On Tuesday 23 July 2002 6:16 am, Cliff Wright wrote:
> On Tue, 23 Jul 2002 14:42:17 +1000
> "Richard Grace" <rgrace@aapt.com.au> wrote:
> 
>   
> > Would one get better performance of all packages requiring
> > PTH if this option is enabled?
> > 
> > Is it worth doing ``make udate'' to remedy this?
> 
> It is only an issue when using pthread compatible routines if the pth I/O
> routines are used directly their is no problem. I ran in to this when
> compiling the milter library for sendmail. It creates a thread to handle
> signals not normally seen, and then put that thread to sleep, which had the
> effect of putting the whole task to sleep, not allowing any normal thread to
> run. So the effect was not a performance issue, it actually broke my program
> that was using the milter library.
>
> Also I tried "--enable-syscall-hard" but the install test indicated this was
> broken.

The pth-syscall package has this option enabled. One day we hope to enable 
this option on the default pth package and remove pth-syscall

The test is non --enable-syscall-hard friendly. 

Nick