[Patch] Partial fix for libstdc++/17038

[Paolo Carlini]

Hi,

this is only a partial fix for mainline and 3_4: in my opinion the problem
cannot be really solved without breaking the API/ABI: __timepunct::_M_put
must return an integer.
Anyway, this is some preliminary, useful, work:
 1- A testcase needing 72 chars, vs 64, the current size of the buffer: I
    have enlarged the latter to 128.
 2- Minimal, but really needed error checking: if strftime/wcsftime
    fails it is compulsory to null terminate __s anyway since later
    we are calling traits::length on it!
 3- A FIXME comment about __timepunct::_M_put ;)

Tested x86/x86_64-linux, gnu/generic locale models.

Paolo.

/////////////////
       
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: CL_17038_partial
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20040825/02016506/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch_17038_partial
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20040825/02016506/attachment-0001.ksh>