Unexpected behavior with TZ

[gmail]

Hello,

While looking after the multibyte test, i notice a behavior with 
printf3.sub that surprises me.

The following screen caps are a good resume :


    [user@pompomgalli] $ cd bash-4.2_build

    [user@pompomgalli] $ ./bash

    [user@pompomgalli] $ echo $TZ

    Europe/Paris

    [user@pompomgalli] $ echo $LC_ALL

    C

    [user@pompomgalli] $ echo $LANG

    fr_FR

    [user@pompomgalli] $ date

    Tue Apr  5 22:04:38 CEST 2011

    [user@pompomgalli] $ TZ=EST5EDT

    [user@pompomgalli] $ export TZ

    [user@pompomgalli] $ LC_ALL=C

    [user@pompomgalli] $ export LC_ALL

    [user@pompomgalli] $ LANG=C

    [user@pompomgalli] $ export LANG

    [user@pompomgalli] $ printf "%-40.50(%a %b %e %H:%M:%S %Z %Y)T 
date-style time\n" 1275250155

    Sun May 30 22:09:15 CEST 2010            date-style time

    [user@pompomgalli] $ date

    Tue Apr  5 16:05:32 EDT 2011

    [user@pompomgalli] $ printf "%-40.50(%a %b %e %H:%M:%S %Z %Y)T 
date-style time\n" 1275250155

    Sun May 30 16:09:15 EDT 2010             date-style time



I see in the CHANGES that bash is able to detect TZ change and call 
tzset() if available,
i then check the build's config.log :


    [root@pompomgalli] # grep -i tz config.log

    configure:13502: checking for tzset
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    | #define HAVE_TZSET 1
    ac_cv_func_tzset=yes
    #define HAVE_TZSET 1



I have probably misunderstood something, or perhaps something is wrong 
with my build environment or there could be a side effect when  building 
in a chrooted jail.
Or perhaps it is an expected behavior with Linux/Glibc arch ? (in all 
this cases, sorry for the disturbance)




> > Another point, i see that the tests/intl2.sub make a direct call to a
> > "/usr/bin/printf" command.
> The intent was to see if and how bash differs from the system printf
> binary (not necessarily coreutils) when changing the locale for the
> duration of the command.  Bash versions before bash-4.2 had a problem
> with assignments to LANG preceding builtins.  The old versions didn't
> temporarily change the locale, though they performed the variable
> assignment.
>
> Chet

Not necessarily coreutils, right ^^.



Regards, Cédric.