[opensuse-factory] UsrMerge preparations
Hi, Quite some packages carry around sections like this: #UsrMerge mkdir $RPM_BUILD_ROOT/sbin ln -s %_sbindir/foo $RPM_BUILD_ROOT/sbin/foo #EndUsrMerge That was added years ago as preparation of replacing /bin, /sbin, /lib and /lib64 with a symlink and merging the content into their counterparts in /usr. I'm not sure how that was supposed to work in practice though. To actually get rid of resp disable that code even for testing one would have to modify a hundred packages. So I'd like to replace those comments with a conditional, ie %if !0%{?usrmerged} mkdir %buildroot/sbin ln -s %_sbindir/foo %buildroot/sbin/foo %endif That doesn't yet answer the question when and how we finally do the /usr merge. However at least it would be possible to flip a switch via prjconf to stop building compat symlinks in a branch project then. Factory as of today would keep building the packages the way they are now, so no change in content at this point. Using the macro also paves the way to actually test a usr merged environment in staging in the future. So unless there are better ideas I'd just go ahead and adjust all packages that have those symlinks. cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.com/ SUSE Software Solutions Germany GmbH, GF: Felix Imendörffer HRB 36809 (AG Nürnberg)
On Tue, Nov 17, 2020 at 8:18 AM Ludwig Nussel <ludwig.nussel@suse.de> wrote:
Hi,
Quite some packages carry around sections like this:
#UsrMerge mkdir $RPM_BUILD_ROOT/sbin ln -s %_sbindir/foo $RPM_BUILD_ROOT/sbin/foo #EndUsrMerge
That was added years ago as preparation of replacing /bin, /sbin, /lib and /lib64 with a symlink and merging the content into their counterparts in /usr. I'm not sure how that was supposed to work in practice though. To actually get rid of resp disable that code even for testing one would have to modify a hundred packages.
So I'd like to replace those comments with a conditional, ie
%if !0%{?usrmerged} mkdir %buildroot/sbin ln -s %_sbindir/foo %buildroot/sbin/foo %endif
That doesn't yet answer the question when and how we finally do the /usr merge. However at least it would be possible to flip a switch via prjconf to stop building compat symlinks in a branch project then. Factory as of today would keep building the packages the way they are now, so no change in content at this point. Using the macro also paves the way to actually test a usr merged environment in staging in the future.
So unless there are better ideas I'd just go ahead and adjust all packages that have those symlinks.
I'd more be in favor of just ripping all that stuff out, but yeah, your plan makes sense. -- 真実はいつも一つ！/ Always, there's only one truth!
Neal Gompa wrote:
On Tue, Nov 17, 2020 at 8:18 AM Ludwig Nussel <ludwig.nussel@suse.de> wrote:
Quite some packages carry around sections like this:
#UsrMerge mkdir $RPM_BUILD_ROOT/sbin ln -s %_sbindir/foo $RPM_BUILD_ROOT/sbin/foo #EndUsrMerge
That was added years ago as preparation of replacing /bin, /sbin, /lib and /lib64 with a symlink and merging the content into their counterparts in /usr. I'm not sure how that was supposed to work in practice though. To actually get rid of resp disable that code even for testing one would have to modify a hundred packages.
So I'd like to replace those comments with a conditional, ie
%if !0%{?usrmerged} mkdir %buildroot/sbin ln -s %_sbindir/foo %buildroot/sbin/foo %endif
That doesn't yet answer the question when and how we finally do the /usr merge. However at least it would be possible to flip a switch via prjconf to stop building compat symlinks in a branch project then. Factory as of today would keep building the packages the way they are now, so no change in content at this point. Using the macro also paves the way to actually test a usr merged environment in staging in the future.
So unless there are better ideas I'd just go ahead and adjust all packages that have those symlinks.
I'd more be in favor of just ripping all that stuff out, but yeah, your plan makes sense.
Longer term sure. Right now you can't just rip out eg /bin/bash. That would immediately bring all package builds to a grinding halt :-) Others might cause more subtle breakage either for package builds or at actual installations. So I'm in favor of using a global switch that allows us to prepare packages without causing breakage for users and packagers and then test the whole thing in staging including openQA etc at some point. cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.com/ SUSE Software Solutions Germany GmbH, GF: Felix Imendörffer HRB 36809 (AG Nürnberg)
On 11/17/20 9:26 AM, Ludwig Nussel wrote:
Neal Gompa wrote:
On Tue, Nov 17, 2020 at 8:18 AM Ludwig Nussel <ludwig.nussel@suse.de> wrote:
Quite some packages carry around sections like this:
#UsrMerge mkdir $RPM_BUILD_ROOT/sbin ln -s %_sbindir/foo $RPM_BUILD_ROOT/sbin/foo #EndUsrMerge
That was added years ago as preparation of replacing /bin, /sbin, /lib and /lib64 with a symlink and merging the content into their counterparts in /usr. I'm not sure how that was supposed to work in practice though. To actually get rid of resp disable that code even for testing one would have to modify a hundred packages.
So I'd like to replace those comments with a conditional, ie
%if !0%{?usrmerged} mkdir %buildroot/sbin ln -s %_sbindir/foo %buildroot/sbin/foo %endif
That doesn't yet answer the question when and how we finally do the /usr merge. However at least it would be possible to flip a switch via prjconf to stop building compat symlinks in a branch project then. Factory as of today would keep building the packages the way they are now, so no change in content at this point. Using the macro also paves the way to actually test a usr merged environment in staging in the future.
So unless there are better ideas I'd just go ahead and adjust all packages that have those symlinks.
I'd more be in favor of just ripping all that stuff out, but yeah, your plan makes sense.
Longer term sure. Right now you can't just rip out eg /bin/bash. That would immediately bring all package builds to a grinding halt :-) Others might cause more subtle breakage either for package builds or at actual installations. So I'm in favor of using a global switch that allows us to prepare packages without causing breakage for users and packagers and then test the whole thing in staging including openQA etc at some point.
Most, if not all of these are from me and there was a wiki once that tracked the packages. Yes using a global switch is a better approach. Before going down the path I'd say it might makes sense to reach consensus that this is going to happen first. The effort got stuck, at the time and then eventually dropped by me, based on some package maintainers not agreeing with the arguments made for the change to dropping /bin and others in favor of /usr based locations. Times have changed and maybe those maintainers have changed their minds or maybe we can have a greater initiative to move this along such that those maintainers cannot be in a position of being blockers. In any event the markers should make things easy to find which was really the original goal ;) Later, Robert -- Robert Schweikert MAY THE SOURCE BE WITH YOU Distinguished Engineer LINUX Technical Team Lead Public Cloud rjschwei@suse.com IRC: robjo
Robert Schweikert wrote:
[...] So I'd like to replace those comments with a conditional, ie
%if !0%{?usrmerged} mkdir %buildroot/sbin ln -s %_sbindir/foo %buildroot/sbin/foo %endif [...] Most, if not all of these are from me and there was a wiki once that
On Tue, Nov 17, 2020 at 8:18 AM Ludwig Nussel <ludwig.nussel@suse.de> wrote: tracked the packages. Yes using a global switch is a better approach.
Before going down the path I'd say it might makes sense to reach consensus that this is going to happen first. The effort got stuck, at the time and then eventually dropped by me, based on some package maintainers not agreeing with the arguments made for the change to dropping /bin and others in favor of /usr based locations. Times have changed and maybe those maintainers have changed their minds or maybe we can have a greater initiative to move this along such that those maintainers cannot be in a position of being blockers.
The good thing is that with the macro we don't have to decide about the usr merge now. Nothing would break in Factory. Those who want to further explore the usr merge can then actually proceed to create a working prototype without having to fight merge conflicts though. cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.com/ SUSE Software Solutions Germany GmbH, GF: Felix Imendörffer HRB 36809 (AG Nürnberg)
On Tuesday 2020-11-17 14:18, Ludwig Nussel wrote:
Quite some packages carry around sections like this:
#UsrMerge mkdir $RPM_BUILD_ROOT/sbin ln -s %_sbindir/foo $RPM_BUILD_ROOT/sbin/foo #EndUsrMerge
So I'd like to replace those comments with a conditional, ie
%if !0%{?usrmerged} mkdir %buildroot/sbin ln -s %_sbindir/foo %buildroot/sbin/foo %endif
The world managed to move many things to /usr/sbin even without needing /sbin symlinks, which is great. There are but a few programs left where the same needs to be done, and/or whose symlinks could just be nuked, probably. What I have left: /sbin# echo * agetty apparmor_parser arp badblocks blkid blockdev blogctl blogd blogger cfdisk chcpu checkproc chkconfig ctrlaltdel debugfs depmod dumpe2fs e2fsck e2image e2label e2mmpstatus e2undo ether-wake faillock fbtest fdisk findfs fsck fsck.cramfs fsck.ext2 fsck.ext3 fsck.ext4 fsck.minix fsck.xfs fsfreeze fstab-decode fstrim halt hwclock init insmod insserv install-info ip ipmaddr iptunnel isserial killall5 killproc ldconfig logsave losetup lsmod mke2fs mkfs mkfs.bfs mkfs.cramfs mkfs.ext2 mkfs.ext3 mkfs.ext4 mkfs.minix mkfs.xfs mkhomedir_helper mkill mkswap modinfo modprobe mount.fuse mount.fuse3 nameif netconfig nologin pam_namespace_helper pam_tally2 pam_timestamp_check pidofproc pivot_root plipconfig poweroff raw rcapparmor rdisc reboot refresh_initrd resize2fs rmmod runlevel rvmtab service setconsole sfdisk showconsole shutdown slattach smart_agetty start_daemon startproc swaplabel swapoff swapon switch_root sysctl telinit tune2fs unix2_chkpwd unix_chkpwd unix_update vhangup wipefs xfs_repair With that, I find it prospective to just keep emptying /sbin and in doing so, drop the #UsrMerge blocks from specfiles, without going to %usrmerged, but straight to "done".
Hi, On 17 Nov 14:55 2020, Jan Engelhardt wrote:
With that, I find it prospective to just keep emptying /sbin and in doing so, drop the #UsrMerge blocks from specfiles, without going to %usrmerged, but straight to "done".
I agree with Jan on this, let's get this done. I signup for fixing part of the possible fallout ;) ismail -- Entrenched in both ancient and modern cultural tradition, the pursuit of happiness is a popular preoccupation around the world. — Ed Diener SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg, Germany GF: Felix Imendörffer (HRB 36809, AG Nürnberg)
Jan Engelhardt wrote:
On Tuesday 2020-11-17 14:18, Ludwig Nussel wrote:
Quite some packages carry around sections like this:
#UsrMerge mkdir $RPM_BUILD_ROOT/sbin ln -s %_sbindir/foo $RPM_BUILD_ROOT/sbin/foo #EndUsrMerge
So I'd like to replace those comments with a conditional, ie
%if !0%{?usrmerged} mkdir %buildroot/sbin ln -s %_sbindir/foo %buildroot/sbin/foo %endif
The world managed to move many things to /usr/sbin even without needing /sbin symlinks, which is great. There are but a few programs left where the same needs to be done, and/or whose symlinks could just be nuked, probably. What I have left:
/sbin# echo * agetty apparmor_parser arp badblocks blkid blockdev blogctl blogd blogger cfdisk chcpu checkproc chkconfig ctrlaltdel debugfs depmod dumpe2fs e2fsck e2image e2label e2mmpstatus e2undo ether-wake faillock fbtest fdisk findfs fsck fsck.cramfs fsck.ext2 fsck.ext3 fsck.ext4 fsck.minix fsck.xfs fsfreeze fstab-decode fstrim halt hwclock init insmod insserv install-info ip ipmaddr iptunnel isserial killall5 killproc ldconfig logsave losetup lsmod mke2fs mkfs mkfs.bfs mkfs.cramfs mkfs.ext2 mkfs.ext3 mkfs.ext4 mkfs.minix mkfs.xfs mkhomedir_helper mkill mkswap modinfo modprobe mount.fuse mount.fuse3 nameif netconfig nologin pam_namespace_helper pam_tally2 pam_timestamp_check pidofproc pivot_root plipconfig poweroff raw rcapparmor rdisc reboot refresh_initrd resize2fs rmmod runlevel rvmtab service setconsole sfdisk showconsole shutdown slattach smart_agetty start_daemon startproc swaplabel swapoff swapon switch_root sysctl telinit tune2fs unix2_chkpwd unix_chkpwd unix_update vhangup wipefs xfs_repair
With that, I find it prospective to just keep emptying /sbin and in doing so, drop the #UsrMerge blocks from specfiles, without going to %usrmerged, but straight to "done".
Could do that but then that would require verifying that each and every package you touch doesn't actually break after removing the stuff in /bin while we're not usrmerged yet. So I wouldn't stop anyone from doing that but to me it feels like it's quicker to not spend thought on it and keep the compat block. Adding the %if is just finger exercise, no brain needed. And once we actually did the merge, the insight whether the stuff in /bin was useless for some packages is irrelevant anyways :-) FWIW the list files I compiled from ARCHIVES.gz is -- aaa_base /sbin/refresh_initrd /sbin/service /sbin/smart_agetty -- apparmor-parser /sbin/apparmor_parser /sbin/rcapparmor -- audit /sbin/audispd /sbin/auditctl /sbin/auditd /sbin/augenrules /sbin/aureport /sbin/ausearch /sbin/autrace -- aws-efs-utils /sbin/mount.efs -- bash /bin/bash -- biosdevname /sbin/biosdevname -- blktrace /bin/blkparse /bin/blktrace /bin/btrace -- blog /sbin/blogctl /sbin/blogd /sbin/blogger /sbin/isserial /sbin/showconsole -- btrfsprogs /sbin/btrfs /sbin/btrfs-convert /sbin/btrfs-image /sbin/btrfsck /sbin/btrfstune /sbin/fsck.btrfs /sbin/mkfs.btrfs -- busybox-coreutils /bin/arch /bin/basename /bin/cat /bin/chgrp /bin/chmod /bin/chown /bin/cp /bin/date /bin/dd /bin/df /bin/echo /bin/false /bin/ln /bin/ls /bin/md5sum /bin/mkdir /bin/mknod /bin/mktemp /bin/mv /bin/pwd /bin/readlink /bin/rm /bin/rmdir /bin/sleep /bin/sort /bin/stat /bin/stty /bin/sync /bin/touch /bin/true /bin/uname -- busybox-cpio /bin/cpio -- busybox-ed /bin/ed -- busybox-findutils /bin/find -- busybox-gawk /bin/awk -- busybox-grep /bin/egrep /bin/fgrep /bin/grep -- busybox-gzip /bin/gunzip /bin/gzip /bin/zcat -- busybox-hostname /bin/dnsdomainname /bin/hostname -- busybox-iproute2 /bin/ip /sbin/ip -- busybox-iputils /bin/arping /bin/ping /bin/ping6 -- busybox-kbd /bin/chvt /bin/deallocvt /bin/openvt /bin/setfont /bin/setkeycodes /bin/setlogcons /bin/showkey /sbin/loadkmap -- busybox-kmod /bin/lsmod /sbin/depmod /sbin/insmod /sbin/lsmod /sbin/modinfo /sbin/modprobe /sbin/rmmod -- busybox-misc /bin/chattr /bin/fatattr /bin/iostat /bin/lsattr /bin/lzop /bin/makemime /bin/mpstat /bin/reformime /sbin/mdev /sbin/mkdosfs /sbin/mke2fs /sbin/mkfs.ext2 /sbin/mkfs.vfat /sbin/setconsole /sbin/uevent /sbin/zcip -- busybox-net-tools /bin/ifconfig /bin/netstat /bin/route /sbin/arp /sbin/ether-wake /sbin/iptunnel /sbin/nameif /sbin/slattach -- busybox-policycoreutils /sbin/restorecon /sbin/setfiles -- busybox-procps /bin/pgrep /bin/pkill /bin/ps /sbin/sysctl -- busybox-psmisc /bin/fuser -- busybox-sed /bin/sed -- busybox-sh /bin/sh -- busybox-sysvinit-tools /bin/fsync /bin/pidof /bin/usleep /sbin/killall5 /sbin/pidof -- busybox-tar /bin/tar -- busybox-tunctl /sbin/tunctl -- busybox-util-linux /bin/dmesg /bin/kill /bin/logger /bin/login /bin/more /bin/mount /bin/su /bin/umount /sbin/blkid /sbin/findfs /sbin/fsck /sbin/losetup /sbin/mkswap /sbin/nologin -- busybox-vi /bin/vi -- busybox-vlan /sbin/vconfig -- ceph-common /sbin/mount.ceph -- cifs-utils /sbin/mount.cifs -- crda /sbin/crda /sbin/regdbdump -- cryptsetup /sbin/cryptsetup -- dash /bin/dash -- davfs2 /sbin/mount.davfs /sbin/umount.davfs -- dbus-1 /bin/dbus-cleanup-sockets /bin/dbus-daemon /bin/dbus-monitor /bin/dbus-send /bin/dbus-test-tool /bin/dbus-update-activation-environment /bin/dbus-uuidgen -- dd_rescue /bin/dd_rescue -- dd_rhelp /bin/dd_rhelp -- device-mapper /sbin/dmsetup -- dhcp-client /sbin/dhclient /sbin/dhclient-script /sbin/dhclient6 -- diod /sbin/mount.diod -- dmraid /sbin/dmevent_tool /sbin/dmraid -- dosfstools /sbin/dosfsck /sbin/dosfslabel /sbin/fsck.fat /sbin/fsck.msdos /sbin/fsck.vfat /sbin/mkfs.fat /sbin/mkfs.msdos -- dracut /sbin/installkernel /sbin/mkinitrd -- drbd-utils /sbin/drbdadm /sbin/drbdmeta /sbin/drbdmon /sbin/drbdsetup -- e2fsprogs /sbin/badblocks /sbin/debugfs /sbin/dumpe2fs /sbin/e2fsck /sbin/e2image /sbin/e2label /sbin/e2mmpstatus /sbin/e2undo /sbin/fsck.ext2 /sbin/fsck.ext3 /sbin/fsck.ext4 /sbin/logsave /sbin/mkfs.ext3 /sbin/mkfs.ext4 /sbin/resize2fs /sbin/tune2fs -- ecryptfs-utils /sbin/mount.ecryptfs /sbin/mount.ecryptfs_private /sbin/umount.ecryptfs /sbin/umount.ecryptfs_private -- elilo /sbin/elilo /sbin/eliloalt -- exfat-utils /sbin/dumpexfat /sbin/exfatfsck /sbin/exfatlabel /sbin/fsck.exfat /sbin/mkexfatfs /sbin/mkfs.exfat -- f2fs-tools-compat /sbin/defrag.f2fs /sbin/dump.f2fs /sbin/f2fstat /sbin/fibmap.f2fs /sbin/fsck.f2fs /sbin/mkfs.f2fs /sbin/parse.f2fs /sbin/resize.f2fs /sbin/sload.f2fs -- fedfs-utils-client /sbin/mount.fedfs -- fillup /bin/fillup -- fuse /sbin/mount.fuse -- fuse-exfat /sbin/mount.exfat /sbin/mount.exfat-fuse -- fuse3 /sbin/mount.fuse3 -- fxload /sbin/fxload -- gawk /bin/gawk -- glibc /sbin/ldconfig -- glusterfs /sbin/mount.glusterfs -- grubby /sbin/grubby /sbin/new-kernel-pkg -- hdparm /sbin/hdparm /sbin/wiper.sh -- hostname /bin/domainname /bin/nisdomainname /bin/ypdomainname -- info /sbin/install-info -- info4 /sbin/install-info4 -- initviocons /bin/initviocons -- insserv-compat /sbin/chkconfig /sbin/insserv -- iputils /bin/clockdiff /bin/tracepath /bin/tracepath6 /sbin/rdisc -- ipvsadm /sbin/ipvsadm /sbin/ipvsadm-restore /sbin/ipvsadm-save -- isapnp /sbin/isapnp /sbin/pnpdump -- iscsiuio /sbin/brcm_iscsiuio /sbin/iscsiuio -- jfsutils /sbin/fsck.jfs /sbin/jfs_debugfs /sbin/jfs_fsck /sbin/jfs_fscklog /sbin/jfs_logdump /sbin/jfs_mkfs /sbin/jfs_tune /sbin/mkfs.jfs -- kbd /bin/clrunimap /bin/dumpkeys /bin/fgconsole /bin/getkeycodes /bin/getunimap /bin/kbd_mode /bin/kbdinfo /bin/kbdrate /bin/loadkeys /bin/loadunimap /bin/mapscrn /bin/outpsfheader /bin/psfaddtable /bin/psfgettable /bin/psfstriptable /bin/psfxtable /bin/resizecons /bin/screendump /bin/setleds /bin/setmetamode /bin/setpalette /bin/setvesablank /bin/setvtrgb /bin/showconsolefont /bin/spawn_console /bin/spawn_login /bin/unicode_start /bin/unicode_stop /sbin/fbtest -- kexec-tools /sbin/kexec -- keyutils /bin/keyctl /sbin/key.dns_resolver /sbin/request-key -- klogd /sbin/klogd -- kpartx /sbin/kpartx -- ksh /bin/ksh93 -- libewf-tools /sbin/mount.ewf /sbin/umount.ewf -- live-net-installer /bin/extend -- lvm2 /sbin/lvchange /sbin/lvconvert /sbin/lvcreate /sbin/lvdisplay /sbin/lvextend /sbin/lvm /sbin/lvmconfig /sbin/lvmdiskscan /sbin/lvmdump /sbin/lvmpolld /sbin/lvmsadc /sbin/lvmsar /sbin/lvreduce /sbin/lvremove /sbin/lvrename /sbin/lvresize /sbin/lvs /sbin/lvscan /sbin/pvchange /sbin/pvck /sbin/pvcreate /sbin/pvdisplay /sbin/pvmove /sbin/pvremove /sbin/pvresize /sbin/pvs /sbin/pvscan /sbin/vgcfgbackup /sbin/vgcfgrestore /sbin/vgchange /sbin/vgck /sbin/vgconvert /sbin/vgcreate /sbin/vgdisplay /sbin/vgexport /sbin/vgextend /sbin/vgimport /sbin/vgimportclone /sbin/vgmerge /sbin/vgmknodes /sbin/vgreduce /sbin/vgremove /sbin/vgrename /sbin/vgs /sbin/vgscan /sbin/vgsplit -- mailx /bin/mail -- makedev /sbin/MAKEDEV -- mawk /bin/mawk -- mcstrans /sbin/mcstransd -- mdadm /sbin/mdadm /sbin/mdmon -- mingetty /sbin/mingetty -- mksh /bin/ksh /bin/lksh /bin/mksh /bin/pdksh -- multipath-tools /sbin/mpathpersist /sbin/multipath /sbin/multipathd -- munin /sbin/rcmunin-cgi-graph /sbin/rcmunin-cgi-html -- munin-node /sbin/rcmunin-node -- net-tools /sbin/plipconfig -- net-tools-deprecated /sbin/ipmaddr -- netconsole-tools /sbin/netconsole-server -- nfs-client /sbin/mount.nfs /sbin/mount.nfs4 /sbin/umount.nfs /sbin/umount.nfs4 -- nfs-kernel-server /sbin/nfsdcltrack -- nilfs-utils /sbin/mkfs.nilfs2 /sbin/mount.nilfs2 /sbin/nilfs_cleanerd /sbin/umount.nilfs2 -- ntfs-3g /sbin/mount.lowntfs-3g /sbin/mount.ntfs /sbin/mount.ntfs-3g -- ntfsprogs /sbin/mkfs.ntfs -- ocfs2-tools /sbin/fsck.ocfs2 /sbin/mkfs.ocfs2 /sbin/mount.ocfs2 /sbin/mounted.ocfs2 /sbin/o2cluster /sbin/ocfs2_hb_ctl /sbin/tunefs.ocfs2 -- ocfs2-tools-o2cb /sbin/o2cb /sbin/o2cb.init /sbin/o2cb_ctl /sbin/ocfs2.init -- ooRexx /sbin/rcooRexx -- open-iscsi /sbin/iscsi-gen-initiatorname /sbin/iscsi-iname /sbin/iscsi_discovery /sbin/iscsi_fw_login /sbin/iscsi_offload /sbin/iscsiadm /sbin/iscsid /sbin/iscsistart -- open-vm-tools /sbin/mount.vmhgfs -- pam /sbin/faillock /sbin/mkhomedir_helper /sbin/pam_namespace_helper /sbin/pam_timestamp_check /sbin/unix2_chkpwd /sbin/unix_chkpwd /sbin/unix_update -- pam-deprecated /sbin/pam_tally2 -- pam_mount /sbin/mount.crypt /sbin/mount.crypt_LUKS /sbin/mount.crypto_LUKS /sbin/umount.crypt /sbin/umount.crypt_LUKS /sbin/umount.crypto_LUKS -- pciutils /sbin/lspci /sbin/setpci -- pcmciautils /sbin/lspcmcia /sbin/pccardctl -- perl-Bootloader /sbin/pbl /sbin/update-bootloader -- plymouth /bin/plymouth -- policycoreutils /sbin/restorecon_xattr -- polkit-default-privs /sbin/chkstat-polkit /sbin/set_polkit_default_privs -- powerd /sbin/detectups /sbin/powerd -- rarpd /sbin/rarpd -- rash /bin/rash -- reiserfs /sbin/debugfs.reiserfs /sbin/debugreiserfs /sbin/fsck.reiserfs /sbin/mkfs.reiserfs /sbin/mkreiserfs /sbin/reiserfsck /sbin/reiserfstune /sbin/resize_reiserfs /sbin/tunefs.reiserfs -- rpcbind /bin/rpcinfo /sbin/pmap_set2 /sbin/rpcbind /sbin/rpcinfo -- rpm /bin/rpm -- rsyslog /sbin/rsyslogd -- rungetty /sbin/rungetty -- scsh /bin/scsh -- scsirastools /sbin/getmd /sbin/mdevt /sbin/sgdefects /sbin/sgdiag /sbin/sgdiskmon /sbin/sgdskfl /sbin/sgevt /sbin/sgmode /sbin/sgraidmon /sbin/sgsafte -- sdparm /sbin/sas_disk_blink /sbin/scsi_ch_swp /sbin/sdparm -- supportutils /sbin/analyzevmcore /sbin/chkbin /sbin/getappcore /sbin/supportconfig -- sysconfig /sbin/ifuser /sbin/rcnetwork -- sysconfig-netconfig /sbin/netconfig -- syslog-ng /sbin/syslog-ng -- syslog-service /sbin/rcsyslog -- syslogd /sbin/syslogd -- systemd-sysvinit /sbin/halt /sbin/init /sbin/poweroff /sbin/reboot /sbin/runlevel /sbin/shutdown /sbin/telinit -- sysvinit-tools /sbin/checkproc /sbin/fstab-decode /sbin/killproc /sbin/mkill /sbin/pidofproc /sbin/rvmtab /sbin/start_daemon /sbin/startproc /sbin/vhangup -- tcsh /bin/csh /bin/tcsh -- tomoyo-tools /sbin/tomoyo-init -- udhcp /sbin/udhcpc -- util-linux /bin/findmnt /bin/lsblk /sbin/agetty /sbin/blockdev /sbin/cfdisk /sbin/chcpu /sbin/ctrlaltdel /sbin/fdisk /sbin/fsck.cramfs /sbin/fsck.minix /sbin/fsfreeze /sbin/fstrim /sbin/hwclock /sbin/mkfs /sbin/mkfs.bfs /sbin/mkfs.cramfs /sbin/mkfs.minix /sbin/pivot_root /sbin/raw /sbin/sfdisk /sbin/swaplabel /sbin/swapoff /sbin/swapon /sbin/switch_root /sbin/wipefs -- vim /bin/ex /bin/vim -- virtualbox /sbin/vboxconfig -- virtualbox-guest-tools /sbin/mount.vboxsf /sbin/vboxguestconfig -- virtualbox-qt /sbin/vbox-fix-usb-rules.sh -- wicked-service /sbin/ifdown /sbin/ifprobe /sbin/ifstatus /sbin/ifup -- xen-tools-domU /bin/domu-xenstore /bin/xenstore-chmod /bin/xenstore-exists /bin/xenstore-list /bin/xenstore-ls /bin/xenstore-read /bin/xenstore-rm /bin/xenstore-watch /bin/xenstore-write -- xfsdump /sbin/xfsdump /sbin/xfsrestore -- xfsprogs /sbin/fsck.xfs /sbin/mkfs.xfs /sbin/xfs_repair -- yast2 /sbin/yast /sbin/yast2 -- zsh /bin/zsh cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.com/ SUSE Software Solutions Germany GmbH, GF: Felix Imendörffer HRB 36809 (AG Nürnberg)
On 17. 11. 20, 14:55, Jan Engelhardt wrote:
With that, I find it prospective to just keep emptying /sbin and in doing so, drop the #UsrMerge blocks from specfiles, without going to %usrmerged, but straight to "done".
As a start, what about a post build check like we have for uses of /usr/bin/env after shebang? This would eliminate a lot, esp. for /bin/bash IMO. regards, -- js suse labs
On 11/18/20 5:18 PM, Jiri Slaby wrote:
On 17. 11. 20, 14:55, Jan Engelhardt wrote:
With that, I find it prospective to just keep emptying /sbin and in doing so, drop the #UsrMerge blocks from specfiles, without going to %usrmerged, but straight to "done".
As a start, what about a post build check like we have for uses of /usr/bin/env after shebang? This would eliminate a lot, esp. for /bin/bash IMO.
I was going to suggest the same thing, but i'll also enable your flag in the dbus-1 package. -- Simon Lees (Simotek) http://simotek.net Emergency Update Team keybase.io/simotek SUSE Linux Adelaide Australia, UTC+10:30 GPG Fingerprint: 5B87 DB9D 88DC F606 E489 CEC5 0922 C246 02F0 014B
Jiri Slaby wrote:
On 17. 11. 20, 14:55, Jan Engelhardt wrote:
With that, I find it prospective to just keep emptying /sbin and in doing so, drop the #UsrMerge blocks from specfiles, without going to %usrmerged, but straight to "done".
As a start, what about a post build check like we have for uses of /usr/bin/env after shebang? This would eliminate a lot, esp. for /bin/bash IMO.
To nag at packagers to patch all script to use /usr/bin/bash? That would be over the top IMO. It isn't wrong to use /bin/sh in shell scripts. That's kind of ABI IMO. We can't change /lib/ld-linux.so.2 hardcoded in binaries either after all. That's why the effort needs to be called usr merge rather than usr move :-) After the merge using either location works as it's the same. cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.com/ SUSE Software Solutions Germany GmbH, GF: Felix Imendörffer HRB 36809 (AG Nürnberg)
On 11/18/20 8:02 PM, Ludwig Nussel wrote:
Jiri Slaby wrote:
On 17. 11. 20, 14:55, Jan Engelhardt wrote:
With that, I find it prospective to just keep emptying /sbin and in doing so, drop the #UsrMerge blocks from specfiles, without going to %usrmerged, but straight to "done".
As a start, what about a post build check like we have for uses of /usr/bin/env after shebang? This would eliminate a lot, esp. for /bin/bash IMO.
To nag at packagers to patch all script to use /usr/bin/bash? That would be over the top IMO. It isn't wrong to use /bin/sh in shell scripts. That's kind of ABI IMO. We can't change /lib/ld-linux.so.2 hardcoded in binaries either after all. That's why the effort needs to be called usr merge rather than usr move :-) After the merge using either location works as it's the same.
The current one for /usr/bin/env is just a warning, plenty of packagers choose to ignore it and we have plenty of cases where it still exists (especially in python libs), so providing it as a warning is probably useful info for packagers should they feel like changing it in a cleanup, at the same time the review team shouldn't block packages with this warning from entering the distro. -- Simon Lees (Simotek) http://simotek.net Emergency Update Team keybase.io/simotek SUSE Linux Adelaide Australia, UTC+10:30 GPG Fingerprint: 5B87 DB9D 88DC F606 E489 CEC5 0922 C246 02F0 014B
On 2020/11/17 05:18, Ludwig Nussel wrote:
Quite some packages carry around sections like this: #UsrMerge mkdir $RPM_BUILD_ROOT/sbin ln -s %_sbindir/foo $RPM_BUILD_ROOT/sbin/foo #EndUsrMerge That was added years ago as preparation of replacing /bin, /sbin, /lib and /lib64 with a symlink and merging the content into their counterparts in /usr. I'm not sure how that was supposed to work in practice though.
---- For many who had a separate physical disk for '/' and '/usr', and booted from the physical disk, it didn't work very well, but that was dealt with by just calling the old config, "not supported", so any problems caused by situations like that could be tossed out with "unsupported config". It's happening again -- with various dirs moving off root and off /usr and under /usr/share (which got so big in my /usr, that I had to split it to /home/share and mount that on /usr/share). Again, some similar potential problem points in that just like a separate /usr can't be mounted until root is available, a separate /usr/share can't be mounted until /usr is mounted. That always meant that tricks like putting 'mount' and/or its libs on /usr, and putting a symlink from /bin/mount -> /usr/bin/mount result in an unbootable system. It seems there's also another move to put /etc/ => /usr/share/etc ? or was that tabled? There was a reason why file locations were specified in the LSB. Violating that will tend to make suse less compatible with other distro's which seems like useless churn, especially given that most authors will still put config files in /etc/. That makes a larger burden for suse to convert all [progs desired to run 'natively'] programs to the new format. That will make force app-writers to create a special version for suse -- something certainly likely to reduce the number of apps that easily work if you move them from many linux's to suse. It's sorta why perl became less popular over time. Different versions of perl became more incompatible w/each other -- especially as you go up the version chain. So more people come to the point of realizing that adopting or upgrading to the latest perl risks more incompatibility with previous versions. So adoption of new perls has declined for the past 10 years or so. In the same way, someone wanting to distribute on linux, will lean toward those distros that provide the most inter-compatibility. Are other Distros moving the /etc contents to the suse location?
On 11/21/20 1:38 PM, L A Walsh wrote:
On 2020/11/17 05:18, Ludwig Nussel wrote:
Quite some packages carry around sections like this: #UsrMerge mkdir $RPM_BUILD_ROOT/sbin ln -s %_sbindir/foo $RPM_BUILD_ROOT/sbin/foo #EndUsrMerge That was added years ago as preparation of replacing /bin, /sbin, /lib and /lib64 with a symlink and merging the content into their counterparts in /usr. I'm not sure how that was supposed to work in practice though.
For many who had a separate physical disk for '/' and '/usr', and booted from the physical disk, it didn't work very well, but that was dealt with by just calling the old config, "not supported", so any problems caused by situations like that could be tossed out with "unsupported config".
It's happening again -- with various dirs moving off root and off /usr and under /usr/share (which got so big in my /usr, that I had to split it to /home/share and mount that on /usr/share).
Again, some similar potential problem points in that just like a separate /usr can't be mounted until root is available, a separate /usr/share can't be mounted until /usr is mounted. That always meant that tricks like putting 'mount' and/or its libs on /usr, and putting a symlink from /bin/mount -> /usr/bin/mount result in an unbootable system.
It seems there's also another move to put /etc/ => /usr/share/etc ? or was that tabled? There was a reason why file locations were specified in the LSB. Violating that will tend to make suse less compatible with other distro's which seems like useless churn, especially given that most authors will still put config files in /etc/. That makes a larger burden for suse to convert all [progs desired to run 'natively'] programs to the new format.
There are equally reasons for splitting up /etc which were mentioned in the thread, and seemed to be significant enough for people to be willing to invest in making the change.
That will make force app-writers to create a special version for suse -- something certainly likely to reduce the number of apps that easily work if you move them from many linux's to suse. in
Not really, in most plain cases packagers will make it work, for 3rd party apps if an app is expecting to use /etc it will still just work unless you are using a system with transactional updates configured.
It's sorta why perl became less popular over time. Different versions of perl became more incompatible w/each other -- especially as you go up the version chain. So more people come to the point of realizing that adopting or upgrading to the latest perl risks more incompatibility with previous versions.
So adoption of new perls has declined for the past 10 years or so. In the same way, someone wanting to distribute on linux, will lean toward those distros that provide the most inter-compatibility.
I'd disagree here atleast about the perl part. I can't remember the last time I did something and thought perl was the best language for the job. There are just better languages around now.
Are other Distros moving the /etc contents to the suse location?
Many are doing the split, but the location is pretty inconsistent so far, fedora / Red Hat were looking at /usr/lib for some reason. But the hard bit is teaching stuff to read from multiple locations i'm sure most distro's will settle on a standard at somepoint and then everyone migrating from whatever they pick now shouldn't be too bad. But splitting system and user config is a pretty different topic to moving files into /usr -- Simon Lees (Simotek) http://simotek.net Emergency Update Team keybase.io/simotek SUSE Linux Adelaide Australia, UTC+10:30 GPG Fingerprint: 5B87 DB9D 88DC F606 E489 CEC5 0922 C246 02F0 014B
On Sat, Nov 21, 2020 at 4:32 AM Simon Lees <sflees@suse.de> wrote:
On 11/21/20 1:38 PM, L A Walsh wrote:
On 2020/11/17 05:18, Ludwig Nussel wrote:
Quite some packages carry around sections like this: #UsrMerge mkdir $RPM_BUILD_ROOT/sbin ln -s %_sbindir/foo $RPM_BUILD_ROOT/sbin/foo #EndUsrMerge That was added years ago as preparation of replacing /bin, /sbin, /lib and /lib64 with a symlink and merging the content into their counterparts in /usr. I'm not sure how that was supposed to work in practice though.
For many who had a separate physical disk for '/' and '/usr', and booted from the physical disk, it didn't work very well, but that was dealt with by just calling the old config, "not supported", so any problems caused by situations like that could be tossed out with "unsupported config".
It's happening again -- with various dirs moving off root and off /usr and under /usr/share (which got so big in my /usr, that I had to split it to /home/share and mount that on /usr/share).
Again, some similar potential problem points in that just like a separate /usr can't be mounted until root is available, a separate /usr/share can't be mounted until /usr is mounted. That always meant that tricks like putting 'mount' and/or its libs on /usr, and putting a symlink from /bin/mount -> /usr/bin/mount result in an unbootable system.
It seems there's also another move to put /etc/ => /usr/share/etc ? or was that tabled? There was a reason why file locations were specified in the LSB. Violating that will tend to make suse less compatible with other distro's which seems like useless churn, especially given that most authors will still put config files in /etc/. That makes a larger burden for suse to convert all [progs desired to run 'natively'] programs to the new format.
There are equally reasons for splitting up /etc which were mentioned in the thread, and seemed to be significant enough for people to be willing to invest in making the change.
That will make force app-writers to create a special version for suse -- something certainly likely to reduce the number of apps that easily work if you move them from many linux's to suse. in
Not really, in most plain cases packagers will make it work, for 3rd party apps if an app is expecting to use /etc it will still just work unless you are using a system with transactional updates configured.
It's sorta why perl became less popular over time. Different versions of perl became more incompatible w/each other -- especially as you go up the version chain. So more people come to the point of realizing that adopting or upgrading to the latest perl risks more incompatibility with previous versions.
So adoption of new perls has declined for the past 10 years or so. In the same way, someone wanting to distribute on linux, will lean toward those distros that provide the most inter-compatibility.
I'd disagree here atleast about the perl part. I can't remember the last time I did something and thought perl was the best language for the job. There are just better languages around now.
Are other Distros moving the /etc contents to the suse location?
Many are doing the split, but the location is pretty inconsistent so far, fedora / Red Hat were looking at /usr/lib for some reason. But the hard bit is teaching stuff to read from multiple locations i'm sure most distro's will settle on a standard at somepoint and then everyone migrating from whatever they pick now shouldn't be too bad. But splitting system and user config is a pretty different topic to moving files into /usr
Red Hat/Fedora inherited the usage of /usr/lib from upstream projects. Unfortunately, RPM-OSTree reserves /usr/etc for diversions of content installed into /etc for the purpose of being able to do automatic merging of package changes to admin changes on OSTree updates, so they can't use /usr/etc. Other distributions feel that /usr/etc is bad and that SUSE is doing it wrong. They'd rather use the existing convention or create a new dedicated directory path in /usr/share. -- 真実はいつも一つ！/ Always, there's only one truth!
Ludwig Nussel wrote:
[...] So unless there are better ideas I'd just go ahead and adjust all packages that have those symlinks.
Meanwhile most requests went through. Thanks to all package maintainers who accepted them! I've updated https://en.opensuse.org/openSUSE:Usr_merge with some insights and questions that came up in the process. Also, I've hit a problem with update-alternatives. In particular gawk/mawk, mksh/ksh and mailutils/mailx chose to handle both the /bin as well as the /usr/sbin variant via update-alternatives instead of just linking one to the other. When we get rid of one, update-alternatives gets confused and removes both symlinks¹. Any ideas how to fix that? cu Ludwig [1] https://en.opensuse.org/index.php?title=openSUSE:Usr_merge#update-alternativ... -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.com/ SUSE Software Solutions Germany GmbH, GF: Felix Imendörffer HRB 36809 (AG Nürnberg)
On Wednesday 2020-12-02 16:03, Ludwig Nussel wrote:
Also, I've hit a problem with update-alternatives. In particular gawk/mawk, mksh/ksh and mailutils/mailx chose to handle both the /bin as well as the /usr/sbin variant via update-alternatives instead of just linking one to the other. When we get rid of one, update-alternatives gets confused and removes both symlinks¹. Any ideas how to fix that?
I would imagine that, if both are gone, you have free reign to instantiate whatever new objects you like in those places.
participants (8)
-
İsmail Dönmez -
Jan Engelhardt -
Jiri Slaby -
L A Walsh -
Ludwig Nussel -
Neal Gompa -
Robert Schweikert -
Simon Lees