[Patch] Fix libstdc++/11378 (3rd take)

[Nathan Myers]

On Mon, Jul 14, 2003 at 01:21:55PM +0200, Paolo Carlini wrote:
> Nathan Myers wrote:
> 
> >I don't see the above
> >as a reason to delay applying the patch.  Since the problem is in 
> >the existing code, and doesn't affect 11378, there's no reason not
> >to apply it and close 11378.  (I suppose I should file a bug report
> >detailing this problem.)
> >
> Thanks for bringing up the issue! I don't think it's necessary to file a 
> bug report: I'm already working on the problem, which seems quite clear. 
> Indeed, providing a wrapper for write seems quite simple, that for 
> writev is a little more complex, but I have found around some code to 
> study and adapt to our needs (just google for xwritev ;).

Note that in principle you could avoid the memmove by making filebuf's 
buffer pointer handling a lot more complicated.  However I don't think 
it's necessary to optimize for short-write handling.  A short write 
should be a rare event, I think.

Does anybody know a circumstance where short writes are common enough 
that performance in that case matters?

Nathan Myers
ncm-nospam@cantrip.org