On Wed, May 2, 2012 at 5:25 AM, Allan McRae <allan@archlinux.org> wrote:
> If --with-scriptlet-shell is specified, sylimk the shell to /bin/sh
> in order for the "make check" to pass.
>
> Signed-off-by: Allan McRae <allan@archlinux.org>
Hmm. I'm not a fan of doing this in the sed/.in style at all; namely
because pactest can be used to test the installed binary on the system
as well as the one in the tree. Thus, tailoring it to the configure
options without a way of undoing that seems like bad news.
I'd much rather just add a new option to the invocation of pactest
(--scriptlet-shell= or something) and add that to our one automated
invocation in the toplevel Makefile.in; the substitution could thus be
performed there instead.
Finally, do we need to symlink it, or can we simply copy in only that
binary (derferencing symlinks, of course)?
On 02/05/12 22:17, Dan McGee wrote:
> On Wed, May 2, 2012 at 5:25 AM, Allan McRae <allan@archlinux.org> wrote:
>> If --with-scriptlet-shell is specified, sylimk the shell to /bin/sh
>> in order for the "make check" to pass.
>>
>> Signed-off-by: Allan McRae <allan@archlinux.org>
>
> Hmm. I'm not a fan of doing this in the sed/.in style at all; namely
> because pactest can be used to test the installed binary on the system
> as well as the one in the tree. Thus, tailoring it to the configure
> options without a way of undoing that seems like bad news.
>
> I'd much rather just add a new option to the invocation of pactest
> (--scriptlet-shell= or something) and add that to our one automated
> invocation in the toplevel Makefile.in; the substitution could thus be
> performed there instead.
OK, sounds easy enough to do.
> Finally, do we need to symlink it, or can we simply copy in only that
> binary (derferencing symlinks, of course)?
The only thing that could go wrong is a shell being used that does not
support "echo". I'm not sure if that is even possible, but "python -c"
works so...