On 2016-04-28 15:28, Mattia Rizzolo wrote: > Some days ago I finished the work to have it synced (src:hunspell-sv > needed RM for $reasons, and to do so src:ubuntu-keyboard and the > seeds needed changes), and got it synced, so starting for yakkety the > package is the same.
Saw that, great! > On Thu, Apr 28, 2016 at 01:10:52AM +0200, Gunnar Hjalmarsson wrote: >> I gave it another shot in the light of how LO recognizes locales, >> and dropped a few symlinks including el_CY. Attached please find a >> new, slightly reduced list of proposed symlinks > > to me that list look mostly sane, but a bit pops in my mind: for > hunspell-kmr, shall we provide new virtual packages hunspell-ckb and > hunspell-ku? To be quite honest, I don't know how useful the proposed ckb_IQ and ku_TR symlinks are. They do work in the sense that the kmr_Latn dictionary is recognized by LO in case of a ckb_IQ or ku_TR locale, but multiple Kurdish dialects exist, so doing some research before creating such packages may be advisable. On 2016-04-28 17:06, Agustin Martin wrote: > myspell-es ships a bunch of symlinks for that purpose after request > in a Debian bug report, but later found some bug reports in Ubuntu > about just the opposite, > > https://bugs.launchpad.net/ubuntu/+source/espa-nol/+bug/663032 > https://bugs.launchpad.net/ubuntu/+source/espa-nol/+bug/793666 > https://bugs.launchpad.net/ubuntu/+source/espa-nol/+bug/927256 > > myspell-es is the same for all countries, so I think that leaving > symlinks there is by far the lesser evil, even if it adds some > noise. Yeah, I have also seen that kind of complaints in a few bug reports and Ask Ubuntu questions. There are two possible approaches in my mind to reduce the noise/pollution. I just made a Spanish install of Ubuntu, stating New York as the time zone location, and the installer added LANG="es_US.UTF-8" to /etc/default/locale. Changing the installer behavior in this respect, so it instead set LANG to "es_ES.UTF-8", while setting some formats related LC_* variables to "es_US.UTF-8", would make it possible to get rid of quite some polluting symlinks. Another idea is that the source packages, e.g. libreoffice-dictionaries, provide the symlinks in separate binary packages, so for Spanish we would then have: hunspell-es hunspell-es-mx hunspell-es-ar hunspell-es-us ... with the actual dictionary in hunspell-es and each of the other hunspell-es-* packages only containing a symlink and a dependency on hunspell-es. But I suppose that those thoughts are more long term ones, and would need careful consideration. -- Gunnar Hjalmarsson https://launchpad.net/~gunnarhj
