Hacker's Diary
A rough account of I did with Emacs
recently.
- October 31
- Humma. My laptop battery is definitely
somehow ill. Or perhaps it's the part of the laptop that does the
charging. Hrm.
New version of XScreensaver
out, with my fix to vidwhacker included. Yay!
- October 30
- Did some work on SEEKRIT PROJECT X which I
may mention if something comes of it.
Discovered that my laptop has an ill battery. Also discovered that
my laptop is under warranty until April 2003, which fact I will
use to cancel the first discovery if it proves that the ill
battery is, in fact, a dead battery. Or if the screen gets any
worse (the base-to-screen connector is not in the best of
shape).
- October 29
- Back to Dublin. For some reason my spam
quotient has gone WAY up lately. Read some more of the procmail
manual to try and fix this some.
- October 28
- More office calls, also more
partying.
- October 27
- Hrm. Somehow my 2.2.18 tree ended up with a
2.0.36 version header file.
Someone sent me a PGP5.0i-encrypted message, which
GPG refused to decrypt,
cryptically (hah!) complaining about a disabled algorithm and then
claiming my passphrase was wrong. Seems Nick Petrely was right
about Open Source developers being sucky with error
messages. Anyway. The problem was/is that IDEA is patented and
thus isn't included with GPG
by default. And bolting it on is pretty trivial, but noone seems
to have actually documented it. So here's how to do it:
- Fetch the GPG
source.
- Fetch some convenient IDEA code; http://www.gnupg.dk/contrib-dk/
has some that will do nicely.
- You did check the signatures on those downloads, right?
- Put idea.c into the cipher directory.
- Apply the following patch to add IDEA into
GPG's default set of ciphers. I've also signed the GPG-IDEA
patch so you can feel safe and warm.
- ./configure --whatever, make and
optionally make install, and you're good to go.
- Don't forget to re-SUID the gpg executable if you
want to get rid of that pesky message about unsafe
memory.
There. Now you're fully IDEA-enabled, and probably breaking some
laws too. Don't you feel rebellious?
For the remainder of the afternoon, I went west, playing with my
GPS toy along the way, and spent the evening partying in
Ballina. And taking some calls from the office.
- October 26
- Embarked on the 386 recovery plan
again. It's rather tricky to build a zImage kernel with enough bits to boot
the machine - I can't use bzImage without a newer version of LILO
on the box. You also find non-obvious interdependencies, like you
can't build SYSV IPC without /proc support being present
also.
- October 25
- Some discussion on the bbdb-info list about
the right way to handle conflicting expansions, i.e. where (say)
Real Name and mail address both match the
stub-to-be-expanded.
- October 24
- Waiting for the nice telecomms people to
show up and fix my damned ISDN line. I predict one of the
following outcomes:
- The guy has to cancel because a previous customer caused him
GBH and he's now on his way to the emergency room.
- I go outside for thirty seconds and return to find "we
called but noone was there to let us in".
- The line checks out perfectly, the equipment checks out
perfectly, but nothing actually works.
- "Dunno. Never seen one o' them before."
Suffice to say I'm not exactly holding my breath in anticipation
of a broadband connection by this evening.
Rebuilding Linuxconf
(finally!) in the interests of updating my m4mailconf source with the
patch for same provided to me some time ago.
I take it all back. I HAVE ISDN!
The explanation, such as it was, that I received was approximately
that when the Big Fault happened at the end of June (yes, June), a
1,200-pair join had to be dismantled, dried out, and remantled. In
the process of remantling, they crossed a few phone lines, caused
some previously good lines to become useless, and broke a few
connections entirely. Somehow, my house seems to have been on the
receiving end of all three of these. Since the ISDN wasn't in use,
noone noticed, so the problem went unreported.
Building Linuxconf proved
to be more fun than I'd anticipated. Wxxt, the windowing toolkit
it uses, doesn't build out of the box with the latest toys - I'm
guessing a more stringent GCC is to blame. So, I've made a Wxxt Compiler Placating
Patch. You'll need to modify the spec file to include this if
you're building from the spec file.
- October 23
- Upgraded Eris (the desktop box at work) to
RedHat 7.2 using
up2date. Remarkably painless; I thought I'd broken
CPAN in the process but it
turned out to be some duff index files on one of the
mirrors. Shame that CPAN
doesn't seem to notice that by itself.
Went to see Paul Brady at Vicar Street; he
brought on Van Morrison for a few songs, which was pretty
cool.
- October 22
- Emacs
21 was released today, which no doubt means more fun for BBDB.
- October 21
- I still cannot for the life of me figure out
why Netscape insists on
trying to do a DNS lookup on localhost. And, of course,
ends up pulling in www.localhost.com. I've got
as far as determining that it's something to do with Proxy
Autoconfig - I think - but beyond that I'm stumped. Perhaps I
should just punt it and go for Mozilla
instead.
- October 20
- This has been a rather unproductive week on
the hacking front.
- October 19
- No really.
- October 18
- Damn these tourists and their
beer-drinking!
- October 17
- D'oh. All I need to do is build a kernel and use sneakernet to put
it on the 386. Screw the bootloader; it's got one already!
And then I couldn't get 2.2.18 to build, either, so I'm trying
2.2.19 now. What happened all my source trees?
- October 16
- Noticed the laptop running a bit sluggishly;
turns out my Netscape
process had managed to clock itself up to 350
MEG.
Hurf. 2.0.36 will not build without some abuse. Turns out that
mkdep coredumps in the fs directory; I snagged a
newer mkdep and used that instead. Then the compile
bombed out with some arbitrary crap I didn't quite
grasp. Argh.
Fixed the Bobbins link in snorq to not care
if it's a GIF or a PNG, and added a link to the latest from jwzrants.
- October 15
- Oh look. Another card from the phone
company, this time for the ISDN line: we visited but couldn't gain
access. No date, no crew number, no phone call to either of my
daytime contact numbers. Perhaps I should try another letter of
complaint, since that at least got me some positive attention the
last time, plus it'd make a nice list of the continuing recurrence
of the above - there's obviously a significant fault with their
CRM system/practices if they keep making the same basic mistake
with the same customer.
Micromail website update,
the second this month. Isn't it time you bought some books?
Added a rule in my procmailrc to catch spams that put a numeric
tag on the subject line.
Hauled down an old, old kernel to see if I can
reconstruct Fedex's boot sequence, and thus determine what's
busted.
- October 14
- Medic! We have a man down!
Tooled around with the old 386
a little, for no good reason. Built a 2.2 kernel which it refused to boot
from. Hmm. It's crashing while reading the disk, and I'm not sure
I want to start debugging x86 assembly language.
More work on BBDB. I
think I'll try and release 2.34 this week, or the weekend by the
latest.
- October 13
- The actual Meal-which-is-not-the-Meal.
More drink, too.
- October 12
- Many DSPs in pub for Gemma's
"Not The DSP meal
2001". Quote of the night from BPC for his perfect timing and
inflection of the phrase "He's Swiss." - you had to be
there, really.
Fiddling with PGP/GPG. Seems like the keyserver web
is a bit messed up.
- October 11
- The ISDN line is live on the system, it
says, but Sales have handed it off to Faults "just in
case". At this point I can't do anything but
laugh.
Mind you, to give credit where it's due, my local head engineer is
being really nice about all this, in a "it's beyond my
control but I appreciate your annoyance" sort of
way.
- October 10
- "Hi, I'm just checking on the status of
an ISDN line that's supposed to be transferred to me... it's been
transferred?...*ten* *days* ago?...uh, thanks."
Of course, it's not actually working, as you might
expect.
- October 9
- Finally got around to upgrading VM to 6.96 (made awkward
by my leave-mail-on-server hacks to the POP module), probably
shortly before Kyle announces the 6.97
release.
Drove to the office with the GPS toy and the laptop with the
mapping and the plotting and muah bwah hah. Obviously I need a
life. Heck, I was even listening to a badly-mixed '80s CD.
Slurped down some data from GIS Data Depot, which has
all manner of fun freebies categorised by country. Of course, you
still have to read up on the various GIS file
formats. Still, I've now got a DTED map of Ireland, whatever
that is. Oh, and roads. Probably dating from about a decade ago,
I'm sure.
Good grief. The DTED
spec is a big freakin' PDF
file full of IMAGES.
- October 8
- Got my in-car PSU for the laptop
today. Muahaahha!
Citizen Kane problem was down to it being a no-region disk, which
my RPC-II DVD drive apparently didn't like. Google, download, flash the
firmware to RPC-I, and we're all happy
again. Whee!
Did some work on BBDB's
completion function. I've sort of implemented it along the lines
of jwz's intertwingle idea, at
least as far as the concept of a unique person being Name + Email
goes. Now I need to optimise it and make sure it at least behaves
approximately like people expect the existing completion to
work. This isn't necessarily a rewrite; it's more of a
verification that the completion works. For now,
anyway.
- October 7
- Humma. Looks like I'll need to grab the Linuxconf source to rebuild
my tarball and SRPM files for m4mailconf.
Looks like the Citizen Kane DVD simply doesn't like my DVD
drive. I'll have to reboot to Windows to make
completely sure, though.
Cleaned up a mass of duplicate files on the laptop, in the
interests of not wasting quite so much space. I should probably
just get a bigger hard drive for it, though.
- October 6
- Fiddling about with Richard Gooch's
devfs in combination with my 2.4.10 kernel. Looks like a neat
idea!
Modified the diary code a little to not insert blank lines unless
necessary (hard to explain unless you actually use the code!).
Modified snorq
to add Last-Modified dates to entries where they're
available.
Discovered a minute bug in RedHat's gpm setup
routine while continuing the devfs fiddling: you can set the mouse
device using DEVICE=/dev/whatever in
/etc/sysconfig/gpm and /etc/sysconfig/mouse, but
the gpm init script does if [ -n $DEVICE ]
instead of if [ -z $DEVICE ].
Ran up Emacs 19.34 to see
what I've broken. Amazingly, BBDB
works.
devfs is indeed pretty tidy, but as yet I'm not going to roll
right into it. In testing what worked, I noticed that Xine was complaining about
being unable to play DVDs. Switching back to a 2.4.9 kernel without devfs doesn't
appear to have fixed the problem, either, so I'm not sure what's
going on there. Might just be the DVD I was trying to run (Citizen
Kane).
Someone sent me a patch for m4mailconf to make it build
under current versions of Linuxconf. I've offered the
entire project to him as I don't use Linuxconf quite as much as I
used to.
Finally managed to get WineX (Transgaming's hack) to run MadOnion's 3DMark2000 toy, by
switching off access to DGA - it seems that something in either
Transgaming's code or the GATOS ati.2 drivers blows up. The
resulting system clocked a whopping... 6 3DMarks. This by
comparison to, um, 500 (I think) in native Windows
mode.
Found a bug in gnuserv, too - gnuserv-running-p always
returns t after you've called gnuserv-start,
regardless of whether the process is actually running or not. tsk
tsk.
- October 5
- Managed to get both Oracle and SMB authorization built into Apache HTTPD, neither of which
is quite what I wanted.
Spent the evening in the pub helping someone I hired at a previous
job drown her sorrows on being made redundant.
And the phone repair man called to the house. Again. At 11:35 on a
weekday, without making any contact with me prior to the call to
say he'd want access to the premises. Now, are these people
lemmings or what?
- October 4
- It seems I'm not the only person who thinks
eircom provide a
less-than-sterling service.
I got a disk in the post from Tony Firshman
containing the contents of my old QL microdrives, including my
fabled 68000 assembler written in SuperBASIC. A comment in the
file indicates that it's version 2, written in December
1991.
Modified VM's MIME
parsing so that it'd correctly display Sircam
messages... damn virus authors don't adhere to standards, mutter
mutter.
- October 3
- Played with Transgaming's WineX for a
bit. It still crashes my X server, after tying it up for a very
long time. I should probably be a good netizen and debug
it.
- October 2
- Whoops. One minor bug in the Micromail update - due to,
would you believe, an omitted period. D'oh.
Phone line still a nuisance. Eircom guy says they're putting
their new magic debug toy on it. I made sure he had both my
cellphone and
office numbers before he hung up.
Persistent networking annoyance: firstly, the aforementioned
failure of ppp-watch to return on my main server;
secondly, the insistence of pump (I think) that the
dial-up connection I made last night is authorative on what should
be in my resolv.conf, despite the fact that the interface
said data came from no longer exists.
Grabbed the kernel patch for
2.4.10 and noticed that it has a new USB-serial driver which may
actually correspond to the toy I have here. Woo!
- October 1
- Added Euro (€) prices to my brother's diving
website. Using some Emacs
lisp, of course.
Also converted Micromail's
site to Euros, which was "slightly" more difficult.
Eircom left a message on my
phone again. The landline. You know, the one I said not
to call during work hours. Smart folk, them.
previous month | current month | next month