Re: [Debian Installer] General release plan for RC1

2006-10-08 Thread Attilio Fiandrotti

Frans Pop wrote:

=== D-I STRING FREEZE: Thu 12 Okt, 00:00 UTC - Sun 22 Okt, 00:00 UTC ===

Hi folks,

This mail really should have gone out at least two and probably three 
weeks earlier. Blame goes to the discussions on d-{private,vote,wherever} 
which resulted in a serious drop in my motivation in recent weeks and in 
me dragging my heels on preparing the release.

My apologies for letting you all down.


GENERAL STATUS
==
The most important work planned for RC1 has been done and the installer in 
general works well, although some finishing touches and testing are still 
planned in the run up to Release Candidate 1 of the installer.


I don't have a hard planned release date yet, but it will probably be 
first week of November. Detailed planning and updates will be sent to 
d-boot and d-release lists.

General status and preparations for Etch RC1 can be found on [1].

As it looks now RC1 will be released with kernel 2.6.17, unless 2.6.18 is 
ready to migrate to testing in time for us to make the switch before the 
release. If not, we plan to switch to 2.6.18 ASAP after RC1 and release 
RC2 ASAP after 2.6.18 does migrate to testing.


How can you help?
-
If you have some time, please help us by testing the installer. We need 
testing especially for the non-mainstream architectures and new features 
recently introduced in the installer (e.g. encrypted partitioning).


Especially welcome is also specific testing by e.g. maintainers of 
packages/functionality used in the installer: RAID/LVM support, PCMCIA, 
filesystems (JFS, XFS, ...), bootloaders, etc.

Please check if the bits *you* care about are supported well!

Watch out for calls for testing on d-d-a!


GRAPHICAL INSTALLER
===
The graphical installer now, for the first time, uses udebs built from 
official GTK library packages (2.8.x), which means we will be able to 
drop the various unofficial packages we have been using up to Beta 3. 
Thanks to Dave Beckett, Loïc Minier and Josselin Mouette for helping us 
get this far!


The GNOME maintainers have decided to stay with 2.14 and GTK 2.8.x for 
Etch. This means that g-i will also stay with the current 2.8.x libs, so 
until the release we need to focus on that and leave work on 2.10.x for 
post-Etch.

There is one RC issue that will need to be resolved before RC1: #390683.


Just today mike emmel fixed the "boom" bug, it should technically 
possible switching to GTK+ 2.10.x.
Moreover, after some debugging from my side yesterday mike possibly 
found where to look to fix #390683.
If mike manages to fix #390683, we should be able to backport the fix 
into our patch tarball for GTK 2.8.20, unless (i guess) the fix is 
located in that huge block of code added in july to corrctly implement 
painters, which our tarball currently does not include because it's 
dated May 2006.
In the latter case, we should perform a from scratch backport of the 
gtkdfb backend found in current GTK 2.10 CVS to 2.8.20.
In the case a clean backport should be overcomplicated (i failed in this 
some weeks ago), would this be a valid reason to switch to a clean GTK 
2.10.x set of libs for the d-i only?


cheers

Attilio


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: [HELP] Is separate lib package needed for libcairo directfb variant

2006-06-19 Thread Attilio Fiandrotti

Frans Pop wrote:

Hi all,

Dave Beckett is preparing a new release of libcairo2 (see [1]) after a 
request from the Debian Installer team to add a udeb compiled against the 
directfb backend and not against the X backend. This udeb will be used by 
the d-i team for the graphical installer.




This is also where it gets really complex as compiling pango specifically 
against libcairo2-directfb would imply that we also need a separate 
variant of pango, but so far we've been running a "normal" pango library 
against a cairo2-directfb library in the graphical installer without any 
obvious problems (like missing symbols).


It may be that we're missing obvious stuff here, so comments/suggestions 
from experienced library wizards are very welcome.


i'm not a library expert at all, so i may be wrong or missing something 
important, but pango should be totally indipendent from specific 
backends used by both cairo and gtk.


[EMAIL PROTECTED]:/usr/lib$ ldd libpango-1.0.so
linux-gate.so.1 =>  (0xe000)
libgobject-2.0.so.0 => /usr/lib/libgobject-2.0.so.0 (0xa7e9b000)
libgmodule-2.0.so.0 => /usr/lib/libgmodule-2.0.so.0 (0xa7e97000)
libdl.so.2 => /lib/tls/i686/cmov/libdl.so.2 (0xa7e93000)
libglib-2.0.so.0 => /usr/lib/libglib-2.0.so.0 (0xa7e0b000)
libm.so.6 => /lib/tls/i686/cmov/libm.so.6 (0xa7de5000)
libc.so.6 => /lib/tls/i686/cmov/libc.so.6 (0xa7cad000)
/lib/ld-linux.so.2 (0x7000)

we always used gtkdfb 2.0.9 library with the pango library provided by 
existing pango udeb, and judging from experiments both davide and i made 
with GTKDFB 2.9.x and GTKDFB 2.8.17, current standard pango udeb works 
well with more recent GTKDFB libraries too.


A libcairo2-directfb library deb could also be generally useful for other 
projects based on directfb.


i think embedded linux applications developers will greatly benefit of 
precompiled [gtk|cairo]dfb libraries


Attilio


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]