Re: [Boost-build] Patches to python toolset

From: Bojan Resnik (resnikb_at_[hidden])
Date: 2006-04-03 04:06:30

> There's one issue -- the Python module can be now initialized several times,
> with different versions for different compilers. For example:
>
>   using python : 2.3 : /some/root : : :
>           <toolset>gcc <toolset-gcc:version>3.3 ;
>
>   using python : 2.3 : /some/other/root : : :
>           <toolset>gcc <toolset-gcc:version>4.0 ;
>
> But your code will return just one version. What do you think be the right
> solution for you case?

  As I understand it, the conditionals are used to select correct
headers and libraries, and the interpreter is not used. The primary
purpose of having the path to the interpreter is to enable user
actions to execute python code without having to find the interpreter
themselves. I think the only relevant part is the version of the
interpreter. Perhaps the rule that returns the interpreter should have
an optional argument to specify the version of the interpreter to be
returned. If there is no interpreter with that version, the rule
should return the interpreter with the next greater version?

> >   Please find attached the diffs for python.jam and python-config.jam.
>
> Except for the issue above, the patch looks good. For future patches, though,
> it's better to not use UPPERCASE_VARIABLES for your own variables, this way
> they can be confused with environment variables. Instead, prepend "." to
> names of module-global variables as documented in hacking.txt

  Thanks, I'll change that.

--
Bojan Resnik