[php-maint] Plans for wheezy

[Raphael Geissert]

Hi everyone,

Although sid has been kept up to date thanks to Ondrej (and recently Sean), I 
think it is time to define the goals we want to achieve for wheezy, and what 
it would take to get there.

I can think of a few, but please comment and compliment:

* Ship 5.4.x
Requires: another transition, there's an ABI bump
We would need to prepare packages of 5.4 in experimental. The sooner the 
better. The longer we can make it available for testing will help stabilise 
it. At least the preparations should be made. I haven't followed the commits 
for a few months now, so I don't know how the code is, quality-wise.
The request for mysqlnd comes again, and even though I don't really like it, 
we will probably have to make the switch soon.

* Trimming SAPIs by shipping some extensions as additional packages
There are a few extensions which I doubt anyone actually uses. With the 
current packaging approach we have no idea who uses what extensions.
My proposed solution would be to first add virtual packages for each extension 
included in the core (excluding main stuff like: core, sockets, pcre, session, 
zlib, etc.)
Then we would ask package maintainers to add dependencies as needed, so that 
we can eventually split some of them from the SAPIs. Versioned dependencies 
are rare, so we would be fine with virtual pkgs for now.

* Include an extensions manager, finally
There's not much to say about this, I think. The basic approach I think is 
very well understood. What language is used by the manager, and the way 
conflicts would be declared need to be decided, though.
My original work is rather messy, so it definetely needs a rewrite.

* Improve build-time testing
I can think of testing the pgsql-related extensions and enabling the http-
based tests. Not sure if we also want to test the ldap extension while at it.
I remember reading about an internal http server being shipped so that at 
least some of the http tests could be run, but we really need to consider 
specific SAPIs (especially apache.)

* Reduce build time
The testing phase is taking the most time nowadays. Some time ago I wrote a 
patch that allowed for multiple processes to be started in parallel. It had a 
few issues with the collection of test results, but nothing that can't be 
fixed.
http://marc.info/?l=php-internals&m=126680115113101
I'm rather hopeless on the long-awaited rewrite (*cough* extensions manager 
*cough*) mentioned on the thread.


Anything else?

It seems like the usual burn-ratio of PHP has hit Ondrej, so I'm going to try 
to find more time to get things moving here.
(yeah, right, especially because I wrote most of this email like five months 
ago :-/)

Cheers,
-- 
Raphael Geissert - Debian Developer
www.debian.org - get.debian.net