On Fri, 2013-04-26 at 09:09 -0400, Jonathan Kamens wrote: > On 04/26/2013 08:58 AM, Cristian Sava wrote: > > Webalizer is enabled via /etc/sysconfig/webalizer but I don't understand > > why /etc/cron.daily/00webalizer has this line: > > > > [ "z$WEBALIZER_CRON" != "zyes" ] && exit 0 > > > > that I think it should be (and it is working this way) > > > > [ "$WEBALIZER_CRON" != "yes" ] && exit 0 > > > > What is the mistery? > > > The two versions you quoted are functionally equivalent on Linux. If > there's a z before the variable expansion and a z before the word > "yes", then the two z's cancel out and don't affect the string > comparison. > > As for why they are there, it's an old shell programmers' trick to > avoid causing syntax errors when doing string comparisons to empty > strings. Let me explain... > > If you do this: > > [ $WEBALIZER_CRON != yes ] > > and the variable WEBALIZER_CRON is empty, then the command above will > result in an error, because when it's evaluated, it will look to the > shell as if there is nothing to the left of the '!=' and therefore > there's a missing expression in the statement. > > This, however, is fine: > > [ "$WEBALIZER_CRON" != yes ] > > because the quotes ensure that the statement will be evaluated with an > expression to the left of the != even if the expression is just an > empty string. > > This is fine too: > > [ z$WEBALIZER_CRON != zyes ] > > because if the variable is empty, the expression to the left will be > "z" rather than an empty string. > > It's unnecessary to use both the quotes and the extra character on > both sides of the expression. I'm not sure, but I think it may have > been necessary to use both at some point in the past, because I think > the middle example above, with just the quotes, might actually have > been buggy in some old versions of the "test" program that evaluates > such expressions. This caused shell programmers to get into the habit > of using both the quotes and the extra character. I doubt it's been > necessary to do that for many years, though for all I know webalizer > may support some UNIX versions that are so old that they still have > problems in this area. Given that possibility, there's certainly no > harm in writing the test the way it's written in the file you looked > at. > > jik > Thank you very much for your explanation, C.Sava -- test mailing list test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
