Date: Tue, 03 Apr 2007 16:03:27 -0700 From: Julian Elischer <julian@elischer.org> To: Julian Elischer <julian@elischer.org> Cc: freebsd-stable@freebsd.org, "Andrey V. Elsukov" <bu7cher@yandex.ru>, julian@freebsd.org, JoaoBR <joao@matik.com.br> Subject: Re: ipfw add pipe broken in RELENG_6 Message-ID: <4612DD3F.1010401@elischer.org> In-Reply-To: <4612238D.6070504@elischer.org> References: <200704011107.22750.joao@matik.com.br> <200704030239.l332dC89044128@lava.sentex.ca> <4611FAB6.4090201@elischer.org> <46120138.5090702@yandex.ru> <4612238D.6070504@elischer.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Julian Elischer wrote: > Andrey V. Elsukov wrote: >> Julian Elischer пишет: >>> somewhere between my MFC testing and the commits there seems to have >>> been a screwup. >>> I think it happenned because I reverted a MFC out of my list of MFC's >>> to do after >>> I had done some tests because they causled a failure and I hadn't >>> realised that >>> they affected this code too. >>> >>> I'm doing testing now and should be able to confirm this in a short >>> while. >> >> Hi, Julian! >> >> Seems, that converting from time_second to time_uptime broke >> `ipfw -t show'. >> Now we have one PR: >> http://www.freebsd.org/cgi/query-pr.cgi?pr=111177 >> >> And i see similar problem on the my CURRENT. I have committed a fix in -current please check ip_fw2.c version 1.162 If this meets with general happiness I will re-MFC this 1.112 along with this fix to it. >> > > yeah I seem to have MFC'd a BUG.. > I confirmed they behaved the same. but I was looking at binary data > and didn't notice the date.. > > I'm preparing a revert, however it is a no-win situation as that leaves > you open to sessions timing out immediately when the time is changed > forward. > > So, which is more important? > accurate timeouts or accurate reporting? > > Ont thing that COULD be done would be to add the boot-time to the reported > times. this would never let a session time out too early, but would give > slightly misleading report numbers if the time is adjusted. They would > 'adjust' to the same offset against the 'new time'. > i.e. if it shows 2 hours before "now" before the change it will show 2 > hours before the "new now" after the time is changed. This may be > acceptable.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4612DD3F.1010401>