I committed the patch in r26917.
On Tue, 02 Oct 2007, Ben Collins-Sussman wrote:
> This patch looks pretty good to me. Can you commit it, and see if it
> resolves the weird behavior that Mark was seeing?
>
> On 10/1/07, Daniel L. Rall <dlr@finemaltcoding.com> wrote:
> > On Mon, 01 Oct 2007, Eric Gillespie wrote:
> >
> > > Jack Repenning <jrepenning@collab.net> writes:
> > >
> > > > On Oct 1, 2007, at 6:32 PM, Eric Gillespie wrote:
> > > >
> > > > > I advocate setting line buffering in
> > > > > svn_cmdline_init, something like:
> > > > >
> > > > > setvbuf(stderr, NULL, _IOLBF, 0);
> > > > > setvbuf(stdout, NULL, _IOLBF, 0);
> > > >
> > > > +1 on stdout. But isn't stderr ordinarily character-buffered, for
> > > > reasons having to do with ensuring you can report program crash even
> > > > if the crash is because you can't allocate a line buffer to report it?
> > >
> > > When connected to a tty, yes; good catch. We should use _IONBF
> > > for stderr.
> >
> > Patch attached. Rather than operating on stderr, svn_cmdline_init()
> > takes an ERROR_STREAM parameter. Return values from setvbuf() appear
> > to be best-off ignored, I suppose.
- application/pgp-signature attachment: stored
Received on Thu Oct 4 01:17:52 2007