Date: Fri, 3 Sep 1999 18:07:10 -0700 (PDT) From: John Polstra <jdp@polstra.com> To: jeff@mountin.net Cc: stable@freebsd.org Subject: Re: buildword curiousities Message-ID: <199909040107.SAA10009@vashon.polstra.com> In-Reply-To: <3.0.3.32.19990903155015.0142d100@207.227.119.2> References: <3.0.3.32.19990903155015.0142d100@207.227.119.2>
next in thread | previous in thread | raw e-mail | index | archive | help
In article <3.0.3.32.19990903155015.0142d100@207.227.119.2>, Jeffrey J. Mountin <jeff@mountin.net> wrote: > > No matter since my main point is that in both cases deleting section of the > source tree and re-cvsup'ing did not help. Gave up and blew away the > entire source tree, reinstalled the source from 3.2R, and cvsup'd again. > > The first time there were no problem cvsup'ing. Recently there have been > problems with cvsup6 hanging at times and needing to ^C and reconnect. > During the day it can be slow, but even when it's saturating a 128K ISDN > connection it will hang at times, especially if there are a lot of changes > (oddly enough last night it sent everything to bring 3.2R source up to > -stable). Stops dead and never finishes, even hours later. Hmm, when you suspect problems on a CVSup server, you can do us all a big favor by writing to the site's maintainer and asking him to check and make sure things are running smoothly. The maintainers are listed along with all the CVSup mirror sites in section 25.4 of the FreeBSD Handbook. I'm the maintainer of cvsup6, so consider me contacted. :-) I checked the log files and couldn't find any sign of trouble. It seems to be ticking along nicely. Perhaps your network path to cvsup6 is bad lately. Try a couple other mirror sites and see if any of them work better for you. (Just add "-h cvsup5.freebsd.org" to your cvsup command line to try cvsup5, for example.) The next time you get one of those hangs on cvsup6, please leave it hung and drop me an e-mail. If I'm around at the time, I'll login to cvsup6 and see what your connection looks like from its end of the pipe. > It may due to passive mode, If it is then it's a problem with your network stack probably. Try multiplexed mode (the default in CVSup-16.0, which I see you're using). It's the best all-around mode to use, and it should work in all situations. > but one has to wonder if the ^C leads to some corruption. It shouldn't. CVSup is carefully designed to do all its work in temporary files, renaming each one atomically to the actual name only after it has confirmed that the MD5 checksum is correct. At worst you might end up with a few stray temporary files lying around. But it tries hard to clean those up as well, even if you kill it with ^C. > Oddly, one time I found a badly mangled file that had not changed > for over 6 months and was never touched by cvsup or moved, barring a > hardware change. From time to time I receive reports of mangled files. But they always (always, always, every time) have turned out to be caused by HW problems or kernel problems. You can tell by making a hex dump of the offending file with "hd", and looking at where the damage begins. I can almost guarantee that it will begin at an offset which is a multiple of 4K bytes (usually it's a multiple of 8K bytes). That's a dead giveaway that the VM system or your hardware has screwed you. There's nothing CVSup or any other program can do to protect you against that. If the MD5 checksum says the file is fine (when it is still in the kernel buffers, no doubt) but then the kernel and HW don't manage to get the bits written correctly onto the disk -- well, the application did its best. John -- John Polstra jdp@polstra.com John D. Polstra & Co., Inc. Seattle, Washington USA "No matter how cynical I get, I just can't keep up." -- Nora Ephron To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199909040107.SAA10009>