On Fri, Dec 5, 2014 at 11:12 PM, Adam D. Barratt <a...@adam-barratt.org.uk> wrote:
> On Fri, 2014-12-05 at 22:56 +0100, Michael Hanke wrote: > > The mixture of in-place modification at runtime and patches has proven > > to be unreliable. Hence to move towards putting as much > > as possible into patches. > > Right, but the section I mentioned, which is removed in the patch, was > only added in the -4 package I previously unblocked. So removing it > again with no explanation seems slightly odd. > Even in -4 it wasn't really added. If you inspect the long sed expression that was removed in -4 (which, in combination with a changed configuration file, was the cause for #769100), you'll see that all configuration variable are set to identical values. The section was "added" as the actual configuration file had no corresponding variables to modify anymore. > (Looking at it, yes the items it contains are indeed in the other patch, > but that just makes things even more confusing. Why was it added in the > first place, and why isn't the removal documented?) > I hope I answered why it was "added". As for why it was "removed" or turned into a patch: Only for having the entire configuration setup in a patch (as in-place modification is apparently fragile). As the freeze policy states that only the diff between the version in testing and the one to be unblocked is relevant, I went for this modification in order to have the end result less complicated. Granted that the way towards it was a bit convoluted. I am sorry that you had to witness it that closely. In any case, this may be academic, as condor appears to have picked up a > dependency on the new version of globus-io, which is still blocked. > That is unfortunate. I hope for the best. thanks in any case, Michael
