Red Hat Bugzilla – Bug 1315336
Regression: screen update breaks screen protocol
Last modified: 2017-02-07 07:54:27 EST
Description of problem:
It seems that the screen update between RHEL-7.1 - RHEL-7.2 broke the screen protocol. This can cause serious trouble to customers/users. Imagine scenarios where there are long term running processes in screens. By performing the screen update, these screens are no longer accessible and the processes running there are inaccessible.
What worse, the screen client hangs upon attempt to connect to the screen server and it is not possible to terminate it (i.e. it doesn't respond to CTRL+C). The only way how to terminate it is to login via another shell and kill the processes by hand.
This causes really bad user experience and can cause serious harm. What worse, only the release from the NVR was bumped which doesn't signal that the update uses incompatible protocol. Not counting the scenarios where the update process is initiated from the screen itself and user disconnects from the screen during the update.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. install screen-4.1.0-0.19.20120314git3c2946.el7.x86_64
3. disconnect from screen
4. update to screen-4.1.0-0.21.20120314git3c2946.el7.x86_64
5. screen -r
screen client hangs, screen server not accessible
screen session reconnected
I guess the protocol was broken by some backport. I think we should provide specific screen protocols downstream hacks, to make new screen clients backward compatible to the older server.
Screen downgrade make the running screens again accessible.
Probably dupe of bug 1253697, maybe fixed by screen-4.1.0-0.22.20120314git3c2946.el7, but I haven't checked it.
I have check this issue in RHEL-7.3 with screen package in the latest version in RHEL (screen-4.1.0-0.23.20120314git3c2946.el7_2).
This issue is not presented here. It looks like it was solved in bugzilla #1253697. This solution is already presented in RHEL-7.3 branch.
Because of this, I am closing this bugzilla as DUPLICATED of 1253697.
Feel free to reopen it in case you think, that described issue still persist.
*** This bug has been marked as a duplicate of bug 1253697 ***