svn commit: r333351 - head/usr.bin/grep

[Alexey Dokuchaev]

On Tue, May 08, 2018 at 09:36:21AM -0500, Kyle Evans wrote:
> On Tue, May 8, 2018 at 5:58 AM, Alexey Dokuchaev <danfe at freebsd.org> wrote:
> >>
> >> -     if ((f = fopen(fn, "r")) == NULL)
> >> +     if (strcmp(fn, "-") == 0)
> >> +             f = stdin;
> >
> > This makes sense: when `fn' is "-", `f' is stdin.
> >
> >> -     fclose(f);
> >> +     if (strcmp(fn, "-") != 0)
> >> +             fclose(f);
> >
> > But not this one: why are you checking `fn' again?  Shouldn't you
> > fclose(f) if it's not stdin?
> >
> >         if (f != stdin)
> >                 fclose(f);
> >
> 
> You say potato, I say potato. =) In this case, it's low overhead in a
> not particularly performance critical bit and drawing a connection
> between this and the opening of 'f' above in an extremely obvious way.

Well, I'm not worried about the overhead or performance issues, they are
negligible.  I just find second strcmp(fn, "-") to be semantically wrong
(and that's why you need implicit "there's only one way to get stdin here"
assert).  You assign `f' to stdin based on `fn' being "-", but you
fclose(f) when it's not stdin; the value of `fn' is irrelevant this time.
As a nice bonus, you only spell strcmp(fn, "-") once and do not need to
implicitly assert that there's only one way to get stdin here.

> This also might get ripped out soon -- we'll see how things go.

I see, understood.

./danfe