When I run emerge-webrsync with FEATURES="webrsync-gpg", I got this non-critical error message: tar: portage/metadata/.checksum-test-marker: timestamp 2017-06-29 21:25:21 is 45313578.601960136 s in future. Whole output: * Running emerge-webrsync Fetching most recent snapshot ... Trying to retrieve 20160121 snapshot from http://distfiles.gentoo.org ... Fetching file portage-20160121.tar.xz.md5sum ... Fetching file portage-20160121.tar.xz.gpgsig ... Fetching file portage-20160121.tar.xz ... Checking digest ... Checking signature ... gpg: VAROVÁNÍ: přístupová práva pro domovský adresáře nejsou bezpečná „/var/lib/gentoo/gkeys/keyrings/gentoo/release“ gpg: Podpis vytvořen Pá 22. leden 2016, 01:57:53 CET pomocí klíče RSA s ID uživatele C9189250 gpg: Dobrý podpis od „Gentoo Portage Snapshot Signing Key (Automated Signing Key)“ [neznámý formát] gpg: VAROVÁNÍ: Tento klíč není certifikován důvěryhodným podpisem! gpg: Nic nenaznačuje tomu, že tento podpis patří vlastníkovi klíče. Otisk primárního klíče: DCD0 5B71 EAB9 4199 527F 44AC DB6B 8C1F 96D8 BF6D Otisk podklíče: E1D6 ABB6 3BFC FB4B A02F DF1C EC59 0EEA C918 9250 Getting snapshot timestamp ... Syncing local tree ... tar: portage/metadata/.checksum-test-marker: timestamp 2017-06-29 21:25:21 is 45313578.601960136 s in future
Not an "error message". The file is put there on purpose to ensure mirrors update properly.
So remove it from the emerge-webrsync tarballs or whatever. Having random nonfatal errors printed during normal operation is not a good thing.
On the portage side, we can add --warning=no-timestamp to the tar options.
There's a patch in the following branch: https://github.com/zmedico/portage/tree/bug_572578 I've posted it for review here: https://archives.gentoo.org/gentoo-portage-dev/message/0cb1a34e5e6d980535234a8e5e757079
I have set this timestamp marker to be no longer in the future, but LATER than the original creation of the file. Old, far-future: Thu Jun 29 19:25:21 UTC 2017 (epoch 1498764321) New, recent-past: Sun Jan 17 03:06:40 UTC 2016 (epoch 1453000000) Any OTHER mtime on the file should be considered an error (as the mirror is not propogating mtime changes) Inside the file, there is also a timestamp; if it is NOT recent, then the mirror is not propogating changes based on checksums).
