On 12/20/2015 01:24 AM, Michał Górny wrote: > On Sat, 19 Dec 2015 15:51:29 -0800 > Zac Medico <zmed...@gentoo.org> wrote: > >> On 12/16/2015 09:49 AM, Michał Górny wrote: >>> On Wed, 16 Dec 2015 14:38:03 +0100 >>> Alexander Berntsen <berna...@gentoo.org> wrote: >>> >>>> -----BEGIN PGP SIGNED MESSAGE----- >>>> Hash: SHA512 >>>> >>>> On 16/12/15 06:33, Zac Medico wrote: >>>>> Disable Manifest "stable mtime" behavior by default, and add a >>>>> corresponding egencache option. >>>> This message tells me nothing about why we need to do this. >>> >>> We need do this because we changed the behavior and the new behavior is >>> counter-intuitive. We already had a number of bugs caused by it, >>> and while it's used by Infra, it's at least unexpected when someone >>> manually runs 'repoman manifest'. >> >> Just because it was historically buggy does not mean that it will always >> be that way. I believe that it will be very safe once we've fixed it to >> include the mtimes of all relevant directories in the max mtime calculation. >> >>> I mean, before this all started 'repoman manifest' just updated >>> the Manifest which meant it's mtime changed. Nowadays, it also sets >>> mtime to some value in the past, which means running 'repoman manifest' >>> may result in updated Manifest having mtime older than the old >>> Manifest. As a result, people using rsync are in trouble. And this has >>> been reported too by overlay owners. >> >> We should get our facts straight. It's not possible for the updated >> Manifest to have an older mtime than the old manifest, because the mtime >> of the old Manfiest is included in the max mtime calculation. > > Unless you remove the old Manifest (and distfiles) to have Portage > refetch the files and recalc the digests.
Removing the Manifest will bump the mtime of its parent directory, and that mtime is included in the max mtime calculation. -- Thanks, Zac