[OE-core] [CONSOLIDATED PULL 00/40] Review and ACK

[Phil Blundell]

On Thu, 2013-07-18 at 12:01 +0100, Paul Eggleton wrote:
> On Thursday 18 July 2013 11:52:22 Burton, Ross wrote:
> > On 18 July 2013 07:34, Saul Wold <sgw at linux.intel.com> wrote:
> > >   valgrind: added perl dependency
> > 
> > I presume the situation here is that a target perl wasn't actually
> > built, so adding an explicit runtime dependency causes it to be built
> > and therefore available in the feed for rpm to find.
> > 
> > To me this says that the #!-to-dependency magic from rpmdeps isn't
> > really useful, as we then have to go and put explicit dependencies
> > back in to ensure the requirements are actually built.
> 
> I have to say I too have wondered this. Perhaps these would be better 
> implemented as QA warnings (that could be defaulted to errors) rather than 
> just silently adding the dependencies.

Seeing this discussion makes me wonder: if the target perl hasn't
actually been built at this point, how does rpmdeps know what package to
add a dependency on in the first place?  Does it just have some random
hard-coded list somewhere?

p.