[Libreoffice] [REVIEW] 3-5, 3-5-0, 3-4 corrupted Indic characters and crash under Windows fdo#44208, fdo#45107

[Andras Timar]

2012/1/25 Caolán McNamara <caolanm at redhat.com>:
> On Wed, 2012-01-25 at 16:44 +0100, Andras Timar wrote:
>> Result on Windows is confusing. My test language was Hindi (hi).
>>
>> Case 1 - without the patch in officecfg
>> ...
>> default-font entry for "hi" "IN" "" added
> ...
>> searching for a node for "hi" "" "" for type UI_SANS
>
>> BTW it was not called from outdev3.cxx
>
> Righteo, so might as well get the backtrace from unotools when the
> locale Language is "hi" and locale Country isEmpty() and see where that
> bare "hi" comes from.
>
> I suspect that "hi" was basically injected in from from GetUILocale
> maybe and *probably" your /path/to/3/user/registrymodifications.xcu has
> a oor:name="UILocale" ... <value>hi</value>, right ? not
> <value>hi-IN</value>.
>

Yes, it is hi, not hi-IN.

> If that's the case, then fair enough, I'm happy that there's nothing
> horribly busted here and its simply an unexpected side-effect of
> renaming hi-IN etc to hi uncovering that the config for default fonts
> was over-specialized.
>
> Assuming that UILocale is set as "hi" in the config then I suspect that
> comes originally from code like optgdlg.cxx:1482 aLangString =
> seqInstalledLanguages[d-1] where the name of the language comes from the
> name of the installed language packs.

I still don't understand what's happening, i.e. in good case it reads
UI_SANS list for "hi" once, then several time for "en", and it looks
good, and in bad case it reads UI_SANS list for hi_IN several times
and still, we get corrupted characters.

Cheers,
Andras