[Dextrose] [Fwd: Dextrose patch queue]

[Bernie Innocenti]

On Sat, 2010-11-27 at 20:57 +0100, Sascha Silbe wrote:
> Excerpts from Bernie Innocenti's message of Fri Nov 26 22:56:15 +0100 2010:
> 
> > > 6b2b970 Fix duplication of OLPC mesh icons
> > >0521ff7 Indicate inactive state in mesh device icon  
> > 
> > Have these bugs been fixed in a different way upstream?
> 
> Don't know, but Dextrose disables mesh anyway.

Ah, I think this bug would also affect the ad-hoc icons (unless the code
has been duplicated).


>  Personally I'd rather
> drop pre-802.11s mesh support altogether (at upstream).

Draft or not, 802.11s is not being used by anyone except for OLPC... Now
even OLPC dropped it... let's face it: it was a fascinating idea, but
not very useful in practice and certainly not worth its cost huge in
terms of complexity and fragility.


> > > sugar
> > > -----
> > > 42d2d96 Click on journal icons with a exclusive time frame
> > > - should use launcher feedback or fix race condition instead of fixed timeout
> > 
> > Indeed. The actual bug may have been fixed some time ago, can't remember
> > the details.
> 
> It has [1,2]. Patch marked as to-be-dropped.

Go ahead.


> > > e15f79e3 Add font dpi schema
> > > - configure X server instead?
> > 
> > Sayamindu would now better. This is complicated shit!
> 
> Yeah, I would like to see a complete rationale for this change.

The sugar-settings-manager part of this patch in already in git. By
itself, it wasn't sufficient to fix the bug in Browse with form buttons
which is now visible in F14-S0.90.

Eventually, we dropped sugar-settings-manager instead of trying to
figure out which part of our XSETTINGS protocol implementation was not
good enough for xulrunner.


> > > da33877 Change the method to add new actitivities in the activities list in home view
> > > - rationale? fix sorting instead?
> > 
> > Yes, the activities were appearing in backwards order (oldest first).
> 
> SL #1660 [4]. It would be nice if aa gave the real fix a try. The
> problem I had doesn't seem to happen for the Journal sorting, so he
> probably knows how to avoid it.

Sure, let's try to improve the fix (within a reasonable amount of
effort)


> > Using an activity is ok with me, but remember that backups must work
> > with the *current* schoolservers in deployments. No special server-side
> > software.
> 
> How does Backup+Restore work with current XS? Does it integrate with
> Sugar 0.82 based builds? If so, why was it dropped for 0.84? Or what
> changed that broke the integration?

The journal backup offered by the OLPC builds is a daily cronjob that
has no UI and no knowledge of the journal format. There's no real
restore functionality, except for a web page that lets you retrieve one
file at a time from the schoolserver.

In Dextrose, we maintained the automatic backups and added a UI to
trigger a backup manually. We also added a UI to restore the entire
journal from the backup.

These two features, combined, allows us to upgrade from any older
version of Sugar without loosing data. Datastore format upgrades are
taken care of transparently at startup.

It's hard to beat the simplicity and performance of rsync. On the server
side, there's nothing but a restricted shell account. It works with any
version of Sugar, past and future. Daily backups are hard-linked
together to save space. A typical datastore of 500MB can be backed up
over the network in less than one minute if there aren't too many new
objects.

I'm ok with moving this functionality to the Backup and Restore
activities, but switching to a higher-level storage format may not buy
us any.


> We should also have a good look at the DataManager activity [5].

Looks *very* promising. We were just looking to add a UI for browsing
the objects stored in the XS. We were thinking of mounting it with fish
or something like that, and make it visible with a device icon.


> > > 54932cb Backup translations
> > > - can we use upstream Pootle for this?
> > 
> > I don't think Pootle can be made to work with code which isn't yet in
> > git.
> 
> Then we'll put it in git. It doesn't need to be the same source as we use
> for building images, only contain the patches that need translations.

Ok. Do you know if Pootle support translations done directly in git or
does one have to always go through the web interface?


> > > 2d03290 + 09fd691 + e8b1da6 + f31bd6f + 080d1f1 + 3f7ec91 + ecda730 + 548dee3 accessibility series
> > > - check approach
> > > - do we need to patch Sugar or can we package it separately?
> > > - style cleanup
> > > 70db1a3 accessibility translations
> > 
> > This may take quite some work. Would you like to take it?
> 
> I'll eventually need to. ;)
> If you could initiate discussion with the Design Team and Sayamindu,
> that would be quite appreciated and a very good start. OLPC is going
> to need at least the virtual keyboard support for their tablet model, so
> maybe we can offload parts of it to them.

I know close to nothing about the implementation. The author of most of
this work is Esteban Arias. He might be interested in doing some of the
needed cleanups himself.


> These patches are only about a few tweaks for people with physical
> disabilities, BTW. Making Sugar work properly with AT-SPI & co (e.g. for
> blind people) is going to be much more invasive.

Yes, my understanding is that more work would be needed, but for the
time being it's quite a step forward. The large fonts and high-contrast
scheme are appreciated by teachers with glasses. The virtual keyboard is
appreciated even vy some users with no disabilities: it saves the day
with broken keys on the XO membrane keyboard.

-- 
   // Bernie Innocenti - http://codewiz.org/
 \X/  Sugar Labs       - http://sugarlabs.org/