Bugzilla – Full Text Bug Listing
|Summary:||Test installation and upgrade on Solaris 10|
|Product:||ThinLinc||Reporter:||Peter Åstrand <firstname.lastname@example.org>|
|Component:||Other||Assignee:||Peter Åstrand <email@example.com>|
|Status:||CLOSED FIXED||QA Contact:||Bugzilla mail exporter <firstname.lastname@example.org>|
Works (not text though, but that has its own bug).
Whatever you do when you test this bug, DO NOT FORGET TO ENABLE SSH TCP PORT FORWARDING! Starting a session seems to work. I'm too lazy to test printers as CUPS is not installed. Trouble starting the installer when locale is set to something else than C, but that has its own bug.
A few notes, to reduce future pain: * Make sure you remove ALL references of nis in /etc/nsswitch.conf. * Had to install lsof from sunfreeware.com. * Sane PATH: export PATH=/usr/sbin:/usr/bin:/opt/sfw/bin:/usr/sfw/bin:/usr/local/bin * Start the installation with: DISPLAY="" PYTHON=/usr/sfw/bin/python ./install-server * To create a test user: useradd user1 passwd user1 mv /home /oldhome mkdir /home mkdir /home/user1 chown user1 /home/user1 * sshd_config contains stupid: AllowUsers root AllowTcpForwarding no Works though.
Reproducible UI fail when upgrading on Solaris - the dialogue box with the pkgrm messages is not cleared before the install progress bar is shown.
Also, vsmserver started from tl-setup.py refuses to connect to an existing agent. Restarting vsmserver manually fixes this problem.
Upgrading from 3.1.2 to 3.2.0 (3128) fails because uninstall scripts in 3.1.2 doesn't work. Running install-server a couple of times will let it continue on further, but it'll eventually stop on removing CENDthinlinctladm. You'll have to edit /var/sadm/pkg/CENDthinlinctladm/install/preremove by hand to remove the bad if statement to make the problem go away. Unless we decide to patch the preremove script from install-server, I don't see how we can solve this by code. We can make sure that it won't happen again by always testing that our packages uninstall cleanly on Solaris as it is a vital part of the upgrade routine. Installing from scratch works like a charm, with DISPLAY=''. The problem mentioned in comment 7 is still there.
Clean installation worked fine, with the exception of the known issues: - You have to restart vsmserver because of a race - You have to manually kill tlstunnel and start tlwebadm because of another race
Upgrade from 3.2.0 seems to work fine as well. Running sessions are kept functional.
Situation remains as in previous tests.
Testing performed on sunray.lkpg.cendio.se with a fresh install of Solaris 10 u10, the latest Solaris 10 version at the time of the tests. * Configured SSH to allow port forwarding, as per Platform Specific Notes. * Installed ThinLinc 4.0.0 in text mode, works great. * Rolled back to pre-ThinLinc snapshot, installed 3.4.0, upgraded to 4.0.0. Works great. Most comments in comment #4 are no longer valid, no need to set set PATH, DISPLAY or PYTHON. Regarding users and home directories: /home is a mount point for automount, so don't move it around. First time, run this: echo "* sunray:/export/home/&" >> /etc/auto_home svcadm restart autofs Then for each user: useradd <user> passwd <user> mkdir /export/home/<user> chown <user> /export/home/<user>
Tested upgrade on "sunray". Checked: * tlwebadm * VSM * Logging in, starting Java desktop * Resizing the session * Mahjongg * Firefox * /usr/openwin/demo/worm Solaris is as fun as always...
Clean installation and upgrade from 4.1.0 both work fine.
Clean install and upgrade from 4.1.1 works fine with 4.2.0rc3. This issue still affects us though: - You have to restart vsmserver because of a race