Re: sh -e changes broke pkgsrc

[James Chacon <jmc%netbsd.org@localhost>]

On May 26, 2008, at 4:55 PM, David Holland wrote:

> On Mon, May 26, 2008 at 11:41:40PM +0200, Matthias Drochner wrote:
> > > there's another unclear issue. In bin/32282, if I understand
> > > correctly, dsl@ says that -e failures inside a () subshell should
> > > cause the outside shell to exit also
> >
> > A subshell invocation is a command as any other.
>
> So dsl is wrong? (I just want to be as clear as possible, since this
> whole subject is extremely murky.)
>
> > So in the light of what was said before -- if the subshell
> > return status was caught by a "&&" or so, the outside
> > shell should not exit. Otherwise, it should.
>
> Well, no, because a subshell isn't a simple command. The standard says
> that only failing simple commands cause exits... the question is
> whether the failure inside the subshell kills off only the subshell or
> the entire caboodle. (Imagine for a moment that () wasn't implemented
> by forking; I think that's the premise.)

I got into a long discussion with dsl about this when retrofitting  
make a couple of years ago. The shell has 2 notions of sub-commands:

()
{}

We came to the conclusion (as I recall) that -e should transition it's  
behavior inside () for a simple command executed there (and thereby  
setting it's exit status), but otherwise the same rules applied (where  
compound statements were not affected by -e)

James