[RFC] Default constructed stringbufs and stringstreams

[Paolo Carlini]

Paolo Carlini wrote:

> ...I gonna show you a patch later today...

I'm attaching it below: besides fixing the problem, it also fixes most of
the spurious regressions that happen if we artificially give a non-zero
capacity to a default-constructed string (more about these issues soon,
of course...)

I like it *so* much and, if you can't spot something wrong with it, I'd
like to commit it later today...

Paolo.

//////////////////
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: CL_stringbuf
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20040929/53f8abb4/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch_stringbuf
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20040929/53f8abb4/attachment-0001.ksh>