Re: Login time

[Federico Mena Quintero <federico ximian com>]

El mi�11-04-2007 a las 18:52 +0200, Soeren Sandmann escribi� It will depend on the number of system calls per time of course, but
> my experience is that strace times should not be trusted _at all_. In
> one case I saw a factor of 6 difference between running with strace
> and running without.

Hmm, interesting.

What I like about strace is that:

- It makes it easy to monitor all the session's processes [1].  Adding
checkpoints to application code to look for interactions is then easy.
This is how I found the mess with bonobo-activation-server a while back.

- It's easy to see when a process goes to sleep, waiting for another one
to respond.  Then you see the other process wake up, process the
request, reply, and go to sleep.  Then you see the original process
waking up from the poll()/read()/whatever.  This is what made me want to
figure out how much time we spend in roundtrips to gconfd.

- Along with the checkpoints, you get the usual embarrassment of
excessive syscalls, which is always fun to look at :)

[1] Caveat: glibc or the kernel are buggy and it deadlocks if you ptrace
a program which calls setuid().  I had to #ifdef the calls to seteuid()
in gnome-vfs to make strace and gdb work.

  Federico