Personal experiences/anecdotes of the bzr review processes

[Mark Hammond]

I noticed that the bazaar review process has been discussed here recently,
and a recent experience of mine is a reasonable example of my experiences
submitting patches, so I thought it might be worth sharing my perception of
the process.

The anecdote is the thread "[MERGE] add
win32utils.get_local_appdata_location()" and I hope I don't come across as
pointing fingers at any people, just at processes.  Here is some background
of the issue itself and the patch:

* bzr provides a function to query the windows API for the "app data"
directory.  It does this by calling a single windows API function with a
particular param.
* It turns out Windows provides *two* app-data locations, and a bzr-svn
feature request correctly suggests some data from bzr-svn be stored in the
second.
* Jelmer suggests bzr provide this location, I agree to assist by providing
a patch in an effort to help both bzr-svn and bzr itself.
* Importantly, this entire feature involves calling the *exact same* API
function as in the first point, but with a different param.
* Possibly foolishly, I add tests for both the existing calls to that API,
and for the new calls to that same API with the different param and for
exposing a different way of calling the exact same windows API function
(ctypes vs pywin32).  OTOH, I feel confident that if I submitted that patch
*without* adding tests even for the existing non-tested cases, it would have
been rejected.

The timeline for this patch:

* 23 August, I submit a patch
* 27 August, John votes "resubmit" with (mostly valid; some even pep8)
stylistic issues.
* 27 August, I ask for clarification.
* 29 August, John clarifies.
* 30 August, I resubmit
* 4 Sept, John asks for another tweak in a comment block
* 4 Sept, I resubmit
[Note: no 'approve' from John ever comes]
* 19 Sept, Ian reviews and notes tests fail on non-windows platform.
* 19 Sept, I resubmit
* 19 Sept, Ian votes approve
[first approval!]
* 23 Sept, Aaron re-reviews, picks up on various stylistic issues, it seems
we are back at square one. 
* 23 Sept, my frustration grows :)

At this point the patch is fundamentally the same as I submitted it (give or
take some double blank lines as per pep8 and skipping the new tests in some
cases), bzr-svn has long ago submitted the change to take advantage of this
feature when it deigns to appear, and I'm awaiting instructions for yet
another resubmission request.  While I understand the importance of keeping
a close eye on quality, the reality is from my personal POV, the cost of
submitting improvements that aren't *critical* for my work are very close to
being too high to bother with...
 
I hope this is constructive,

Mark