Re: [forum] Re: restest hangs the server (was: Some hard numbers on XFree86 memory usage ...)

On Mon, 2003-04-21 at 14:25, Mark Vojkovich wrote:
> On 21 Apr 2003, Owen Taylor wrote:
> 
> > On Mon, 2003-04-21 at 09:25, Michel Dänzer wrote:
> > > [ please follow up to the devel list only if you agree that this is more
> > > appropriate there ]
> > > 
> > > On Mon, 2003-04-21 at 08:13, Aidan Kehoe wrote:
> > > >  Ar an 19ú lá de mí 4, scríobh Mark Vojkovich :
> > > > 
> > > >  >     http://www.xfree86.org/~mvojkovi/restest.c
> > > > 
> > > > That locks up my X server (except for the mouse cursor) on NetBSD 1.6P,
> > > > btw. 
> > > 
> > > I experienced the same on Linux/PPC, thought it was architecture
> > > specific, but maybe not?
> > > 
> > > > No problem when it's run from a virtual console with DISPLAY set
> > > > appropriately. Where would one go to fix this?
> > > 
> > > The first thing to do would be to find out where it hangs I guess.
> > > Funnily, the server would still respond to the keyboard, at least I
> > > could switch VTs, but it wouldn't draw (everything) anymore.
> > 
> > It doesn't actually hang the server, rather what
> > happens is that it outputs to terminal with a server
> > grab.
> > 
> > So, if you run it in a xterm (etc.), you get a:
> > 
> >  X server blocks on restest
> >  restest blocks on terminal
> >  terminal blocks on X server
> 
>    That doesn't happen with xterm over here.  

It's going to be an intermittent problem; whether it happens
not is going to depend on various buffers sizes (stdio,
terminal, Xlib.) I suspect if you set things up so restest
outputs sufficiently large quantities of text, you'll see it.

I don't see any way that a terminal program could avoid
this problem. Xterm has no way of knowing when it is going to block 
in an Xlib call because the X buffers are full and the server is
grabbed.

Regards,
                                         Owen