[RFC, trunk] What about removing _M_buf_size_opt?

[Paolo Carlini]

Hi,

while working on the issues which I mentioned earlier today, it
occurred to me that perhaps for trunk we can remove _M_buf_size_opt
completely!

In the spirit of Nathan's maxim indeed: "... cut and _cut_ and *cut*!"

As already pointed out by Benjamin some time ago (0 or 1 in the
unbuffered case?!?) its very semantics its quite confused: this is at
the root of the bug which I fixed for 3.3 in the previous posting.

I'm attaching here the preliminary patch which I have successfully
regression tested.

Opinions?

Paolo.

/////////
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch_size_opt
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20030329/ba6e6363/attachment.ksh>