Bug 1021583

I realize that the current code uses "chmod -R" to minimize the number of calls, but here's some data:

[agrimm@dhcp129-30 ~]$ for x in $( seq 1 10000 ); do touch file$x; done
[agrimm@dhcp129-30 ~]$ time for x in $( seq 1 10000 ); do chmod +x file$x; done

real	0m4.576s
user	0m0.268s
sys	0m0.896s

Real time is about the same for the shell if you stat every file:

[agrimm@dhcp129-30 ~]$ time for x in $( seq 1 10000 ); do if [ -x file$x ]; then chmod +x file$x; fi; done

real	0m4.776s
user	0m0.357s
sys	0m0.962s

It gets much faster if you don't have to chmod:

[agrimm@dhcp129-30 ~]$ time for x in $( seq 1 10000 ); do if [ \! -x file$x ]; then chmod +x file$x; fi; done

real	0m0.085s
user	0m0.072s
sys	0m0.011s


And just for good measure, this measures a double loop with some math, so that we can attempt to count the cost of the extra shell loop that we'd be introducing:

[agrimm@dhcp129-30 ~]$  time for x in $( seq 0 99 ); do for y in $( seq 1 100 );  do z=$(( 100*x + $y )); if [ \! -x file$z ]; then chmod +x file$z; fi; done; done

real	0m0.256s
user	0m0.146s
sys	0m0.041s

This isn't quite as scientific as we can get, but knowing the overhead of executing a process on a busy node, I'm pretty convinced that the "-x" test is worth it.

One more test:

first script: one thousand lines of:
chmod +x file1

second script: one thousand copies of:
for x in file{1,2,3,4,5}; do
	if [ \! -x $x ]; then
		chmod +x $x
	fi
done

The first take about 0.44 seconds on my machine, while the second takes about 0.065.

Commit pushed to master at https://github.com/openshift/origin-server

https://github.com/openshift/origin-server/commit/2bc19d0fc0f47263108d7f09b4c95b4b19cddacf
Bug 1021583 - Stop chmod'ing all cron scripts

* stat script files before chmod'ing

Tested on devenv_3927.
The chmod code is changed.
It saves much time to check the mode than directly +x.
[walter@dhcp-10-238 abc]$ time for x in `seq 1 10000`; do chmod +x file$x ; done

real	0m9.293s
user	0m0.626s
sys	0m1.853s

[walter@dhcp-10-238 abc]$ time for x in `seq 1 10000`; do [ -x file$x ] ; done

real	0m0.099s
user	0m0.081s
sys	0m0.017s

It will obviously reduce the time when lots of the files already have an "x" mode.

Mark the bug as verified.