Re: fix stage17 regression

Eric Blake <ebb9 <at> byu.net> writes:

> 
> The actual code fix adds a single conditional of when not to expand an argv 
> wrapper onto the argument collection stack.  The bulk of this patch is making 
> sure we never regress on this tricky corner case again.

It turns out the regression was in stage 18, not 17.  Andreas' report from git 
bisect was off by one commit.

-- 
Eric Blake