Phillip J. Eby wrote:
> At 11:44 PM 4/2/2007 -0400, Mike Taylor wrote:
>> The reason we set rt.py to use the system python is so that the test
>> running tool is not using the code that it is testing (and to not have
>> to have two copies of Chandler Python installed.)
> 
> I don't understand.  Why would two copies of Python be necessary to run
> Python in a subprocess?

I think that is a red herring.

Some reasons I have for preferring ./tools/rt.py to run system python:

* tools should not depend on Chandler unless they have to
* In case Chandler is so borked that python won't run properly I want rt
still to function (tbox will rely on this) - think about case where
chandler python hangs on startup
* I want rt.py to run fast, using python release bits - which would not
work if you did make install DEBUG=1
* I don't think you can put Chandler python on the bang line in
$CHANDLERBIN/(debug|release)/RunChandler(.bat) format

I think I had some other reasons when we started on rt work, but I can
not remember all of my reasons.

Anyway, the issue here is that we need to build Chandler python with
ctypes. I'll file a bug if there isn't one already.

-- 
  Heikki Toivonen


Attachment: signature.asc
Description: OpenPGP digital signature

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Open Source Applications Foundation "chandler-dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/chandler-dev

Reply via email to