1 |
Beso <givemesugarr@×××××.com> posted |
2 |
d257c3560811230324r6f2d2a31p74a546e8be289cd2@××××××××××.com, excerpted |
3 |
below, on Sun, 23 Nov 2008 11:24:23 +0000: |
4 |
|
5 |
> by the way i'm still waiting for the kernel upgrade script post... :-) |
6 |
|
7 |
Well, I had made an assumption that proved true for changing kernel |
8 |
versions at the time, but invalid over the longer term. Specifically, I |
9 |
had the scripts copying over the outputdir (which I set so the sources |
10 |
themselves stay clean, everything I've done is in the outputdir), so a |
11 |
new build would only have to rebuild where the code had changed. Well, |
12 |
all that broke early in the 2.6.27 series, between rc1 and rc2, when x86 |
13 |
moved its headers around. For several rcs I was unable to compile at |
14 |
all, until I figured out it was still trying to use the old headers |
15 |
from .26, mixed up with the ones from .27! That was NOT working! |
16 |
|
17 |
So, I had to figure out the problem (which early on I thought might be |
18 |
the scripts, it was, but not quite as I expected), then fix the scripts |
19 |
so they cleared the outputdir properly (while still saving and |
20 |
transferring the .config in the same dir), then test my changes thru |
21 |
several rcs... |
22 |
|
23 |
Meanwhile I found a real bug late in the .27 development cycle and was |
24 |
pressed to bisect it, learning a bunch about git and git bisect in the |
25 |
process... and updating my scripts with git versions... all within only a |
26 |
couple weeks before target release, since I hadn't been able to test |
27 |
earlier since I couldn't compile. |
28 |
|
29 |
Oh, and along in there I typoed a variable name and a script that was |
30 |
supposed to clean the outputdir in my git version, instead tried to |
31 |
"clean" /! Yes, I brown-bag-errored a rm -rf /! Fortunately I caught it |
32 |
before it got thru with /home, but unfortunately after it killed /bin, |
33 |
/dev, /etc... But fortunately my backups weren't too out of date. |
34 |
Still, I had to recover in the middle of that already time-pressed bisect |
35 |
cycle. But I did, and I finished the bisect, thereby tracing the problem |
36 |
to a single commit, and they figured out the problem and patched it |
37 |
before 2.6.27 release. |
38 |
|
39 |
And I found a bug to bisect in the .28 rc series as well. It was keeping |
40 |
my machine from properly hibernating, which meant after every failed test |
41 |
during the bisect cycle, there was a good chance the RAID-6 wasn't synced |
42 |
at the crash, and I had it doing the resync when I rebooted. |
43 |
|
44 |
But it's all working relatively good again now... and I'm getting back to |
45 |
trying to catch-up on all these things! =:^) |
46 |
|
47 |
-- |
48 |
Duncan - List replies preferred. No HTML msgs. |
49 |
"Every nonfree program has a lord, a master -- |
50 |
and if you use the program, he is your master." Richard Stallman |