREAD_END_BLOCK;
  return resultobj;
fail:
  SWIG_PYTHON_THREAD_END_BLOCK;
  return NULL;
}

also, it seems we need to leave the macro

#define SWIG_PYTHON_INITIALIZE_THREADS  PyEval_InitThreads()

or the test suite fails in many places. If needed, the user can do

  c++ -DSWIG_PYTHON_INITIALIZE_THREADS="" ...

and that cancels it. Anyway, the docs says it is save to call

(Continue reading)

Louis Pecora | 8 Dec 2005 22:17

Picon

Picon

Solution to some SWIG Install problems on Mac OS X

If you are a newbie or non-guru (like me) and are are having some 
problems installing SWIG on Mac OS X (10.3.9, but version may not matter 
for the problems below), you should be aware of some problems I had and 
their solutions.  Many thanks to several people on the email lists for 
Pythonmac-sig (pythonmac-sig <at> python.org) and SWIG 
(swig-user <at> lists.sourceforge.net) for many valuable hints, clues and 
pointers.

The installation of SWIG on the Mac is the same three-step process as 
for Unix (% is the prompt):

% ./configure

% make

% make install

I had the follow problems with each step and I have the solutions.

(1) running ./configure resulted in a file not found.  When the script 
listed the file only a partial path name appeared.
   Solution:  directories (folders) and files with spaces and special 
Mac symbols (e.g. option-whatever) can cause problems.  I had some 
spaces in my folder names that held the SWIG files.  When I removed 
those and made more Unix-like names, ./configure worked (after one more 
fix below).

(2) running  'make'  caused a parse error in CParse when compiling 
cscanner.c.
    Solution:  when I looked back at ./configure I see that g++ was the

(Continue reading)

Ames Andreas | 9 Dec 2005 14:55

RE: Swig/Python thread support

> -----Original Message-----
> From: Marcelo Matus [mailto:mmatus <at> acms.arizona.edu] 
> Sent: Wednesday, December 07, 2005 8:22 PM
> To: Ames Andreas
> Cc: swig-user <at> lists.sourceforge.net; swig-devel <at> lists.sourceforge.net
> Subject: Re: [Swig-user] RE: [Swig-devel] Swig/Python thread support
> 
> Ok, I remembered the main reason why we can't do this:
> 
> #      define SWIG_PYTHON_THREAD_BEGIN_BLOCK   { 
> SWIG_Python_Thread_Block _swig_thread_block
> #      define SWIG_PYTHON_THREAD_END_BLOCK     }
> #      define SWIG_PYTHON_THREAD_BEGIN_ALLOW   { 
> SWIG_Python_Thread_Allow _swig_thread_allow
> #      define SWIG_PYTHON_THREAD_END_ALLOW     }
> 
> the code should work in plain C, and we have some gotos 
> inside there, ie:
> 
> {
>   SWIG_PYTHON_THREAD_BEGIN_BLOCK;
>   ...
>      if (error) goto fail;
>   ...
>   SWIG_PYTHON_THREAD_END_BLOCK;
>   return resultobj;
> fail:
>   SWIG_PYTHON_THREAD_END_BLOCK;
>   return NULL;
> }

(Continue reading)

Marcelo Matus | 9 Dec 2005 18:54

Picon

Re: Swig/Python thread support

Ames Andreas wrote:

>>-----Original Message-----
>>From: Marcelo Matus [mailto:mmatus <at> acms.arizona.edu] 
>>Sent: Wednesday, December 07, 2005 8:22 PM
>>To: Ames Andreas
>>Cc: swig-user <at> lists.sourceforge.net; swig-devel <at> lists.sourceforge.net
>>Subject: Re: [Swig-user] RE: [Swig-devel] Swig/Python thread support
>>
>>Ok, I remembered the main reason why we can't do this:
>>
>>#      define SWIG_PYTHON_THREAD_BEGIN_BLOCK   { 
>>SWIG_Python_Thread_Block _swig_thread_block
>>#      define SWIG_PYTHON_THREAD_END_BLOCK     }
>>#      define SWIG_PYTHON_THREAD_BEGIN_ALLOW   { 
>>SWIG_Python_Thread_Allow _swig_thread_allow
>>#      define SWIG_PYTHON_THREAD_END_ALLOW     }
>>
>>the code should work in plain C, and we have some gotos 
>>inside there, ie:
>>
>>{
>>  SWIG_PYTHON_THREAD_BEGIN_BLOCK;
>>  ...
>>     if (error) goto fail;
>>  ...
>>  SWIG_PYTHON_THREAD_END_BLOCK;
>>  return resultobj;
>>fail:
>>  SWIG_PYTHON_THREAD_END_BLOCK;

(Continue reading)

Marcelo Matus | 9 Dec 2005 19:58

Picon

Re: Swig/Python thread support

Oh, well, still we have the dispatch case:
{
  SWIG_PYTHON_THREAD_BEGIN_BLOCK;
  argc = PyObject_Length(args);
  for (ii = 0; (ii < argc) && (ii < 2); ii++) {
    argv[ii] = PyTuple_GetItem(args,ii);
  }
  if (argc == 1) {
    int _v;
    _v = (SWIG_AsVal_int(argv[0], 0) == SWIG_OK);
    if (_v) {
      SWIG_PYTHON_THREAD_END_BLOCK;
      return _wrap_overtparams1__SWIG_2(self,args);
    }
  }
  ....
  SWIG_PYTHON_THREAD_END_BLOCK;
  SWIG_SetErrorMsg(PyExc_NotImplementedError,"No matching function for 
overloaded 'overtparams1'");
  return NULL;
}

the problem is that  eliminating SWIG_PYTHON_THREAD_END_BLOCK in C++
will produce different results in C and C++, ie, in the first case the 
blocks
is ended before dispatching the method, in the second, the block will be
closed after the call.  In theory we can change the code to produce
a more friendly code for threads, but the dispatch generator is
common in C++ and is not easy to change since
it must work for all the languages.

(Continue reading)

Ames Andreas | 9 Dec 2005 20:14

RE: Swig/Python thread support

> -----Original Message-----
> From: Marcelo Matus [mailto:mmatus <at> acms.arizona.edu] 
> Sent: Friday, December 09, 2005 6:54 PM
> To: Ames Andreas
> Cc: swig-user <at> lists.sourceforge.net; swig-devel <at> lists.sourceforge.net
> Subject: Re: [Swig-devel] Swig/Python thread support
> 
> but we call PyEval_InitThreads only when the module is 
> loaded/imported, 
> is that still dangerous?
> 
> I mean, if you can't never call it safely, nobody should call 
> it ever, 
> right?

Ok, it seems I've found a good reference for the problem at hand, please take a look at
http://www.python.org/peps/pep-0311.html.  According to this, what you are doing with
PyEval_InitThreads seems to be the right thing.  Sorry for the fuzz.

Calling it in the init code should ensure that it is either called from python or from a thread holding the GIL
(anything else would be a bug in the calling code).  If it's called from a thread holding the GIL, it is
essentially a noop (see Python/ceval.c); if it's called from python, it is either called from a thread
that holds the GIL, so see above, or there is no GIL yet, which actually means there is no other python thread
(because thread creation calls PyEval_InitThreads).  In this case it is reasonable to confer the 'main
thread' badge on the calling thread.  Once again, sorry for wasting your time.

cheers,

aa

(Continue reading)

Group

gmane.comp.programming.swig.

Project Web Page

SWIG is a tool that simplifies the integration of C/C++ with common scripting languages

Search Archive

Page

1 | 2 | 3

Articles in period: 23

December 2005

Language

Change language

Options

Current view: Threads only / Showing only 20 lines / Not hiding cited text.
Change to All messages, whole messages, or hide cited text.

Post a message
NNTP Newsgroup
Classic Gmane web interface
XML

XML

RSS Feed
List Information

About Gmane

Recent entries

Fwd: Re: std::vector of class with no default constructor (23 May 2013)
Can SWIG Python use Python 3 common ABI? (23 May 2013)
std::vector of class with no default constructor (22 May 2013)
Are there any SWIG users within Brisbane? (16 May 2013)
Resolve struct property for BLOB buffer -> Java (16 May 2013)
Problem wrapping a struct (14 May 2013)
Problem wrapping a struct - FIXED (14 May 2013)
Eigen3 + std::vector + SWIG typemap -> Python (13 May 2013)
wrapper for std::tr1::array (13 May 2013)
Question: adding python instance variable to C++ class? (13 May 2013)
Berkelium (13 May 2013)
headers.h: not such file or directory (12 May 2013)
Special Qt mode for SWIG (12 May 2013)
Needed support to handel the Nested classes and Threading (11 May 2013)
PHP generated code in c++ for functions outside class (10 May 2013)
PHP generated code in c++ for functions outside class (10 May 2013)
Migrating from Swig 1.3.33 to 2.0.9 (7 May 2013)
C/C++ to Java : Use a Java method from C by overloading a C++ object (30 Apr 2013)
c++ -> python: Double delete in _wrap_delete when using a typemap with (29 Apr 2013)
Python overloaded method wrapping does not check types properly (22 Apr 2013)
tcl export std vector (22 Apr 2013)

Archives

35	May 2013
35	April 2013
74	March 2013
57	February 2013
100	January 2013
20	December 2012
76	November 2012
104	October 2012
73	September 2012
64	August 2012
85	July 2012
101	June 2012
169	May 2012
90	April 2012
119	March 2012
68	February 2012
74	January 2012
68	December 2011
160	November 2011
141	October 2011
101	September 2011
122	August 2011
158	July 2011
200	June 2011
88	May 2011
113	April 2011
150	March 2011
110	February 2011
159	January 2011
91	December 2010
59	November 2010
87	October 2010
64	September 2010
98	August 2010
85	July 2010
96	June 2010
144	May 2010
151	April 2010
128	March 2010
77	February 2010
105	January 2010
64	December 2009
128	November 2009
107	October 2009
150	September 2009
195	August 2009
158	July 2009
151	June 2009
97	May 2009
164	April 2009
208	March 2009
165	February 2009
151	January 2009
106	December 2008
100	November 2008
126	October 2008
129	September 2008
78	August 2008
140	July 2008
147	June 2008
156	May 2008
263	April 2008
186	March 2008
138	February 2008
335	January 2008
121	December 2007
163	November 2007
189	October 2007
182	September 2007
187	August 2007
222	July 2007
204	June 2007
268	May 2007
244	April 2007
172	March 2007
126	February 2007
210	January 2007
129	December 2006
309	November 2006
282	October 2006
168	September 2006
157	August 2006
174	July 2006
193	June 2006
160	May 2006
232	April 2006
211	March 2006
39	February 2006
29	January 2006
23	December 2005
3	November 2005
201	October 2005
215	September 2005
158	August 2005
106	July 2005
177	June 2005
136	May 2005
118	April 2005
228	March 2005
165	February 2005
143	January 2005
265	December 2004
263	November 2004
177	October 2004
197	September 2004
172	August 2004
167	July 2004
153	June 2004
206	May 2004
178	April 2004
221	March 2004
161	February 2004
180	January 2004
129	December 2003
175	November 2003
134	October 2003
135	September 2003
265	August 2003
190	July 2003
200	June 2003
252	May 2003
206	April 2003
259	March 2003
193	February 2003
187	January 2003
143	December 2002
266	November 2002
176	October 2002
251	September 2002
154	August 2002
1	January 2002

Design

Zawodny | Right Menu | Left Menu

Your Own Design

Paste the URL of your CSS below. Download CSS template