Source

extradoc / sprintinfo / ddorf2006 / planning.txt

Sprint Planning DDorf  
=========================================

Before the sprint, there were two somewhat-similar, somewhat-conflicting
goals: finish off wp7 functionality (stackless stuff, and more...) and
getting the 0.9 release out of the door.  We decided that the priority
of this sprint, or at least the next two days should be on wp7 and
particularly stackless related features. 

time plan: monday afternoon: breakday 
           tuesday afternoon: EU technical project planning 


Some things that don't really have issues yet:

 * the ext compiler
   (armin, carl, michael, arre, holger) 
   * design user level usage + implement it   
   * Document using ext-compiler getting-started
   * readline support for translated PyPy 

 * documentation

 * running the compliance tests and getting results on the web again

   (DONE) compliance is at 89% **Holger and Anders**

 * ootypesystem stuff

   ootype is mostly complete (r_dict missing)
   **Nik and Antonio are working on ootype completion ** 
   ** Maciek will join  later **

Looking at the 0.9 issues, critical first:

* 194	Advance rctypes approach to become usable

  Basically done (ask Armin). 
  open sub-issue: what about releasing the GIL? 

* 197	implement __del__ support for framework gcs

  Needs more testing (particularly with the stacklessgc), but we think this
  works.

  **Arre, cfbolz to do more testing**

* ???   expose coroutine cloning to app level.
 
  (Armin) 
  * settle on APIs on RPython level 
  * experiment a bit with exposing it to applevel
  * document for 0.9 release 
  
  Shouldn't be too hard, needs to be done though.

* 198	implement tasklet pickling

  (christian) 

  - (in-progress) finish needed object type pickling/unpickling
  - (DONE) finish and integrate explict resume points support
  - (christian, samuele) use the latter to implement restarting after unpickling
    (likely hack! hack! hack! until it works)

  Tasklet cloning has become completely different, that kind of works now
  (needs more testing).

  (armin) write up documentation on how to use tasklet cloning 


* 199	finish app-level stackless support
  (stephan is working on it) 
  This should be rated 'critical'!  Not sure what the status is.

  Greenlets could be done quickly.

Now the bugs:

* 137	Control-C runtime support

  it is quite hard. 

  (arre, cfbolz) 

* 181	time module is incomplete

  Will be done by the SoC ctypes project we hope :)

* 182	compiler crash on windows

  We need a windows user for this?

* 183	windows build needs superuser

  This too.

* 184	compiled pypy-c is host dependent

  How much of this do we want to do for 0.9?

* 195	Make website doc generation more reliable

  (done, could be better) Holger

* 21	_file.py needs more tests

  Yes, it does.

* 22	app-level docstrings missing everywhere

  Should be an easy sprint task.

* 3	fix real bugs exposed by test_descr

  We can do this now we have weakrefs, but it's a bonus point type task.
  
* 95	create a glossary for common PyPy terms

  Should be done, at last.

Features:

* 101	Tool for discovering REV that broke translation	

  We're surviving without it ok, eric's nightly benchmarks help.

* 103	finish posix module

  SoC!

* 145	implement -m command line option

  Probably not critical for 0.9.

* 200	moving gcs

  Another GC bonus point.  Probably not going to happen for 0.9.

* 34	running translator goals per revision

  See comments on issue101.

* 41	roadmap till PyPy 1.x

  Well, yes.  We need to have this by the end of the sprint, I guess.

* 44	have a doc chapter for CPython-dev'ers

  This is part of getting the ext-compiler to be usable in some sense.
  There should be an "0.9 documentation" issue.

* 74	come up with nice pictures/diagrams

  It's not so much coming up with these things as actually linking to them
  from the docs that remains to be done.

Wish:

* 133	remove unused files

  Part of preparing for the release.

* 4	code integrity in documentation

  No closer to fixing the hard problems at the moment?

* 6	Allow users to specify 'allwaysnosy'

  postponed to 1.0, probably involves updating the tracker schema
  on a larger scale.