James Carlson wrote: > Nicolas Williams writes: > > On Sun, Jan 04, 2009 at 11:26:52PM +0100, Roland Mainz wrote: > > > ... does anyone have suggestions for which scripts a compilation would > > > be usefull ? > > > > You could analyze boot performance to find the SMF start method scripts > > that delay the most dependents by the most time. Those would be the > > scripts to target first. > > I think that's assuming, of course, that the delay is due to active > shell interpretation of the script rather than the work being done by > the executables invoked by the script (or deliberate sleep(1) > invocations). Otherwise, if active shell work isn't the issue, then > that analysis won't reveal useful bits. > > Do we have _any_ boot time scripts that do any non-trivial work at the > shell level? Which ones?
That's one of my thoughts behind the question (I've seen possible suspects in the networking area and the upcoming firewall shell scripts are half a |fork()|-bomb (yes, I complained about that during code review, right now I am drafing-up some dtrace-based shell profiling tools to get some data where improvements can easily be done)) ... > It probably won't help you much, but I'd expect class action scripts > (used during package install/upgrade) to be where a lot of the > script-level work gets done, and even there, it's usually something > like nawk that does the heavy lifting, not /sbin/sh. Right... but the reason for that was simply that "awk" is only available when /usr is being mounted and the old Bourne shell is very limited when someone wishes to do some complex tasks. > And > (unfortunately for this effort), that's mostly being discarded due to > IPS. Slightly offtopic: Which is IMO an IPS design issue - as I said earlier IPS _could_ support shell scripting quite easily without breaking portabilty, performance, package delivery or the current IPS state machine (the problem is that the current Sys5 package scripts can randomly create/modify/move/delete files - if this capabilty is being "tracked"/"controlled" by the IPS system then scripting for IPS would be implementable). > Another *huge* user of shell scripting is LU. But that's closed > source and on a dead-end trip as well. Well, one of the ideas behind shipping "shcomp" was to allow shipping closed-source shell scripts without shipping the script in a plain form. ---- Bye, Roland -- __ . . __ (o.\ \/ /.o) roland.mainz at nrubsig.org \__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer /O /==\ O\ TEL +49 641 3992797 (;O/ \/ \O;)
