On Sat, 15 Sep 2018 16:27:28 +0200
Michał Górny <mgorny@gentoo.org> wrote:

> On Fri, 2018-09-14 at 23:44 +0100, James Le Cuirot wrote:
> > There's no need for two separate sed calls here.
> > ---
> >  eclass/python-utils-r1.eclass | 6 +-----
> >  1 file changed, 1 insertion(+), 5 deletions(-)
> > 
> > diff --git a/eclass/python-utils-r1.eclass b/eclass/python-utils-r1.eclass
> > index e3cf82b4b58f..121f2382ba78 100644
> > --- a/eclass/python-utils-r1.eclass
> > +++ b/eclass/python-utils-r1.eclass
> > @@ -1247,11 +1247,7 @@ python_fix_shebang() {
> >  			if [[ ! ${error} ]]; then
> >  				# We either want to match ${from} followed by space
> >  				# or at end-of-string.
> > -				if [[ ${shebang} == *${from}" "* ]]; then
> > -					sed -i -e "1s:${from} :${EPYTHON} :" "${f}" || die
> > -				else
> > -					sed -i -e "1s:${from}$:${EPYTHON}:" "${f}" || die
> > -				fi
> > +				sed -i -e "1s:${from}\( \|\$\):${EPYTHON}\1:" "${f}" || die  
> 
> Hmm, I wonder if we could just utilize '\b' here.

I also considered it but that would match /, which is no good. For
example, it would replace python in /usr/python/foo.

I am now working on a new approach to better accommodate my cross stuff
that would supersede this anyway but it's not ready yet.

> >  				any_fixed=1
> >  			else
> >  				eerror "The file has incompatible shebang:"  

-- 
James Le Cuirot (chewi)
Gentoo Linux Developer