Re: [README] New development cycle

[Teika Kazura <teika lavabit com>]

Hi. 

On Wed, Nov 19, 2008 at 03:56:50PM +0100, Janek Kozicki wrote:

> And also I wanted to encourage Scott Scriven...
> ...
> testing environment (which allows easy restarting of sawfish 
> and testing it, without closing your entire session).

Scott, (do you hear me?) how unstable can your experimental codes?
If it sometimes crashes the X server, it's a disaster, and we've got to
refine librep. It's quite a task, but I'd say it's the fastest.
But if it is not so unstable, then many workarounds are conceivable,
putting together small hacks. (For example, I once considered 
--lisp-dir & --custom command line option. The latter overrides 
~/.sawfish/custom . Or a simple daemon process can restart the stable 
WM when a USB flash RAM is inserted.)

Anyway, if you feel like, let us hear you. It's good to share the
information.

Timo wrote On Thu, 20 Nov 2008 06:52:23 +0200:
>> if there is a regression, then everyone can go to wiki and try
>> unapplying each patch to find the root cause of the problem.
>
> It is easier to unapply commits with svn (or whatever DVCS one happens
> to be using on top of svn).

Right. At the same time, having patch on the web remains valuable, because
you can find it easily by searching.

> How about a feature branch just for it[Scott's tab] then?

The only difficult experiment now is the tab. So a branch sounds nice.

But if we've got many branches, it is difficult to juggle. (Let's face
the fact; we are short of developers.) So if the tab is put in a branch,
how about giving a label (or category) 'in continued test phase'
to other patches now in 'experimental', and to stop the current
experimental release?

Anyway, it often seems that patches in 'experimental' needs not 
only testing, but also additional explanations, probably by testers.
It is also better that testers write their name in the patch page, 
similar to 'Assigned to:', so that the progress can be seen.

Teika kazura