20074 - RFE: Support for Solaris >= 2.7 Unicode locales (e.g. like en_US.UTF-8)

Status: RESOLVED INVALID

(SeaMonkey :: General, defect, P3)

Description

[Roland Mainz]

Solaris >= 2.7 comes with a en_US.UTF-8 locale (and other Unicode locales like
de.UTF-8, en_US.UTF-8, fr.UTF-8 and so on...).
I think we should verify that Mozilla uses the special features of this locales
if theya're available.

Comment 1

[leger]

Sending over to mcafee

Comment 2

[Chris McAfee]

tao, is this your area?

Comment 3

[Roland Mainz]

Small comment: I'll suggest a Solaris 2.7 SPARC with (at least) kernel patch
106541-05 (I'll recomment the current 106541-08 or better: MU4), because before
this patch some nice things like scanf() didn't work properly with *.UTF-8
locales.
But after all: This locale is _very_ usefull. Maybe someone hacks bash2 to work
with this locales - then this is near "perfect" :-)

Comment 4

[tao]

Unless *.UTF-8 variants do something (such as collation) different than normal
locales, this RFE would mean proper mapping of the xyz.UTF-8 to xyz. We are
using unicode across our modules already. All locale sensitve data buffer are
passed as ucs-2 and stored as UTF-8 :-)


Naoki:

Comments?

Comment 5

[nhottanscp]

Does that provide UTF-8 clipboard (that would be usefull)?

Comment 6

[Roland Mainz]

Take a look at http://soldc.sun.com/articles/wp-unicode.html, maybe you'll find
there some usefull answers. I'm still learning how these UTF-8 locales are
working...

Comment 7

[tao]

It seems to me the *.UTF-8 locales affect the runtime messages, native widgets,
and IME.

1). All UI strings come from Mozilla are already encoded in UTF-8 and
    passed to the native widgets in UCS-2 which might cause corrupted
    UI strings in native widgets. We'll need to test this on 2.7.
    XPWidgets are not affected by this locale.
2). Other locale categories such as collation, date, etc are not affected by
    this since we are using the system locale utils. Naoki, would you confirm
    this?
3). Printing. Since our internal buffer are ucs-2 encoded, we'll need to test
    the printing as well.
4). IME. CC frank for comments.

Pass to Naoki for comments first.

Comment 8

[Roland Mainz]

Please check your Solaris 2.8 patches. I strongly suggest to apply at least
kernel patch 106541-08 before making any assumptons/tests with the *.UTF-8
locales. Better is to apply patch cluster MU4 (Maintaince Update 4) which fixes
(AFAIK) all known issues with *.UTF-8-locales.

Comment 9

[Roland Mainz]

> Please check your Solaris 2.8 patches...

AAhhgrrr, my mistake. I mean Solaris 2.7, not 2.8.
Those problems should be wiped-out in 2.8 FCS...

Comment 10

[nhottanscp]

Answering tao's second question, date format and collations are affected by
lang/country information but not by charset. So I think UTF-8 locale has no
effect on those. But it may affect the sort order of out of locale characters
(e.g. order of Japanese characters when using fr.UTF-8).

Comment 11

[tao]

How about the IME, Frank?

Comment 12

[Frank Tang]

I think this is an invalid bug. This bug does not report ANY wrong behavior. It
simply assume we didn't do the right thing w/ UTF-8 locale. I will mark this bug
invalid for now. Please open if you KNOW for sure which code is broken (tell me
which file which line) or what feature is broken (give me test cases and test
procedure, expecting result and actual result).

We do convert IME string from the locale encoding to UCS2. We may have
copy/paste and window title, but that part is not UTF-8 specific at all.

Look at mozilla/intl/uconv/src/unixcharset.properties which map UNIX locale name
to a charset name we used internally. Submit diff to me if I miss anything
there.

Please do not report this kind of meta guess bug. If you see some wrong
behavior, report them into individual bug report. If you just guess we didn't do
the right thing, try it first.

Comment 13

[Roland Mainz]

I didn't file this as a bug, this was a RFE (Request for Enhancement) - mainly
to _inform_ you that there's a new feature in Solaris >= 2.7 which can be used
to simplify work/development of Mozilla and to verify that this feature won't
break anything within Mozilla.