On Mon, 15 Sep 2014 10:28:16 +0100 Georg Rudoy <0xd34df...@gmail.com> wrote:
> Let's limit our sample to Gentoo tree then. How frequently arches list > shrinked as a result of bumping the version (because the upstream has > chosen so, not because of insufficient resources to keep testing all > previously stated arches)? They do not shrink only due to a result of bumping. In the ideal case it is indeed announced and we can do it at bump time; in the less ideal case they rather shrink due to us finding out that a particular major feature does no longer work on an architecture. This could thus also happen due to an incompatibility with one or another dependency for which no further architecture specific code is written, or perhaps due to an architecture specific bug that has arisen. In the ideal case, shrinking wouldn't be a problem; in the less ideal case, having no keywords is preferable to shrinking as continuously testing live ebuilds isn't affordable.
