Re: nits to please a kernel hacker

[Theodore Tso <tytso@mit.edu>]

On Mon, Sep 29, 2008 at 04:01:07PM -0700, Roland McGrath wrote:
> 	sh: /usr/local/bin/staprun: No such file or directory
> 	Pass 5: run completed in 0usr/10sys/11real ms.
> 	Pass 5: run failed.  Try again with more '-v' (verbose) options.
> 	Running rm -rf /tmp/stapXpCOU6
> 	-bash-3.2$ 
> 
> Ok, there is finally.  Two -v's just to see the command line it tried to run?
> Really?  Even though you deserved it for that smarmy damn error message,
> I'm feeling a little bad about the tire-iron.  So, I'm working with ya!

Note that run stap, "try again with more '-v'", rinse, repeat loop is
really annoying not just to kernel developers, but also to System
Administrators.  (Especially ones that were dragged kicking and
screaming to Linux and are longing for Dtrace.)

Stupid question, is there a reason than messages from stderr are
filtered out unless a sufficiently high -v is given?  I can see
wanting to filter out warning messages, but are there normally lots of
error messages sent out to stderr that need to be filtered?

Here's a potentially stupid suggestion --- why not pipe stdout and
stderr into a perl script which is resposible for doing the filtering?
With a perl script it might be possible to do more intelligent
filtering based on regexp's and then translate some very
unintelligible error messages from the compiler, linker, or whatever
into more user-friendly error messages.

						- Ted