Idiot me has been wasting my time (months actually) bothering upstream, when it turns out the problem is caused from a gentoo patch. As verified by upstream, weekly rotation is SUPPOSED (on a computer running continuously) to rotate logs on Sunday, and then every Sunday after that (assuming logrotate is run every day). The patch logrotate-x.y.z-weekly.patch breaks that behavior. It rotates logs on random days (on one computer I had five different log files getting rotated on four DIFFERENT days of the week). I have recompiled logrotate without this patch, and it works as intended (the next Sunday, all of the files got rotated, and they have rotated successfully since then). Please remove this patch from the ebuild. My apologies again to all of the upstream developers I have bothered about this over the past several months.
Which version of logrotate is affected?
(In reply to comment #1) > Which version of logrotate is affected? logrotate-3.7.1-r2, logrotate-3.7.7, and logrotate-3.7.8. $ grep weekly /usr/portage/app-admin/logrotate/logrotate-3.7.* /usr/portage/app-admin/logrotate/logrotate-3.7.1-r2.ebuild: epatch "${FILESDIR}"/${P}-weekly.patch /usr/portage/app-admin/logrotate/logrotate-3.7.7.ebuild: epatch "${FILESDIR}"/${P}-weekly.patch /usr/portage/app-admin/logrotate/logrotate-3.7.8.ebuild: "${FILESDIR}"/${PN}-3.7.7-weekly.patch \
The bug in 107906 is a much worse bug, so I'm not going to remove the patch unless that one is fixed. Unfortunately, I'm completely swamped right now, so I won't get to it for at least a week. If you can verify that the bug in 107906 is no longer an issue, I'll remove the patch immediately.
(In reply to comment #3) > The bug in 107906 is a much worse bug, so I'm not going to remove the patch > unless that one is fixed. Unfortunately, I'm completely swamped right now, so > I won't get to it for at least a week. If you can verify that the bug in > 107906 is no longer an issue, I'll remove the patch immediately. Three comments. 1. I have no idea what the submitter of bug 107906 was talking about (did anyone attempt to duplicate this bug?). 2. A bug that did what he claimed would have been noticed elsewhere (every distribution uses logrotate, and no one else uses a patch like that). 3. Logrotate works fine as it is supposed to without the patch. One one machine, for example: ls -al /var/log/cron* /var/log/messages* -rw------- 1 root root 35549 2010-08-03 04:30:01 /var/log/cron -rw------- 1 root root 238 2010-07-14 03:10:01 /var/log/cron-20100714.gz -rw------- 1 root root 5049 2010-07-18 03:10:01 /var/log/cron-20100718.gz -rw------- 1 root root 8551 2010-07-25 03:10:03 /var/log/cron-20100725.gz -rw------- 1 root root 8383 2010-08-01 03:10:03 /var/log/cron-20100801.gz -rw------- 1 root root 106856 2010-08-02 17:44:59 /var/log/messages -rw------- 1 root root 39024 2010-07-12 03:10:02 /var/log/messages-20100712.gz -rw------- 1 root root 44053 2010-07-18 03:10:01 /var/log/messages-20100718.gz -rw------- 1 root root 54232 2010-07-25 03:10:02 /var/log/messages-20100725.gz -rw------- 1 root root 63435 2010-08-01 03:10:02 /var/log/messages-20100801.gz You can see when I re-installed it without the patch. Both sets of logs went to Sunday rotation (as they are supposed to) and have been rotated each Sunday afterwards (as they are supposed to).
Hello Daniel can we get a (maybe masked) revbump for this to give it a try? Debugging things happening a weekly basis is quite annoying. Thanks, Michael
(In reply to comment #4) > (In reply to comment #3) > > The bug in 107906 is a much worse bug, so I'm not going to remove the patch > > unless that one is fixed. Unfortunately, I'm completely swamped right now, so > > I won't get to it for at least a week. If you can verify that the bug in > > 107906 is no longer an issue, I'll remove the patch immediately. > > Three comments. > > 1. I have no idea what the submitter of bug 107906 was talking about (did > anyone attempt to duplicate this bug?). > > 2. A bug that did what he claimed would have been noticed elsewhere (every > distribution uses logrotate, and no one else uses a patch like that). > > 3. Logrotate works fine as it is supposed to without the patch. > > One one machine, for example: > > ls -al /var/log/cron* /var/log/messages* > -rw------- 1 root root 35549 2010-08-03 04:30:01 /var/log/cron > -rw------- 1 root root 238 2010-07-14 03:10:01 /var/log/cron-20100714.gz > -rw------- 1 root root 5049 2010-07-18 03:10:01 /var/log/cron-20100718.gz > -rw------- 1 root root 8551 2010-07-25 03:10:03 /var/log/cron-20100725.gz > -rw------- 1 root root 8383 2010-08-01 03:10:03 /var/log/cron-20100801.gz > -rw------- 1 root root 106856 2010-08-02 17:44:59 /var/log/messages > -rw------- 1 root root 39024 2010-07-12 03:10:02 /var/log/messages-20100712.gz > -rw------- 1 root root 44053 2010-07-18 03:10:01 /var/log/messages-20100718.gz > -rw------- 1 root root 54232 2010-07-25 03:10:02 /var/log/messages-20100725.gz > -rw------- 1 root root 63435 2010-08-01 03:10:02 /var/log/messages-20100801.gz > > You can see when I re-installed it without the patch. Both sets of logs went to > Sunday rotation (as they are supposed to) and have been rotated each Sunday > afterwards (as they are supposed to). > if I may add... 4. The patch logrotate-x.y.z-weekly breaks also logrotate -f behavior (no logs created/found). Fine without patch
Still valid with 3.8.1?
