On Saturday 11 May 2019 03:40:46 am Gene Heskett wrote: > On Friday 10 May 2019 05:06:25 pm Gene Heskett wrote: > > On Friday 10 May 2019 12:05:32 pm Chris Hassell wrote: > > > I think you just need to attack some debian-package-creation > > > goodness and cd down into "packaging/deb". > > > > > > From there ... outside of documentation and examples ... you have > > > these only to change: > > > > > > ~/.../packaging/deb #1$0 [..local:3_5]@BSL$ grep -n amandabackup * > > > > > > buildpkg:11:amanda_user=amandabackup > > > postinst.src:12:amanda_user=amandabackup > > > postrm.src:9:amanda_user=amandabackup > > > postrm.src:56: # only call deluser when amandabackup is > > > in $group preinst.src:10:amanda_user=amandabackup > > > preinst.src:19:# We require amandabackup to have a specific uid > > > because buildtime uid is rules:37:AMANDAUSER=amandabackup > > > > > > And you can even use the handy-dandy sed command line if you like: > > > > > > % sed -i -e 's/amandabackup/amanda/g' packaging/deb/* > > > > > > That's for the old issue... but for the new one (the prefix being > > > /usr/local).. you can just change the variable there in the > > > "rules" file (a package's make-and-package script). > > > > > > > > > 14 # These are variables that the user can override. They get > > > used in various 15 # places during configure, build, and install. > > > 16 PREFIX=/usr > > > 17 BINDIR=$(PREFIX)/bin > > > 18 triplet := $(shell dpkg-architecture -qDEB_HOST_MULTIARCH) > > > 19 ifeq ($(triplet),x86_64-linux-gnu) > > > 20 LIBDIR=$(PREFIX)/lib/$(triplet) > > > > > > Again that's packaging/deb/rules ... line #16 changed to wherever > > > you like your root-install to start. I guess PREFIX=/usr/local > > > should do it. > > > > > > -- CH > > > > > > On 5/10/19 8:14 AM, Gene Heskett wrote: > > > > > > On Friday 10 May 2019 10:13:30 am Chris Hassell wrote: > > > > > > > > > > > > On 5/9/19 11:31 PM, wrote: > > > > > > > > > <snip> > > > but scanimage -L locks up tight useing zero cpu and is unkillable, > > > I'll have to reboot to get rid of it. This kernel is a preempt-rt > > > kernel, and doesn't have some of a normal kernels bells and > > > whistles, like usbhid-ups is missing, and that speaks to my ups. > > > > > > > > > > > > I have a kernel like that too with Manjaro/Arch... non custom. > > > They're nice but boy they do take their time if something else is > > > competing for resources. Do a "dmesg" and see if something has > > > bit down hard on your kernel's locks and crashed/paniced some > > > thread. > > > > > > > > > > > > dmesg is rather non-commital, as I'd expect for a kernel that cam > > > make sub 10 microsecond latencies in an old phenom. Normal kernels > > > on an amd phenom are several hundred milliseconds. > > > As for dmesg, no unusual other than it doesn't find a scanner at > > > the same usb address as the Brother MFC-J6920DW printer. So I am > > > going to shut down, swap sata1 and sata0 cables and boot back to > > > wheezy where all this works, and see if the detection during dmesg > > > is different. > > > > > > > > > > > > Maybe using "ps l<pid>" will show the symbol/wait-channel that > > > it's stuck in. That's definitely one way to see why a process > > > (if you can find the one) is sitting at zero cpu. "Pstree" is fun > > > to use because it can show the 'leaf' ... the latest and lowest > > > child that was created that all the rest of the process group are > > > waiting on. That's usually the one in the spotlight... or the > > > do-nothing-spotlight in this case. > > > > > > > > > > > > Unforch, no pstree seems to be available, ISTR i looked yesterday. > > > Ahh, found it, its in psmisc.deb. And its not reporting some > > > things, I have a bash script that tells tde-kmail to go get the > > > mail when a mail comes in and is written to /var/mail/$name. htop > > > sees it. but pstree doesn't. > > > > > > pstree spits out 100 lines of hits, while htop says there are 111 > > > tasks ATM. But on a 5th run of pstree, mailwatcher is there. A > > > Head scratcher. > > > > > > But I think I need to get sane, or scanimage -L working next as > > > thats screw up several other things, like gimp and the *office* > > > stuffs. I can kill the ups stuff thats spamming the logs. I've > > > also posted to the emc-dev list about this kernel lack of a couple > > > modules. So I'll expect things will eventually get fixed. > > > > [...] > > One diff though, I > > > > > have never hand installed with a null prefix, always been > > > /usr/local. Using your procedure, how would I make it work to > > > build and install in /usr/local? > > > > Needs two more $varnames in the rules file, $USERNAME and $GROUPNAME > > so that one does not have to troll all the way thru the rules file > > to change all the names of the debs from amanda-backup-client to > > amanda-client yadda yadda. > > > > But this needs to be in the git clone as downloaded or possibly as a > > patch to be applied by autogen autogen or thereabouts. Otherwise > > its a lot of work, only to be rewritten back to the original by the > > fresh git clone. Or can I just do a make clean and rerun the > > packaging command to change both the prefix and the packaging names? > > > > > > I've done the rules edits, now lets see what I get starting with > > "bash autogen". > > > > But that blew up, and I don't see why, so rules > > and /var/log/amanda.install.log from yesterday's fail is attached. > > And I can't find the build.log if there is one. > > > > One other comment, I only tried dump once, 20 years ago, such a > > disaster I've used tar ever since, but I see it wasting time, a lot > > of it, testing dump. You may want to take a survey and see if among > > us, there is anybody using dump. It might be worthwhile to drop it. > > > > This build baled out here: And this time, maybe. I keep setting > > bashes damned history up to 10,000 lines, and the next time I look > > its back to 1000 which is not enough by a hell of a long row of > > apple trees. So this time I set it unlimited. And in the first 50 > > lines I see its back to amandabackup for a user. ctl+c and I give up > > for today. I have to go pay the rent on a wheelchair that hasn't > > been used since last summer. Same with her car she'll never drive > > again but won't let me sell it. 2007 Toy RAV-4 with about 63k miles > > on it. > > > > So 3 things, make prefix=/usr/local, make user=amanda. And make > > debs "amanda-server-version.deb" and "amanda-client-version.deb" > > and "amanda-common-version.deb", that last I don't see being built. > > > > Put it in the git repo as we can always move prefix. Or make prefix > > part of the autogen command line, or maybe even a user generated > > file the package builder can refer to, hell even an env $var, > > whatever floats the boat best. Your call, but it needs to be > > adjustable for those building their own. > > > > > Copyright 2019 by Maurice E. Heskett > > > Cheers, Gene Heskett > > I just found something else, the use of localhost in the server build > should also be adjusted to be either the alias or the fqdn of the > machine doing the server role. I've adjusted the rules file here by > using geany's search and replace to replace all instances of > amanda-backup with $(AMANDAUSER), changed line 16 to /usr/local and > another clean from the git clone build is underway. So far its only > complaint is that smbclient, which has never been used here, does not > seem to be smbclient.
Thats because it doesn't even exist there as its looking for it in /usr/local/bin, when its actually in /usr/bin. So its both an erronious message, its likely an erronious use of $(PREFIX) to look for it in the rules file, so that breaks when prefix is changed. Sigh. > Since Tridgel took on a helper samba/cifs has > been buggier than ten day old roadkill. I use, in the rest of my > networking here, sshfs, and it Just Works. But thats just my opinion, > beating all the variations of samba/cifs and NFS out of use. I have a > root dir "sshnet" and all my other live machines are mounted to > subdirs there. And I don't have to be root to mount them. I have > known_keys in my home dir that make it all login-less. The thing is I > have write perms to anything on those machines that I as first user > can write too. Sweet. Painless. Just Works. > I have all that in /etc/rc.local but stretch does not appear to > execute /etc/rc.local. So I now run bin/mount-machines after a reboot: > #!/bin/bash > sshfs gene@shop:/ /sshnet/shop > sshfs gene@lathe:/ /sshnet/lathe > sshfs gene@GO704:/ /sshnet/GO704 > sshfs pi@picnc:/ /sshnet/picnc > > And like magic they all appear, fully accessible to mc for anything I > own. > But that enough preaching. Back to the build, which errored: > > dh_install -v --sourcedir=/home/amanda/amanda/build/debian/tmp > > >> /home/amanda/amanda/build/debian/dpkg.log 2>&1 > > debian/rules:122: recipe for target 'binary-arch' failed > make: *** [binary-arch] Error 255 > dpkg-buildpackage: error: fakeroot debian/rules binary gave error exit > status 2 > > From line 121: > binary-arch: build > echo "---->dh_testdir: " >> $(log) > > What is error 255? > > > > Cheers, Gene Heskett > > > > Copyright 2019 by Maurice E. Heskett > > Cheers, Gene Heskett > > Copyright 2019 by Maurice E. Heskett > Cheers, Gene Heskett Copyright 2019 by Maurice E. Heskett Cheers, Gene Heskett -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Genes Web page <http://geneslinuxbox.net:6309/gene>
