Bug 9780 - issues with time_get, time_put not being tied to correct _punct facet.
Summary: issues with time_get, time_put not being tied to correct _punct facet.
Alias: None
Product: gcc
Classification: Unclassified
Component: libstdc++ (show other bugs)
Version: 3.4.0
: P3 normal
Target Milestone: ---
Assignee: Benjamin Kosnik
Depends on:
Reported: 2003-02-20 18:36 UTC by Benjamin Kosnik
Modified: 2016-01-16 05:57 UTC (History)
2 users (show)

See Also:
Known to work:
Known to fail:
Last reconfirmed: 2005-09-26 00:37:58


Note You need to log in before you can comment on or make changes to this bug.
Description Benjamin Kosnik 2003-02-20 18:36:00 UTC

And followups.

Seems like an issue with __timepunct not really being used  correctly with time_get and time_put facets. 

When fixing this, please double check the other facets for this behavior as well. Likely candidates are money_put, money_get, num_get, num_put, etc: anything that uses a *_punct facet for data collection.

gcc-3.2, gcc-3.3, gcc-3.4 not applicable to releases before gcc-3.2

x86/linux, gnu locale model

#include <iostream>
#include <locale>

int main ()
     const std::locale loc ("de_DE");

     const std::time_put<char> &tpc =
         std::use_facet<std::time_put<char> >(loc);

     std::tm t = std::tm ();

     for (t.tm_wday = 0; t.tm_wday != 7; ++t.tm_wday)
         tpc.put (std::cout.rdbuf (), std::cout, ' ', &t, 'A') = '\n';


#include <iostream>
#include <locale>

int main (int argc, const char *argv[])
     const std::locale l1 (argc > 1 ? argv [1] : "");
     const std::locale l2 (argc > 2 ? argv [2] : "");

     const std::locale l1_l2 =
         std::locale (std::locale (std::locale::classic (),
                      &std::use_facet<std::time_put<char> >(l2));

     const std::time_get<char> &tg =
         std::use_facet<std::time_get<char> >(l1_l2);

     const std::time_put<char> &tp =
         std::use_facet<std::time_put<char> >(l1_l2);

     std::tm t = std::tm ();

     std::ios::iostate err = std::ios::goodbit;

     // benign but unnecessary
     // std::cin.imbue (l1_l2);
     // std::cout.imbue (l1_l2);

     tg.get_weekday (std::cin.rdbuf (), std::istreambuf_iterator<char>(),
                     std::cin, err, &t);

     if (err & std::ios::failbit)
         return 1;

     tp.put (std::cout.rdbuf (), std::cout, ' ', &t, 'A') = '\n';
Comment 1 Benjamin Kosnik 2003-03-11 15:05:45 UTC
Responsible-Changed-From-To: unassigned->bkoz
Responsible-Changed-Why: Mine.
Comment 2 Benjamin Kosnik 2003-03-11 15:05:45 UTC
State-Changed-From-To: open->analyzed
State-Changed-Why: No longer super-concerned about this, although it is a bug. 
    It doesn't apply to money or numeric facets, because they have their own *punct facets, and this kind of code would not be attempted (instead, the proper *punct facet would also be used).
    I consider the lack of timepunct in the standard a bug. 
    However, things should still work, even with the current bogus design of time facets. I'll work on a hack to fix this.
Comment 3 Andrew Pinski 2003-12-05 03:16:06 UTC
Any news on this.
Comment 4 Benjamin Kosnik 2003-12-05 07:00:13 UTC
Subject: Re:  issues with time_get, time_put not being
 tied to correct _punct facet.

>Any news on this.

this is a fatal flaw in the standard. It can be worked around, much like
money_get and money_put are able to be used: by also imbuing the
moneypunct (and here, __timepunct) facet.

You can suspend this if you'd like.

Comment 5 Andrew Pinski 2003-12-19 09:58:03 UTC
Suspending based on Benjamin's comments.