Re: Gawk match() strange behaviour

[Alain Ketterlin]

Hi, thanks for your help.

> > The following program:
> >
> > {
> >      r = match($0,/^ */,t);
> >      print "R=" r " S=" RSTART " L=" RLENGTH;
> > }
> >
> > produces this (< signals  input, > signals output)
> > <
> > > R=-1208966831 S=-1208966831 L=1208966850
> > < random
> > > R=1 S=1 L=34
> > <  random
> > > R=1 S=1 L=2

> I could not reproduce this using either stock gawk 3.1.5 or the current CVS
> sources.  I suggest that you try building from scratch from the CVS archive
> on savannah.gnu.org.
>
> For the empty line I get
>
>         R=1 S=1 L=0

Things are getting strange (for me, I mean :). I just noticed that
the locale has an impact.

With gawk-3.1.5 (compiled from the tarball), under en_US.utf-8 I get:
-from an empty line: R=1 S=1 L=18
-from a line containing "random" (no space at beginning): R=1 S=1 L=34
-from "  random" (two spaces at beginning): R=1 S=1 l=2 (correct)
Under en_US.iso-8859-1, everything is ok. So it seems that utf-8
input is the problem.

With gawk-stable checked out from savannah, everything is correct,
under both locales.

Thanks for your help.

-- Alain.