Bug 248688 – genkernel-9999 + --integrated-initramfs + openrc-0.3 = uncolored output in console while booting

• Status: RESOLVED FIXED
• Product: Gentoo Hosted Projects

[Description Alexey Shvetsov 2008-11-24 21:48:48 UTC]

output in console while booting is b&w
see attached screenshot =)

Reproducible: Always

[Comment 1 Alexey Shvetsov 2008-11-24 21:49:21 UTC]

Created attachment 173277 [details]
output in console

[Comment 2 Andrew Gaffney (RETIRED) 2008-11-24 21:50:40 UTC]

I can tell you now that --integrated-initramfs has nothing to do with it.

[Comment 3 Andrew Gaffney (RETIRED) 2008-11-24 21:51:21 UTC]

Have you tried booting with a kernel that doesn't have an initramfs?

[Comment 4 Alexey Shvetsov 2008-11-24 21:56:25 UTC]

(In reply to comment #3)
> Have you tried booting with a kernel that doesn't have an initramfs?
> 

hmm thats strange since i use v86d + uvesafb
previosly i use kernel + initramfs and all was ok
if i boot kernel only with v86d part of initramfs than all ok too 

[Comment 5 Alexey Shvetsov 2008-11-24 21:57:19 UTC]

btw i saw this both in x86 and x86_64

[Comment 6 Mike Auty (RETIRED) 2008-12-17 14:32:26 UTC]

Created attachment 175606 [details]
ebuild-bisect tool

Ok,

This also happens with openrc-0.4.0.  I'm pretty sure it's because of commit fa9cbeeda84c645aac2a71f98796c411f378eced (call env -i to clear out the environment).

I found this using a bisection tool I wrote for live ebuilds (should work with subversion and git) called ebuild-bisect, which I've attached here and is also available as an ebuild in my overlay.  Unfortunately that commit didn't run, so I couldn't quite complete the bisection, but it's somewhere between commits 48b282a85d9730075c0c6aa86cb5a5efb802cce1 (good) and 41f44b1d4229b6781ccb2af16e35eb8bbc26d901 (bad).

Unfortunately I'm not sure which environment variable open-rc relies on to determine whether it's being run in a real console (hence colour and fixed widths) or a non-interactive/fake one...

[Comment 7 Andrew Gaffney (RETIRED) 2008-12-17 15:24:52 UTC]

Since making that commit, I've found out that the kernel put key=value pairs from the kernel commandline into the env for pid 1. We already knew this was happening, but we figured it was something that the gk initramfs was doing.

I've reverted that particular change. Please give it a test.

[Comment 8 Mike Auty (RETIRED) 2008-12-18 00:43:09 UTC]

Yep, master seems to fix the problem, all back to colour and the correct width again.  Thanks...  5:)

[Comment 9 Norberto Bensa 2008-12-20 14:37:27 UTC]

*** Bug 251732 has been marked as a duplicate of this bug. ***

[Comment 10 Lars Wendler (Polynomial-C) (RETIRED) 2008-12-24 02:55:04 UTC]

*** Bug 252345 has been marked as a duplicate of this bug. ***

[Comment 11 Lars Wendler (Polynomial-C) (RETIRED) 2009-01-04 11:18:53 UTC]

*** Bug 253678 has been marked as a duplicate of this bug. ***

[Comment 12 Andrew Gaffney (RETIRED) 2009-12-26 22:52:07 UTC]

This was released a while ago.