On Fri, Aug 21, 2015 at 10:05 AM, Yann Ylavic <[email protected]> wrote:
> On Fri, Aug 21, 2015 at 4:36 PM, Ruediger Pluem <[email protected]> wrote: > > > > > > On 08/21/2015 04:31 PM, Yann Ylavic wrote: > >> On Fri, Aug 21, 2015 at 4:29 PM, Yann Ylavic <[email protected]> > wrote: > >>> On Fri, Aug 21, 2015 at 4:25 PM, Ruediger Pluem <[email protected]> > wrote: > >>>> Wasn't it intended for trunk to change the default behavior? > >>> > >>> Yes it was, still when inherit_before is set we need to do that. > >> > >> I meant, we need to do "apr_array_append(p, over->patterns, > >> base->patterns)" and not the inverse. > >> > > > > But apr_array_append(p, over->patterns, base->patterns) means that we do > NOT inherit the stuff from the base > > configuration before the configuration of our section. So the code > before your recent patch was correct. > > Hm, you are right, I was misleaded by the comment in trunk: > /* SubstituteInheritBefore was the default behavior until 2.5.x, ... */ > since SubstituteInheritBefore was *not* the default behavior until 2.5.x. > > To recap, apr_array_append(p, over->patterns, base->patterns) is the > legacy but it is obviously not the inherited base which is merged > before... > > So to keep the existing/prior-to-r1697002 logic in the code (which was > good), I propose to revert r1697002 and rename SubstituteInheritBefore > to SubstituteInheritAfter, WDYT? > It's a toggle, please leave it alone. The default state will be one case in 2.4.x, and a different case in 2.5.x and later, but a user who puts that directive into their config will observe the same behavior when they run their config under 2.4 or later under 2.6.
