Re: Why 'statement may have no effect' lint warning here?

[Wolfgang Laun]

It's not so simple. See below. Is the warning a bug or not? No side effect
- no effect.

function f(x){
   a[1] = "hooey"
}

BEGIN {
i = 1
f(i) == 47 ## Warning here
print a[1]
}
-W


On Tue, 13 Apr 2021 at 10:58, <arnold@skeeve.com> wrote:

> Hi.
>
> Arkadiusz Drabczyk <arkadiusz@drabczyk.org> wrote:
>
> > > > The following code produces 'statement may have no effect' when run
> > > > with --lint:
> > > >
> > > > BEGIN {
> > > >     var = 1
> > > >     var == 1 && var = 2
> > > >     print "var ==", var
> > > > }
> > >
> > > I think that it's a bug that it does so.
> >
> > ok, so just to make sure - is this an actual bug?
>
> Yes. Because:
>
> > > The intent for that warning is soemthing like:
> >
> > > BEGIN {
> > >     var = 1
> > >     var == 2        # no effect
> > > }
>
> > I see why lint would warn me in that case but I still don't understand
> > that part:
> >
> > > See above; if `var == 1' is used standalone as a statement,
> > > it has no effect -- nothing is changed as result.
> >
> > ok, nothing is changed, that is value returned by == is not saved
> > anywhere but it's an expression, right? == operator returns either 0
> > or 1.
>
> Yes, but because the result is thrown away and nothing done with it,
> the statement has no effect: no variables are changed, no 'if' or
> loop statement is affected, etc.
>
> > That
> >
> >     var = 1
> >     var == 2 && var = 3
> >     print "var ==", var
> >
> > would print 1 because var == 2 is not true and `var = 3' would not be
> > executed. So `var == 2' has an effect because it effects the other
> > side of &&.
>
> And that's why I said that the current warning is a bug.
>
> > BTW (or maybe not?) the original snippet I posted cannot be reproduced
> > with Busybox implementation of awk but works with mawk and FreeBSD
> > awk. So maybe doing things like `var == 2 && var = 3' is not specified
> > by POSIX and cannot be expected to work?
>
> You can expect it to work. It's perfectly fine. It sounds to me like
> a bug in Busybox awk.
>
> > Subject: Re: Why 'statement may have no effect' lint warning here?
> > To: Arkadiusz Drabczyk <arkadiusz@drabczyk.org>
> > From: Manuel Collado <mcollado2011@gmail.com>
> > Date: Tue, 13 Apr 2021 09:45:48 +0200
> >
> > [ .... ]
> >
> > The lint warning is 'statement may have no effect'. The key part is the
> > 'may' word. Just a hint that in some cases what the programmer wrote was
> > not what (s)he wanted.
> >
> > A more elaborate example:
> >
> > $ cat no-effect.awk
> > BEGIN {
> >      var = "abcd"
> >      sub("a", "x", var) && sub("c", "y", var)
> >      print "var ==", var
> >
> >      var = "abcd"
> >      sub("e", "x", var) && sub("c", "y", var)
> >      print "var ==", var
> > }
> >
> > $ gawk --lint -f no-effect.awk
> > gawk: no-effect.awk:3: warning: statement may have no effect
> > gawk: no-effect.awk:7: warning: statement may have no effect
> > var == xbyd
> > var == abcd
>
> I disagree that the warning here is valid. At least one of the
> sub() calls will execute, so it has an effect.
>
> I am working to improve the lint warnings. I will  post a patch soon.
>
> Thanks,
>
> Arnold
>
>