Re: CVS commit: src

[Jonathan Perkin <sketch%netbsd.org@localhost>]

* On 2009-02-28 at 09:54 GMT, Adam Hoka wrote:

> Jonathan Perkin wrote:
> 
> > * On 2009-02-28 at 04:50 GMT, Lubomir Sedlacik wrote:
> > 
> > > On Wed, Feb 25, 2009 at 08:38:30PM +0000, Jonathan Perkin wrote:
> > >
> > > > Module Name:    src
> > > > Committed By:   sketch
> > > > Date:           Wed Feb 25 20:38:30 UTC 2009
> > > > 
> > > > Modified Files:
> > > >         src: Makefile
> > > >         src/etc: Makefile
> > > >         src/usr.sbin/postinstall: postinstall
> > > > 
> > > > Log Message:
> > > > Use awk and grep host tools where required.  'build.sh release'
> > > > now works on Solaris (but only with HOST_CC=/usr/sfw/bin/gcc for
> > > > now).
> > > 
> > > Do you really mean OpenSolaris here (with ksh93 as /bin/sh)?
> > 
> > No, it is assumed build.sh is invoked as '/bin/ksh build.sh' or
> > similar on platforms where the host /bin/sh is not good enough.  Sorry
> > if that wasn't clear.
> > 
> > I'd actually quite like to see build.sh made more portable so this
> > isn't necessary, but I don't know if that would receive widespread
> > agreement.
> 
> Crippling scripts to run on pre posix shells is not portability.

Well, we could argue that back and forth, but my point is that I think

  $ ./build.sh -h
  ./build.sh: bad substitution

is a terrible first experience for users of build.sh on Solaris and they
are likely to just give up right there and then instead of finding out all
the awesome things build.sh can do.

Given the intended purpose of the script is to be as cross-platform as
possible, and the relative ease of making it work on crappy /bin/sh's, I
think the benefits outweight any potential loss of aesthetics.

-- 
Jonathan Perkin                                     The NetBSD Project
http://www.perkin.org.uk/                       http://www.netbsd.org/