incubator-stdcxx-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Martin Sebor <se...@roguewave.com>
Subject Re: [PATCH] Windows infrastructure for generating VisualStudio projects and solution(s)
Date Wed, 16 Aug 2006 18:28:40 GMT
Andrew Black wrote:
> Greetings Farid
> 
> My suspicion is that they are supposed to be exported.  My understanding 
> is that those variables tell the locale and localedef utilities exactly 
> what language they are dealing with.

Yes. They determine the locale to be used by the utilities. I haven't
looked at the script but in general, only one of the two variables is
usually used: LC_ALL overrides the setting of any and all other LC_XXX
variables as well as LANG in the environment. LANG provides the default
setting of any unset LC_XXX variables when LC_ALL isn't set.

> On my machine, it appears that the pair of variables are auto-exported 

My understanding of the POSIX requirements is that only variables
from the shell's environment or those explicitly marked for export
are exported to the environment of the child process. See
http://tinyurl.com/6ckth#tag_02_12.

> to the child processes, and I have observed similar behavior to what you 
> observed after explicitly exporting the variables to the child processes.
> 
> I have done a little digging, and it seems to me that the issue 
> originates in the LC_COLLATE section of the out.1 file.  Either the 
> logic to fill the section is incorrect, or the logic to parse the 
> section is incorrect.

Okay, so someone needs to look into this and figure out what's
causing it. Andrew, unless you plan to continue your investigation
into the problem in the immediate future could you please open an
issue in Jira? (Give as much detail as possible, including test
case, stack trace, etc., so that we can readily reproduce it when
we get around to looking into it again).

> 
> Attached is a patch containing changes I made to the run_locale_utils.sh 
> script that I made while trying to track down why I was seeing test 
> failures building locales.  This also incorporates your observations on 
> the missed exports.  If it is to be committed, a changelog entry follows.

This makes sense. Thanks a lot for the patch! It would also be nice
to figure out if we need to be defining and exporting both LANG and
LC_ALL or if LC_ALL alone is enough.

Committed thus:
http://svn.apache.org/viewvc?rev=431972&view=rev

> 
> --Andrew Black
> 
> Log
>     * run_locale_utils.sh (generate_locale): Redirect stderr from 
> localedef utility to $dbgout
>       (test_locale): Export LC_ALL and LANG
>       echo all export directives and values to debug output

Please try to keep this in mind http://tinyurl.com/otfbk.

Martin

Mime
View raw message