svn commit: r303218 - head/tools/tools/locale/tools

[Andrey Chernov]

On 23.07.2016 15:32, Baptiste Daroussin wrote:
> On Sat, Jul 23, 2016 at 02:58:31PM +0300, Andrey Chernov wrote:
>> On 23.07.2016 14:52, Baptiste Daroussin wrote:
>>> On Sat, Jul 23, 2016 at 02:41:39PM +0300, Andrey Chernov wrote:
>>>> On 23.07.2016 14:24, Baptiste Daroussin wrote:
>>>>> Author: bapt
>>>>> Date: Sat Jul 23 11:24:05 2016
>>>>> New Revision: 303218
>>>>> URL: https://svnweb.freebsd.org/changeset/base/303218
>>>>>
>>>>> Log:
>>>>>   Add another hack to add weekday to date format to more locales
>>>>>
>>>>>   While here only initialize the iconv converter when needed
>>>>>
>>>>> Modified:
>>>>>   head/tools/tools/locale/tools/cldr2def.pl
>>>>
>>>> BTW, is there any progress in generating all collates with -m ${area} as
>>>> you planned?
>>>>
>>> you mean the @modifier?
>>
>> No, I mean that:
>>>> Perhaps all uk_UA should be generated with -m ${area} too - Ukrainians
>>>> have extended Cyrillics in their alphabet and I am not sure that their
>>>> order match default Unicode order for Cyrillic. The same about be_BY,
>>>> bg_BG and sr_Cyrl_RS. They may match or may not, I don't check such
>>> deeply.
>>>>
>>>> It seems the safest side will be to generate all with -m ${area}
>>>>
>>> Yes that is my plan, except for the 1 or 2 that cannot be generated
>> >from cldr
>>> directly
>>
> 
> Nope I haven't made any progress, and I actually did a bunch of tests of some
> (honnestly not all locales) trying to get a representative set and could not
> find differnce in ordering between locales generated with -m ${area}.
> 
> It would also needs lots of work on the generation tools.
> 
> I first need to write a tool that compares collation rules for various ${area}
> (not done yet) to see how much this is really needed and use it in the
> generation tools.
> 
> I will try to do that for 12 if it is really worth it (based on the result from
> that comparison tools)

IMHO we don't need comparison tools, just blindly adding -m ${area} to
generating of every collate with different area be enough. CLDR may
involve from time to time including national alphabets changes (which
may happen often enough too), so it will be better to not run such tests
every time with new CLDR import, but just the make code which do it
correctly independently (-m ${area}) and forget about it, it will be
much easier to maintain. Moreover, you'll spend more time writing such
testing tool than adding -m ${area} :)



-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 455 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/svn-src-head/attachments/20160724/4144735c/attachment.sig>