Replay progress report

Eric S. Raymond esr at
Thu Jan 7 11:31:14 UTC 2016

Hal Murray <hmurray at>:
> We should start a collection of debugging routines to print out things like 
> the list of peers.  The idea is to be able to call it from gdb.  In this 
> case, you would have to run your test case in non-replay mode to see what it 
> should look like, then compare with what it says when replay doesn't work.

This is *exactly* what I am working on. Or, rather, was working on before
I got interrupted by urgent GPSD bug reports and concluded I needed to ship
a release. I expect to be back on NTPsec later today.

> > Any time call that is in pure I/O is not captured, precisely in order to
> > avoid problems on replay. 
> What does "pure I/O" mean?

The part of the main select that replay bypasses - that isn't receive() or
the intercepted functions beneath it.  Corresponds closely to what's
in ntp_io.c.
		<a href="">Eric S. Raymond</a>

More information about the devel mailing list