On Thu, Dec 18, 2003 at 01:21:41AM -0800, Simon J. Gerraty wrote:
> >The main reason for new fields and doing it this way was to avoid gutting
> >the code entirely as far as spec'ing a shell and instead hard coding /bin/sh. 
> 
> Keeping the ability to spec a shell is good - Solaris loses unless you
> use ksh for example.

Or xpg4/bin/sh (which is what we suggest in the docs)

> 
> >The main changes have to deal with printing out the command being run vs the 
> >actual command (i.e. so the user doesn't have to see the {} || exit stuff).
> >At this point '|| exit $?' vs '&&' comes out the same except that the exit's
> >are easier to follow in most cases when debugging the shell input script.
> 
> Does -dx show the actual commands executed?

Yes

> 
> >Concerns/comments welcome. Otherwise I'll probably commit these in the next
> >few days barring objections.
> 
> Since we've established that -e just doesn't work (for what make
> wants) - why not remove it all together?  Also why leave the -e's in
> the ksh spec?

Potentially a shell could provide something equivilant and for some versions of
sh/ksh on systems it does work this way. So, it still gives people the option
more than anything. As far as the ksh spec, I'll just make it look like the
sh spec before I commit.

> 
> While you are there - would anyone ever pick csh as the shell to use? ;-)

I kept wondering myself...

James