Re: [pygsl-discuss] New PyGSL release ?

[Pierre S. <p.s...@gs...>]

Sébastien Maret wrote:
>>> Maybe it is the time for a new release ?
>>
>>
>> Agreed. Expect something until the end of the week.
> 
> 
> We may want to wait a little bit indeed...

> I get a number of weird messages when I run the test with  Python-2.4.1 
> on MacOSX 10.4.1. I occurs both when compiled with  gcc-3.3. gcc-4.0 and 
> with Numeric and numarray:
> 
> END   PyGSL_copy_pyarray_to_gslvector In File src/init/ block_helpers.c 
> at line 303
> BEGIN PyGSL_copy_gslvector_to_pyarray In File src/init/ block_helpers.c 
> at line 232
> 
:-)

These messages come from the debug info structure. A bit little
documented, so lets puts light on it.

If you look into the setup file, you will spot
<snip>
macros = macros + [('DEBUG', 1)]
</snip>

This will, as you will know, define -DDEBUG=1 as switch to the compiler,
which will in turn define DEBUG to one for the preprocessor.

PyGSL comes with the following macros, which are shallow warppers around
  fprintf (see also Include/pygsl/intern.h Lines 120-127,
Include/pygsl/utils.h)

FUNC_MESS_BEGIN()
FUNC_MESS_END()
DEBUG_MESS(level, mess, ...)

All these use PyGSL_DEBUG_LEVEL() to see if to do something. This goes
back either to a value defined at compile time, if set to bigger than 0.
If DEBUG is not defined at compile time, all this information will be
left out.

DEBUG=1 is special. In this case, PyGSL_DEBUG_LEVEL() is a static
defined variable. Depending on the value of this variable (which is
unique to each module implemented in C) the above mentioned macros will
print out the information.

The memory location of this variable is registered by init_pygsl() (see
also Include/pygsl/intern.h Lines 135--) via PyGSL_init_debug()
to PyGSL_register_debug_flag which is implemented in
src/init/initmodule.c. This keeps a Python lists of CObjects which hold
pointers to the memory locations of the variables pygsl_debug_level.

If one calls pygsl.init.set_debug_level(level), it will set the values
of all these variables and one can see the aforementioned prints. This
might be an nuisance if it slips out to the user, but is very usefull if
wired errors occur during development.

I now set this level to 0 and disabled the DEBUG define, so that should
be cleared out now.

As you are participating in Fink, I thought that could be of some value
to you.

Best regards
	Pierre