Older blog entries for RoUS (starting at number 53)

post-meridian, EDT

hardware
Well, I upgraded my kernel to the latest RH 7.2 RPMs, X likewise, downloaded and installed the 1.1.20t Savage S3 driver, and added an 'Option "NoAccel"' line to my XF86Config-4 file, and my laptop has managed to stay up without hanging now for almost 72 hours. Fingers crossed..

Linux
Even with the hardware problems I've been having I still have two systems that have been up since the last millenium. :-)

autoresponder
Brian Behlendorf pointed out another no-op situation for autoresponder: it shouldn't bother trying to do anything if there's no address to which a reply can be sent.. Updated to 1.15.2, but no release announcement made yet.

Spackle
Last week I went to add Spackle to Freshmeat, and found that Jeremy Weatherford had created a project there with the same name a couple of weeks ago. We worked it out; he changed his project's name to 'SpackleStats', and we each reference the other's project on our various pages.

post-meridian, EDT

Apache
jhermann, putenv() will certainly do the trick, but it's generally not thread-safe nor terribly modular.. You're working around the API rather than within it. :-/ I'm not even sure it will work on Windows and other non-Unixish platforms.

ante-meridian, EDT

dandt
dandt is a 'download and test' script I wrote for putting the Apache 2.0 package through its paces with the Perl regression test framework. It's getting more and more involved, and AFAIK no-one else uses it but me, but it produces results like those at the Source-Zone site. It still needs a wee bit of work to handle test failures and harness failures differently, and it would be nice if it handled Apache 1.3, but at the moment I'm making what changes are necessary to use it at work (you know, the place that pays me ;-).

Apache
One of the things that has continually come up on the apache-modules mailing list is the fact that Apache 1.[23] calls module init handlers twice, once before daemonising and once after, and it's difficult to tell the difference. D'oh! {sound of manipulative member smacking anteriour cephalic surface} Why didn't anyone think of adding an API routine so modules could find out? I'm adding that for 1.3.next; 1.3 is a 'stable' and 'robust' stream, and not a lot of development energy is being devoted to it, but I think it's going to be around for years yet and this may make life easier for at least some module writers out there..

Athlon CPU problems
Bryce, have you seen this?

ThinkPad woes
Upgraded my ThinkPad T21 from Red Hat 7.1 to 7.2 a couple of days ago, in part to see if it would make the hanging problems I've been seeing go away. No joy. Added the 'Option "ShadowStatus"' line to my XF86Config-4 file, but the display still hangs on a daily basis. The last few haven't really been hangs but X looping intensely -- so much so that I can't even switch to a different virtual console. Thank goodness I can ssh in and kill X (unless it was a hard hang) and do a clean reboot if needed, but this display problem is making the box essentially unusable. :-(

post-meridian, EDT

SourceForge rant
Heh. It's interesting, the number of people who apparently didn't "get" the point of the article. Yes, SF is free, and use of its services are a privilege and not a right. However, once a project has been hosted there, getting it off again is a major hassle, so the 'vote with your feet' solution is of limited utility. They're arguably the most visible host of open software on the 'net; if they 'get away' with jerking around the developers, it sends a signal that's bad for the entire community. And it's hardly in their best interest to alienate the users.

So what's the point of the article? I suppose a notice to developers who use SF but may not have been aware of this change, and that by extension they shouldn't expect any guaranty of service in this or any other areas; in other words, 'beware! this could happen to some facility you care about!'. And that SF is in the service of VA, and not the community. And a twit to SF so that perhaps they'll either answer problems reported to them, or (gasp!) address them. :-)

As for cmiller's remarks.. well, just because he doesn't believe in a cause doesn't make it unworthwhile to others. :-)

ante-meridian, EDT

SourceForge rant
Well, SourceForge closing one of my problem reports without resolution kicked me into posting an Advogato article about the issue. Bah..

B5
Taped the new Babylon 5 movie ('Legend of the Rangers', I think) for a friend last night. I always programme the VCR to start five minutes before air-time, and stop at least five minutes after the published end-time. I was taping at SP quality, and this would be the first time I've found a tape with less than 130 minutes on it.. It ran out of tape and shut off about two minutes before the end of the programme. Fortunately SciFi ran it again immediately, so I was able to scramble another tape in and record it (at EP this time -- quality be buggered). So I'll give him two tapes to-day, and tell him, "Watch this one up to the first advert, and then pop it out and pop this one in to watch the rest." Buggrit! Millenium hand and shrimp!

ante-meridian (quite), EDT

on capitalism
nymia wrote:

This is the thing I like about capitalism, it simply works.

I suspect there are some people in Argentina with differing opinions. :-(

on hardware
I'm finally shifting most of my laptop operations over to the new ThinkPad T21 I bought in September 2001. It has shown a troubling tendency to hang, however -- first the display (magic sysrq seems to get out of that), then a really hard hang that requires power cycling. Right now I'm trying a workaround with a 'Option ShadowStatus' line in my XF86Config file. The 600E that belongs to the company has been running continuously for months; I'd likewise prefer not to be forced to reboot the new one at random intervals. Just added the option yesterday, so it's too soon to tell if it has fixed the problem.

on VMware
I've got VMware installed on my Linux laptop, with a Win2K guest OS. It simply rocks. It doesn't rock perfectly (The Sims can't access the CD/DVD at the low level it wants in order to 'validate' me as a registered user, and the LEGO(tm) Mindstorms 1.5 system can't seem to properly download the RCX through the VMware->Linux->serial port pathway), but it's excellent natheless. At least it's good enough for my main purposes, which is parallel building of Apache on Linux and Windows so I can do the research for my books.

testing, testing.. 1, 2, 3..
Another thing that rocks is the Apache Perl regression test framework. It's excellent for Apache module development, though it's more than a little difficult to wrap your mind around the how-to aspect unless you're a Perl guru or have futzed with the CPAN testing stuff in the past. But if you can express a module's desired behaviour through tests for correctness, you can verify the code progress quickly and automagically. I'm adding sections about it, and sample tests, to my Apache modules book.

ante-meridian, EDT

huh? what?
I haven't had a lot to report in the last couple of months, so I didn't report it. But now I do, and I am.

spackle
A few weeks ago my group ran into an odd problem with our CVS repository: people's diff commit messages were getting entangled. The repository was correct, it was just the mail messages that were sometimes combined. The culprit? An invalid assumption of uniqueness for file naming: the logging script used getpgrp() to form a unique ID for the patch cluster. That works, usually, but not here. See, this repository (being private) was being updated through pserver, and the getpgrp() was always returning the info about inetd's pgrp. Fixed this by generating an MD5 hash of the one constant for a patch cluster, the log message, and using that for the unique ID instead. If you're using the log_accum.pl script from the Apache repository, or any of the others where I've installed it over the years (like Jikes), you're susceptible to this pserver-only bug.

Versionary
Some minor tweaks in here, still pre-alpha. Be more forthcoming about the cause of 'n files here, but none matched the current tag ()' messages -- by reporting missing apps right up front. Also started prototyping correct handling of read-only repositories..

obOpenMouth
I hadn't heard anything from the NordU2002 people about my travel arrangements, so I pinged them yesterday. This morning my mailbox contained a message: 'Oh, there must have been a misunderstanding; make your own travel arrangements and we'll reimburse you after the conference.' Bzzzzt! Not the conditions under which I agreed to go. I don't do 'reimbursement' of airfare or lodging any more; I've been stiffed too many times. Whomever's inviting me gets to pre-pay or direct-bill those. I'm sure not about to put myself in a position to have to eat the cost of a trans-Atlantic flight, even with fares cheaper than they used to be.. I Cced everyone involved on my response, so hopefully they'll get a clew and stick by their original commitment.

post-meridian, EST

obOpen-ness
[No, that's not the anyone-can-play plumbing of the depths of Loch Ness.] Woo-hoo! Released spackle's first alpha to-day! It'll probably break any CVS repository to which it's applied, but that's why it's an alpha and not a beta. :-)

very post-meridian, EST

ApacheRant
mjcox invited me to rant on Apache Week every now and then. Heh-heh. Well, he asked for it. I've rambled a couple of times so far..

obOpenSource
Got some motivation to do a bunch of work on Spackle to-day. The ./configure at least does something useful, even if not complete. Should be ready for a 0.1 release in a few more hours' work..

post-meridian, EDT

obOpenMouth
Well, I'm on the slate to yackity-yack at NordU2002 in Helsinki in February. I'm gonna talk about Synergy in Open Commercialism or some such; at any rate, the middle ground between closed/proprietary software and the militant openness of the GPL.

obOpenSource
Made some progress on Spackle; added the ability to direct the CVS commit messages to different addresses based on a regex match of the files/modules changed. There's still a weirdness interacting with SourceForge's default CVS configuration; there are.. issues.. with a project that frobs the CVS admin files when the repository is on someone else's server. :-D

obOpenFire
About the article about ESR.. oh, never mind. :-/

obOpenSores
jacobo, should I tell you about my misadventures with an automobile battery? Like the one that permanently edited my eyebrows, and involved an oxyacetylene torch? Or would you rather not know? :-)

obApache
We've selected the management company for our ApacheCon events and are working on a contract now. Stay tuned to the announcement list if you care..

health
Don't worry, I won't say anything about it. :-)