Weblogs for forrest
Any valid mail claiming to be from a user @mydomain.com would originate from inside my NATted local network (192.168.xxx.xxx).
Is there a magic incantation I can put in my exim config to reject any outside mail claiming to be from @mydomain.com ?
As all of the commenters suggested, I tried bringing up my box without X.
Now here I have to interject a gripe about Debian: there's no way to bring up a full-featured system without X: there's single-user mode, and then the very next runlevel starts X.
My first linux distro (back in 1995) was Slackware. It was sensibly configured so that runlevel 2 was everything you could possibly run without X, and then runlevel 3 started X. You've got 4 levels between 1 for single-user and 6 for reboot: it seems there should be some useful differentiation of functionality between them, right?
Of course, maybe I'm missing something and there's an easy way to get all your daemons fired up, but I just booted single-user mode for my test.
I booted my uniprocessor 2.6.18 with noapci as a parameter. I couldn't think of much to do in single-user mode, but I started a song with ogg123, backgrounded it, and did an ls -lR /. That ran fine. Then I started X with "init 2" (see gripe above) and I was back to the situation I described before: a blank screen, and when I ssh'd into it, I saw Xorg eating 98% of the CPU.
Then I tried building another SMP 2.6.18 kernel. I had high hopes for "noapci" solving my problem, but I also wanted to tweak some other kernel parameters: I made the Radeon framebuffer device a module instead of compiling it into the kernel, thinking that might help with the video issues.
I booted that kernel with the parameters "noapci vga=ask s" Interestingly, it didn't ask me about vga modes at all, but kept it at the standard 80x25. My last smp kernel only made as far as "waiting for /dev to be populated" (or something like that) so I was quite pleasantly suprised when it booted all the way up to the prompt for my root password.
I got inspired testing my smp kernel without X. I ripped a CD using "abcde -j 4" which allows as many as 4 processes (cdparanoia and oggenc) to run at once. While that was moving right along at a heady clip, I backgrounded it and listened played another ogg file with ogg123. Perfect sound, not even a hint of a hiccup.
So, I booted to X, and ... X came up. I tried the ogg123 test I mentioned in my previous post and it worked fine. I brought Iceweasel up and started to enter a post about how everything was rosy for me now ... but then I needed to go do some other things. I got back to my computer maybe an hour later and started to type but nothing happened. I moved the mouse, but the mouse cursor didn't move. I tried to ssh in, but got "No route to host".
So now I'm back to my working 2.6.16 uniprocessor kernel.
And I'm wondering what to try next.
These are more notes to myself to try and keep track of what I have and haven't tried, but if anyone has a suggestion I'd be glad to hear it.
This machine, gemini, is a dual-Athlon box with a Tyan S2466N-4M mobo.
I have an ATI graphics card and a ViewSonic VX900 LCD monitor: a combo which acts funky in a way that maybe I'll begin to describe once I've discussed the system hangs sufficiently.
When I got this machine I was running a 2.4.x SMP kernel, and even then it hanged (hung?) occasionally. That seemed to be related to it choosing a 3-D screensaver.
So, I just didn't do that.
The SMP kernels seemed to work fine up to 2.6.10. I didn't upgrade for a long time and the next kernel I tried was 2.6.16. With SMP turned on, I could get as far as logging into my GNOME session, but it would hang pretty quickly whenever I started doing anything. I found I could consistently get it to hang by opening a terminal window, playing an ogg file with ogg123, and trying to drag the terminal window. Usually the window would only move a little bit before the system locked up.
I tried turning off "Preempt the Big Kernel Lock" because that sounded particularly ominous, but the result was the same.
Tragically, I'm running a uniprocessor 2.6.16 now, wasting half my processing power.
When I tried upgrading to 2.6.18, I couldn't even run a uniprocessor kernel. The modes of failure were different this time ...
I guess you could say that 2.6.18 uni didn't crash, it boots all the way up but when GDM is supposed to fire up the screen goes blank. I logged in remotely from another linux box I have and "top" told me that Xorg was taking over 90% of the CPU!
... more later ...
Today I finally reinstalled enigmail after maybe a month of not having it. Then I ran apt-get -u dist-upgrade again and it wants to remove it again! No, no, bad apt!
I've tried playing around with pinning, but apparently that can't prevent removal, only up/downgrading. I saw the --no-remove switch in the man page, but I only want to prevent this one package from being removed.
It seems like there should be a way to do this ...
I'd like to be able to use Chinese characters more easily on my sid desktop here, but so far I've only been able to enter Chinese from within emacs.
I used to use the emacs program mew for my mail, but switched to mozilla mail for a number of reasons (especially Enigmail, the amazing GPG plugin). When I did that, I sacrificed my ability to send Chinese e-mail.
There is a debian package called iiimf-htt-server (where iiimf stands for Internet/Intranet Input Method Framework), along with plugins for various input methods (e.g. iiimf-htt-le-newpy is one for Chinese).
I can't figure out how to use the damn thing at all. There is hardly any documentation in English about using Chinese on Linux. There is considerable documentation in Chinese, which I could try to decipher, but it's all geared towards setting up a completely Chinese environment.
Anyone have a clue?
I think it's unfortunate that the weblogs here don't allow comments like the articles: I might mention something like "I had such a hard time doing X", not thinking that the passing comment was worthy of a real question in the articles section, but someone might still have an easier way.
For example I'd like to make a comment on Steve's observations about article popularity (oh, you can't link to individual blog articles either ... not that I want to pick on this exellent site too much. Who knows what will or won't be useful when you first start out?)
Anyway, Steve, I found the java article useful even though I had already installed java "the old way", and had no clue about the existence of make-jpkg, which sounds much easier.
I may in fact upgrade to java 1.5 with that knowlege, which is something I was hesitating to do.
OTOH, I've been running clamav for as long as I've had my mailserver. I glanced at that article and quickly saw there was nothing new for me there. (I've also been running sa-exim as well: that might be a good article topic. Yeah, what are people doing to fight spam?)
BTW, has anyone installed an up-to-date tomcat (5.0.x or 5.5.x) to with a mod_jk connector to a debian-installed apache2? How about a decent install of an up-to-date (3.x) version of the Eclipse java development environment?
I've been using debian since hamm, and I'm loath to change now, but I'm doing more and more java development in my day job and I'd like to have those tools available to me. I'm going to give Ubuntu a try and see if they have better java development support.