Thanks for all the answers. Particularly this: On Fri, Aug 15, 2014 at 9:05 PM, Benjamin Baier <program...@netzbasis.de> wrote: > Here are the newest numbers i can provide for a full build from source. > /usr/src >900MB > /usr/xenocara >700MB > /usr/obj >900MB > /usr/xobj >500MB > /usr/ports >600MB > /usr/ports/pobj can't be big enought...
I decided to wipe the repositories, reformat the partitions with more inodes, and start over. This time, I planned to just try the patch approach. Thought I had planned to, anyway. Gave the src and xenocara partition double the inodes (newfs -i 4096, since the default was 8192), and the ports quadruple the inodes (newfs i- 2048). Forgot to try the populate flag, maybe next time, removed the contents of the xenocara directory on /usr so I wouldn't be locking stuff under the mount. Re-pre-populated all three with the tarballs. And I decided that I might as well update /usr/src, since that seemed to go okay before. export CVSROOT=anon...@anoncvs.au.openbsd.org sudo cvs -d$CVSROOT co -P src So much for my plan to try using patches. But that went okay, and I decided, since I had ended up starting down the path for -current, to keep going with the extra inodes and see what would happen. /usr/ports runs to xmris/scripts, then stalls. And I get the broken pipe. df -ih show me to only be using 45% of the partition and 27% of the inodes. So I tried sudo cvs -d$CVSROOT co -P ports/xmris That goes through to xmris/scripts and returns with no errors. But there is no ports/xmris directory. Just to try something stupid, I tried sudo mkdir ports/xmris; ls -ld ports/x* and tried again to co ports/xmris, but the result was xmris being deleted again. I'll try the co on ports again from a ssh session so I can capture the broken pipe message. -- Joel Rees > On 08/15/14 05:09, Joel Rees wrote: >> >> I'm trying re-learn how to bring a new install up to -stable, and I've >> been following the instructions on >> >> http://www.openbsd.org/stable.html >> >> and >> >> http://www.openbsd.org/faq/faq5.html#Bld >> >> and not doing a very good job of it. The recommended partition left me >> with only 1.4G for /usr, and it was 90% full when I finished unpacking >> the sys, src, ports, and xenocara tarballs. >> >> (ancient IBM thinkpad with only 256M RAM and 20G (17 real gig) or hard >> disk. 860 MHz or so CPU.) >> >> I had saved 2.5G out of the suggested size for /home, so I cut a 1G >> partition for /usr/ports and gave it the default newfs. mount on /mnt, >> "cp -pR /usr/ports/ /mnt/" (I always mess that up -- "mv /mnt/ports/* >> /mnt; mv /mnt/ports/.cvsignore /mnt".) Deleted the original contents >> of /usr/ports, which I now see was a mistake, and mounted the new >> partition on /usr/ports. >> >> And then I did a cvs co on src, ports, and xenocara. >> >> About an hour later, it tells me I have no inodes left on ports. "df >> -ih" tells me I have 398 M used on /usr/ports, which is 42%, but >> 155,676 inodes in use, which is 100%. I forgot to write down what it >> was trying to check out when it ran out. >> >> /usr/src looks like its complete, with 111,613 inodes in use and >> 70,273 free, 1.2G partition with 313M free. I'm thinking that's room >> enough to build the patches and a few other things I need. >> >> What size partition should I cut for /usr/ports, and how many inodes >> should I allocate it? Or should I just not try to bring /usr/ports up >> to stable? >> >> And what can I expect for /usr/xenocara? Just from unpacking the >> tarball, it's using close to 700M on /usr, so I'm planning on cutting >> it a partition, too. >> >> My thinking is to use my remaining 1.5G for a new /usr/ports, give it >> 500,000 inodes and cp -pR again, to save bandwidth on the mirror, then >> take the 1 G partition that would be freed, give it 300,000 inodes, >> and use it for /usr/xenocara. >> >> Can anyone tell me if that will be enough? >> >> Or maybe I should just do it the other way, from the patch sets, I think >> it was.
