Description of problem: All init scripts call (through /etc/profile.d/lang.sh if /etc/sysconfig/i18n is present) unicode_start which causes my monitor (CRT) to flicker/blank shortly, it seems to re-synchronize every time (not sure whether that's harmful or not, I'd be happier if it wouldn't do it that often ;-)). Version-Release number of selected component (if applicable): initscripts-8.45.7-1 How reproducible: Probably easy, it may depend on the graphics card used, I only have one ATI Radeon X800 GTO PCIe with which to test. Steps to Reproduce: 1. attach CRT to main video output 2. boot Actual results: Almost every init script invoked causes the screen to flicker/blank. Expected results: No unnecessary flickering/blanking shortly. Additional info: /etc/rc.d/rc already makes a call to /etc/profile.d/lang.sh via /etc/init.d/functions, you could perhaps subsequently export NOLOCALE=1 to the start/stop scripts which would avoid unnecessarily sourcing lang.sh and calling unicode_start.
There are some fixes in this regard in the F7 devel tree - you may want to try F7 test 4.
Fixes regarding the scripts or regarding unicode_start? I looked at the initscripts there (of course this machine wasn't showing the problem as it has an LCD, not a CRT) and it seemed to me like lang.sh/unicode_start was called there every time as well.
Changes to how/when the lang stuff is called - we don't always evaluate everything if LANG is already set, for example.
From looking at Rawhide's /etc/profile.d/lang.sh (and how it differs to the one in FC-6), unicode_start is explicitely called if a) LANG is set or /etc/sysconfig/i18n or $HOME/.i18n exist and b) LANG is set to a UTF-8 value. I think it's fair to assume that the problem will show on F7 as well. When test4 comes out, I can test a LiveCD on the system, though.
I've just tried it with the F7 test4 Live CD and found the same problem as on FC6, as expected. Shall I clone the bug for Rawhide?
Created attachment 153688 [details] proposed fix Patch for initscripts that sources lang.sh once at the beginning of the rc script and avoids sourcing it again via the functions scriptlet: 1.) rc sources lang.sh 2.) rc exports LANGSH_SOURCED=1 ... n.) init script avoids sourcing lang.sh due to set LANGSH_SOURCED n+1.) init script unsets LANGSH_SOURCED to avoid propagation which could "pollute" further invocations of init scripts ... Makes sense?
Note that the patch is against "upstream" CVS HEAD.
This problem still exists. Shall I just implement what I described in comment #6, try it out and commit it to CVS (and build)?
For rawhide? Commit it, and I'll build at some point in the near future.
Still to elvis CVS? I see that the last change in there is from mid-May, is the repository still active or has the upstream SCM been moved somewhere else?
Yes, elvis.
committed to elvis CVS
Nils, As you already touched lang.sh file, could you, please, also consider the issue reported in the bug #422141 ?
Fixed in 8.64-1.
¿Could this be fixed in initscripts-8.45.17.EL-1.el5.rpm? In RHEL 5.1?
Nothing along these lines has changed in RHEL.
Well, initscripts-8.45.17.EL-1.el5 have the same problem. Flickering start/stop scripts.
Yes, but the fix is too invasive to backport to RHEL 5.
Bug #238300 is a duplicate of this bug. (Meanwhile cheguaka returns to rhgb :p )
*** Bug 238300 has been marked as a duplicate of this bug. ***