[RFC] A QoI problem with filebuf and POD charT != char, wchar_t

[Paolo Carlini]

Paolo Carlini wrote:

> ... well, /almost/ identically, of course: a few sizeof(_CharT) would 
> have
> to be sprinkled here and there.

Hi again.

For concreteness, I'm attaching below a draft patch, already regtested and
lightly tested on a few testcases similar to the posted one.

I briefly recall that the general philosophy is as follows: a codecvt 
facet not
available is basically equivalent to always_noconv. In order to deal 
with both
in the same way we only need to take into account sizeof(char_type). 
When (the
most common case) char_type == char, no computational price is paid since
sizeof(char_type) == 1 and no actual divisions or any other additional code
is really involved.

If no one finds something basically flawed with this approach, I'll 
refine the
patch further, prepare a basic set of new testcases (probably using a 
short type
as char_type), and post the whole in a few days.

Thanks,
Paolo.

/////////////
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch_qoi_filebuf
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20040519/cab40520/attachment.ksh>