Re: fprintf_ln() is slow

[Jeff King <peff@peff.net>]

On Fri, Jun 28, 2019 at 11:03:27AM +0100, Phillip Wood wrote:

> > I considered that, too, but I think it is safe. stdio has its own
> > locking, so every individual call is atomic. The potentially problematic
> > case would be where we switch back from line buffering to no-buffering,
> > and somebody else has written some content into our stack-based buffer
> > (that is about to go out of scope!). But I'd assume that as part of the
> > switch to no-buffering that any stdio implementation would flush out the
> > buffer that it's detaching from (while under lock). Nothing else makes
> > sense.
> 
> The C standard section 7.19.5.6 says that
>   The setvbuf function may be used only after the stream pointed to by
>   a stream has been associated with an open file and before any other
>   operation (other than an unsuccessful call to setvbuf) is performed
>   on the stream.
> 
> The is a note about the buffer that says
>   The buffer has to have a lifetime at least as great as the open
>   stream, so the stream should be closed before a buffer that has
>   automatic storage duration is deallocated upon block exit.
> 
> So changing the buffer in the way that has been proposed is undefined
> behavior on two counts I think.

Oof, thanks for the reference. That is much less safe than I had
imagined. We used to do this kind of setvbuf() munging in vreportf.
Interestingly, it was in released versions for about 2 years, but I
don't recall anybody complaining (we eventually reverted it to have more
flexibility in sanitizing the results before writing them out).

Anyway, I think we're all agreed that's the wrong approach here.

-Peff