Fedora 18 issues with translations and keymaps

Adam Williamson awilliam at redhat.com
Thu Jan 3 08:03:24 UTC 2013


Hey, folks. I'm not really sure how to frame it, but the result of all
my poking about at keyboard layout bugs and related stuff recently is
that I'm pretty sad at the state of support for
anything-but-U.S.-English in Fedora 18.

Here's the tally:

* https://bugzilla.redhat.com/show_bug.cgi?id=889562 - systemd
conversion from xkb to console layouts fails probably more than it
succeeds, when it does, you wind up with U.S. English as your console
layout, not whatever you picked during installation

* https://bugzilla.redhat.com/show_bug.cgi?id=891487 - anaconda doesn't
seem to manage to offer all the keyboard layouts it could do, and some
of the ones it's missing are somewhat important

* https://bugzilla.redhat.com/show_bug.cgi?id=891489 - anaconda's
mapping of 'native' layouts to user's chosen install language doesn't
work in all cases

* https://bugzilla.redhat.com/show_bug.cgi?id=878433 (?) - GNOME has a
weird predilection for coming up with the obscure 'Bambara' layout in
user sessions after an install in a non-English language, but not the
correct layout for that language

* https://bugzilla.redhat.com/show_bug.cgi?id=881624 - X and GNOME
keyboard layouts revert to U.S. English on upgrade to F18

* https://bugzilla.redhat.com/show_bug.cgi?id=854557 - the 'layout
testing' in the keyboard spoke doesn't work at all how you'd expect it
to

* https://bugzilla.redhat.com/show_bug.cgi?id=882440 - in fact, people
have various problems interacting with and understanding the keyboard
spoke at all, really. Several of the issues discussed in this bug are
'greatest hits', especially the lack of a default layout switch command,
the fact that anaconda doesn't automatically start using a layout you
promote to the top of the list, and the lack of any kind of indicator in
anaconda as to what layout is in use

* https://bugzilla.redhat.com/show_bug.cgi?id=859641 - we're picking the
wrong default keyboard for Dutch, apparently

* https://bugzilla.redhat.com/show_bug.cgi?id=867110 - ...and German
(Switzerland)

* https://bugzilla.redhat.com/show_bug.cgi?id=885345 - ...and Dutch
(Belgian)

(there's a few more along those lines which I won't bore anyone with)

In addition to those bugs, we have fairly significant regressions in the
completeness of anaconda translations between Fedora 16 and Fedora 18
(the numbers for F17 for some languages are weird - a lot of languages
show 55% completion for F17 but 90-100% for F16, which seems bogus, as
I'm sure things didn't change that much between F16 and F17, so I'm just
using the F16 numbers. I assume there's some weirdness that explains the
odd 55% numbers for F17, but if not, hey, F17 was kinda boned too...):

Language		F16		F18
Finnish			93%		75%
Indonesian		100%		33%
Kannada			94%		33%
Oriya			94%		27%
Telugu			94%		32%
Bengali (India)		93%		33%
Portuguese		100%		36%
Persian			95%		27%
Malayalam		78%		20%
NorwegianBokmal		92%		55%
Bengali			93%		33%
Sinhala			93%		27%
Serbian			81%		23%
Serbian(Latin)		81%		23%
Hebrew			83%		22%
Catalan			68% (98% F17)	25%
Latvian			88%		20%
Greek			68%		21%
Turkish			79%		21%
Maithili		67%		18%
Asturian		85%		24%

(from https://fedora.transifex.com/projects/p/anaconda/ )

There are several others around the 50-70% complete mark for F16, too, I
just cut off at 67%. It's a fairly long list of languages for which we
previously had tolerably complete translation coverage, but we now have
a level which isn't really usable: basically these languages have gone
from 'covered' in at least F16 (probably F17 too) to 'not covered' in
F18. I want to stress I'm not blaming anyone for this: I don't see how
it could be otherwise, the problem is the huge amount of string churn
caused by newUI, I'm actually more amazed at how many languages _do_
have close-to-100% translations for F18 than how many _don't_, given the
conditions. It's just an unfortunate thing.

anaconda does not list every available translation for the user to
choose from, but I cannot figure out for the life of me how the decision
about which to display is made: it's not that very incomplete
translations are left out, as plenty of very incomplete translations
(including most of the above) are shown. If we filtered out very
incomplete translations we might at least look a bit less silly, but we
don't seem to be doing that. Our language selection screen is definitely
writing checks that our translations can't cash :)

I don't really know where we need to go with this, exactly, but a few
questions arise naturally:

1. In the short term, is the combination of all these factors enough for
us to want to delay F18 further to try and make things suck less?

2. If not, do we want to engage in some Messaging around the F18 release
to emphasize that we know there are all these issues and we'll try to
smooth things out for F19?

3. In the longer term, how can we get anaconda, i18n, systemd, GNOME etc
folks all pointed in the same direction and working so that there's far
less suckage and far more smooth interaction going on here? Should we
try and run some sort of session at FUDCon?

Thanks, folks!
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net



More information about the test mailing list