On 2013/09/21 13:12, Craig R. Skinner wrote: > As the others here with brains have had a chance to sleep on this, > what's the current thinking? > > As I understand it, there are 2 decisions to make: > > 1) How to decide if a $daemon is a script as opposed to a binary > (*) file(1)
too big, too complex. file(1) has to parse a 500K /etc/magic each time it runs, which would be 15+ times for this, and uses about 8MB ram. definitely overkill for identifying the contents of the first two characters of a file. remember this has to run on a wide range of platforms. > (*) dd(d) > (*) sed(1) > Could stat(1) be tasked to switch case on file attributes (e.g: size)? how would size help? > > 2) Whether to check if a script's interpreter is valid > > http://openbsd.7691.n7.nabble.com/etc-rc-d-rc-subr-prefix-pexp-with-script-interpretor-path-td234439.html > > Yes/No/Other? > -- > Craig Skinner | http://twitter.com/Craig_Skinner | http://linkd.in/yGqkv7 > How much does this really help anyway? Porters still need to check the actual string used (for example, it's quite common for perl things to use "perl: <programnanme>"), and it doesn't take much extra typing to add a pexp line to the script in cases where it's needed. I am rather wary of adding complication to rc.subr, iirc one of the conditions when the subsystem was first added was that it remained simple..
