[sldev] [QUESTION] What was issue SL-35450

Ryan Williams (Which) rdw at lindenlab.com
Mon Oct 15 15:29:44 PDT 2007


"Startup crash with non-english locales (linux client)"
"Likely, mozilla beats us to GTK initialization and then GTK calls 
setlocale() on us (way to go, GTK).  More precisely, Mozilla cheerfully 
plays with the locale in multiple places. I believe I've isolated them 
all and wrapped them in code which restores the previous locale now."

Tofu might have more knowledge in his head than notes in JIRA, but 
here's the repro:

* Run the Linux version of First Look 1.13.3.58100 as
LC_ALL=de_DE.utf8 ./secondlife
to ensure that you can reproduce the crash at all; it should crash while 
logging into the world. If you cannot reproduce this crash, you may have 
to install the 'langage-support-de' package or other distro-specific 
German-language support package such that running 'locale -a' lists 
de_DE.utf8 as an option.

-RYaN


alissa_sabre at yahoo.co.jp wrote:
> I need a help from a Linden.
>
> Quick question:
>
>     What was the issue SL-35450?
>
> Longer question:
>
> In the viewer source file linden/indra/llwindow/llwindowsdl.cpp, there
> is a method LLWindowSDL::gatherInput().  In the method, there is a GTK
> event loop to drive SL viewer's embedded web browser.  It is enclosed
> by some unusual codes with the following comment:
>
>     // Yuck, Mozilla's GTK callbacks play with the locale - push/pop
>     // the locale to protect it, as exotic/non-C locales
>     // causes our code lots of general critical weirdness
>     // and crashness. (SL-35450)
>
> So, the enclosing codes solve a problem identified by SL-35450.  To
> understand the problem, I disabled the "push/pop" codes, and tested
> the resulting viewer.  Unfortunately (or, fortunately?  :-) the viewer
> ran with no weiredness or crash.  I guess I didn't hit the trigger of
> the problem in my test run.
>
> Searching the public JIRA with SL-35450 found nothing.
>
> I guess this is an LL internal bug tracking number before the viewer
> source was open-sourced.  I want to know what was the weired and/or
> crashing behaviour before the fix and possibly a repro so that I can
> observe what happens if I run the GTK event loop in non C locale.
>
> Background:
>
> I'm recently making some research/experiment on use of GTK immodule
> API to replace XIM in the Linux version of the SL viewer.  (See JIRA
> issue VWR-2261 filed by Tofu.)
>
> One of the possibility under eveluation is to use GTK event loop for
> keyboard handling entirely.  To do so, GTK event loop should be run
> under the user's locale, since most of the Chinese/Japanese/Korean
> immodule doesn't work in C locale, and many immodules interact with
> GDK event handling layer directly (through event hooks and signals,
> for example).
>
>     Alissa
> --------------------------------------
> Easy + Joy + Powerful = Yahoo! Bookmarks x Toolbar
> http://pr.mail.yahoo.co.jp/toolbar/
>
> _______________________________________________
> Click here to unsubscribe or manage your list subscription:
> /index.html
>   



More information about the SLDev mailing list