svn commit: r214045 - in stable/8/tools/regression/bin/sh: builtins errors execution expansion set-e

[David O'Brien]

On Sat, Oct 23, 2010 at 11:20:24PM +0200, Jilles Tjoelker wrote:
> On Mon, Oct 18, 2010 at 11:10:32PM +0000, David E. O'Brien wrote:
> > +++ stable/8/tools/regression/bin/sh/builtins/command10.0	Mon Oct 18 23:10:32 2010	(r214045, copy of r204802, head/tools/regression/bin/sh/builtins/command10.0)
...
> > +++ stable/8/tools/regression/bin/sh/builtins/command9.0	Mon Oct 18 23:10:32 2010	(r214045, copy of r204801, head/tools/regression/bin/sh/builtins/command9.0)
[..]
> These work on stable/8, but for the wrong reason. In stable/8, 'command'
> does not allow executing builtin utilities, and therefore 'eval' cannot
> be found. This is a different error than the expected passing of a
> non-number to 'shift', and does not allow testing for the bug this is
> supposed to check for.
[..]
> Because this does not cause any extra failures, I don't think it is
> necessary to revert it. However, it seems unusual to MFC a test for a
> bugfix without also MFCing the bugfix.

I MFCed all passing test cases.  If the test case passes when it should
not, it seems to me the test case isn't providing sufficient coverage.

That said, I would like to MFC the various ash bug fixes that are missing
from 8.1.  I know you've MFC'ed some of the ones in HEAD, but are ones
not MFCed that you feel should be?  Or ones that should not be?  Are the
missing fixes just due to lack of time to MFC?

thanks,
-- 
-- David  (obrien at FreeBSD.org)