[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: PROCINFO["FS"] and other bugs
From: |
Aharon Robbins |
Subject: |
Re: PROCINFO["FS"] and other bugs |
Date: |
Tue, 7 May 2002 17:30:11 +0300 |
Greetings. Re this:
> From: address@hidden (Kenny McCormack)
> Subject: Re: PROCINFO["FS"] and other bugs
> Date: 6 May 2002 11:52:29 -0500
>
> In article <address@hidden>,
> Stepan Kasal <address@hidden> wrote:
> ....
> >I've read the source of gawk-3.1.0. I've created a few patches.
> >Every patch has a comment at its own beginning.
> >
> >gawk-3.1.0-procinfo_fs.patch fixes a bug
>
> I looked at this and I have a few comments:
> 1) What is the point of PROCINFO["FS"]? Isn't it equivalent to
> just checking to see if FIELDWIDTHS is set or not?
No. It's possible for both FS and FIELDWIDTHS to have nonempty
values; PROCINFO["FS"] tells you which method gawk is currently using.
{ ....
FS = # whatever
...
FIELDWIDTHS = # whatever
...
FS = FS # go back to FS spliting
...
FIELDWIDTHS = FIELDWIDTHS # go back to FIELDWIDTHS splitting
...
}
> 2) I think the real problem is that setting the variable via the
> "-v" command line option, doesn't trigger the normal
> re-evaluation. I'm wondering if that wouldn't be the better
> place to fix this.
Nope. During -v processing gawk's state is, shall we say, "fragile".
In particular, as is pointed out in the original note, -v processing
happens *before* PROCINFO is created. Just invoking the set_FS routine
would probably crash gawk or some such.
Arnold