time_get::date_order() should return actually date order taken from locale

• Key: STDCXX-459
• Type: Improvement
• Status: Open
• Priority: Minor
• Assignee: Unassigned
• Reporter: Farid Zaripov
• Component/s: 22. Localization
• Affects Version/s: 4.1.2, 4.1.3, 4.1.4
• Fix Version/s: 4.3.0

Description

The current implementation of the time_get::date_order() always returns no_order. It should return date order used by a locale facet (http://www.mail-archive.com/[e-mail]/msg03734.html).

Comments

[Farid Zaripov]

The proposed patch attached.

[Farid Zaripov]

Another option might be extend the struct _rw_time_t with d_order_off field and determine the dateorder value (or two values for narrow and wide characters) in __rw::_rw_get_timepunct().

[Farid Zaripov]

The updated patch is attached.

[Farid Zaripov]

ChangeLog:

STDCXX-459

• include/loc/time_get.h (_rw_get_timepunct): Removed unnecessary declaration of the function.
  (do_date_order): Function outlined and moved from here...
• include/loc/_time_get.cc (do_date_order):... to here. Used __rw_get_dateorder() to obtain the dateorder value.
  (__rw_get_dateorder): Added declaration of the new function.
  (__rw_get_timepunct): Removed unnecessary _RWSTD_EXPORT attribute.
• src/time_get.cpp (__rw_get_dateorder): Added new function to obtain the dateorder value.
• src/time_put.cpp (__rw_get_timepunct): Removed unnecessary _RWSTD_EXPORT attribute.

[Martin Sebor]

The patch looks good to me in general but I have a few compatibility concerns with it:

1. It removes RWSTD_EXPORT from __rw::_rw_get_timepunct(). Is this forward binary compatible on Windows?

2. It adds a new function, _rw::_rw_get_dateorder(). Is this forward binary compatible?

3. It outlines the virtual member function time_get::do_date_order(). Is this binary compatible?

[Farid Zaripov]

1. This change is not necessary (and even more, it's incorrect, because __rw_get_timepunct() should be exported if the user instantiates the std::time_get<user_char, user_iterator>).

2. The __rw_get_dateorder() should also be _RWSTD_EXPORT to be accessible from std::time_get<user_char, user_iterator>. And this change is not a forward binary compatible.

3. I think yes. The outlining doesn't add's the new symbol in the library. The only what may change - it's a return value from the date_order(). But since the all possible returning values are valid (the user of the library shouldn't rely on the always returned no_order from date_order()) this change is safe.

[Farid Zaripov]

Scheduled for 4.3.