Greg
Greg's diary
August 2005
Translate this page
Select day in August 2005:
Su Mo Tu We Th Fr Sa
1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31
Select month:
2005 Jan Feb Mar Apr
2005 May Jun Jul Aug
2005 Sep Oct Nov Dec
Today's diary entry
Diary index
About this diary
Previous month
Next month
Greg's home page
Greg's photos
Network link stats
Greg's other links
Copyright information
    
Groogle

Monday, 1 August 2005 Echunga
Top of page
next day
last day

Meeting day today, involving a rather heated discussion about code quality, descending even as far as hearing opinions that “we don't want no steenking ANSI-compliant code”. Sigh.

Tim Stoakes reported some problems with the build system on Microsoft during the meeting, and later sent me some output. It wasn't pure Microsoft: it was Cygwin, which I have used in the past. Looked like time to try it again, and spent some time installing a surprising amount of software. I still doubt that it can make up for the basic deficiencies of the platform, though.


Tuesday, 2 August 2005 Echunga
Top of page
previous day
next day
last day

More work on the Microsoft breakage today. Completed my installation of Cygwin and rediscovered its strangenesses. For one thing, it doesn't just bend the slashes the right way round, it creates a completely new view of the world. I have a directory C;\home\grog on the box, and Cygwin has a directory /home/grog, but they have nothing to do with each other, and pwd doesn't help. It seems that Cygwin maps each Microsoft “drive” to a file in the directory /cygdrive/.

I was also unable to get Emacs to work correctly: in a shell window, for some reason, the sequence c-x c-c (exit) didn't work; it seems to ignore the c-c altogether. It's also completely unclear how to access the menu bar at the top of the window, and of course using Alt for Meta doesn't work. With some advice, got hold of a native W32 version of Emacs, which worked, but of course trips over the dichotomy of file naming.

Decided that I should really get Samba working again, and installed that on wantadilla, only to run into problems starting it. These proved to come from the new one-interface firewall setup I had installed in June along with the ADSL line.

Managed to fix that half way and then was able to access the sources where I needed them. Discovered that the version of Cygwin that I had downloaded (the newest, obviously) was not compatible with the version we're using in development. In particular, the C compiler is now called gcc (surprise, surprise), whereas it used to be called cl.exe. The old version appears to have a header file direct.h, which I thought had died out in the Dark Ages; now it's called (apparently compatibly) dirent.h. All good changes in my opinion, but it leaves us with the question of how to address our build process.

Scratched my head a bit about that and took up Tim's suggestion to access a development machine in town using vncviewer—but which one? Installed the version in /usr/ports/net/vnc, but it didn't understand the syntax that Tim gave me. Tried deeveear, my SuSe box that is destined to become a DVR, and that worked fine. Still, accessing across a remote link is slower, and the lack of appropriate environment gets on my nerves more and more.


Wednesday, 3 August 2005 Echunga Images for 3 August 2005
Top of page
previous day
next day
last day

More work on the build process under Microsoft, enough to have to recognize that Cygwin has changed enough to break our sources. Since I could, at best, only build and not test, decided to just fix things for the old version we're using. That wasn't helped by the fact that quartz, the only machine with this software to which I have access, went off the net, and nobody could tell me why.

Moved on to my suspended investigation of MySQL, and made some headway on that front. It's been a while since I've used any version of SQL, and either things have changed a lot, or I have forgotten a lot. Still, it's getting comfortable again.

The sale of Telstra is in the news again; I'm sure it's no coincidence that I got a call from Marilyn of Alexander Downer's office, asking if Telstra had taken care of my complaint. They hadn't, of course: my complaint is that the Customer Service Guarantee has such ludicrously low penalties that it's worth their while to offend against it, that I thought that the penalties should be much higher, that that's a matter for legislation, and that Alexander Downer has already told me he thinks they're high enough ($12 a day for loss of phone service. Sure). I also told her I suspected that Telstra would not even honour the CSG this time (because they were less than a whole calendar day over the limit). She's obviously not the brightest spark, and later told me that I would be compensated for the failure, but she did also tell me that the CSG was Telstra's idea (thus, I suppose the low penalties), and that Telstra treated us as part of the Adelaide metropolitan area (which means that the people who handle the complaints are from the metropolitan area; for purposes of the CSG, we're out in the sticks, which gives them more time to fix problems).

In the evening, finally got round to suspending my data projector from the ceiling. Buying a projector isn't normally the end of the story: they're available in 800x600 resolution for as little as $1000, but the cheapest screens cost $250, and mountings cost about the same. I could see no good reason for that, and bought one on eBay for only $80. I was correspondingly concerned that it might be difficult to use, but nothing could be further from the truth: it even came with the correct screws to fit the threads in the projector, and it swivels on three axes. About the only problem I had was that the bracket came with no form of instructions. Still, with a bit of attention, was able to get it mounted nicely:


This should be Projector-mount-1.jpeg.  Is it missing?

Image title: Projector mount 1          Dimensions:          2048 x 1536, 352 kB
Make a single page with this image Hide this image
Make this image a thumbnail Make thumbnails of all images on this page
Make this image small again Display small version of all images on this page
All images taken on Wednesday, 3 August 2005, thumbnails          All images taken on Wednesday, 3 August 2005, small
Diary entry for Wednesday, 3 August 2005 Complete exposure details

 

The key here is to loosen everything and screw to the projector first. Then you turn the swivel mount so that the sides are roughly parallel with the projector body (though this is only important for ease of adjustment and optics), then tighten all the screws.

The mount came with an extensible rod which could mount the projector between 50 and 100 cm from the ceiling, not what I wanted. I was able to mount it without this component, though. There's a basic problem with adjustment that the mounting point isn't above the centre of gravity of the projector, so until it's tightened it hangs at an angle:


This should be Projector-mount-5.jpeg.  Is it missing?

Image title: Projector mount 5          Dimensions:          2048 x 1536, 288 kB
Make a single page with this image Hide this image
Make this image a thumbnail Make thumbnails of all images on this page
Make this image small again Display small version of all images on this page
All images taken on Wednesday, 3 August 2005, thumbnails          All images taken on Wednesday, 3 August 2005, small
Diary entry for Wednesday, 3 August 2005 Complete exposure details

 

Got things mounted nicely in less than an hour, a good time for me. Next step is to route the cables through the ceiling:


This should be Projector-mount-7.jpeg.  Is it missing?

Image title: Projector mount 7          Dimensions:          2048 x 1536, 288 kB
Make a single page with this image Hide this image
Make this image a thumbnail Make thumbnails of all images on this page
Make this image small again Display small version of all images on this page
All images taken on Wednesday, 3 August 2005, thumbnails          All images taken on Wednesday, 3 August 2005, small
Diary entry for Wednesday, 3 August 2005 Complete exposure details

 

All in all, an expenditure of $80 for the mounting and $20 (a bucket of paint) for the screen, instead of $500 if I had done it the normal way. I'm relatively satisfied.


Thursday, 4 August 2005 Echunga
Top of page
previous day
next day
last day

The Microsoft machine on our network is back (we had another ADSL2+ hang), so did more work on the build. What a lot of time it takes! Apart from the pain of using the “tools”, there's the compilation speed itself: on the Microsoft box, a 866 MHz Pentium III, it takes nearly 15 minutes; on my ancient 400 MHz Apple laptop, it only takes 6 minutes. Why does it take such a long time to compile things on Microsoft platforms? Here's the time the build took on four different platforms:

Platform Total time User Sys total * clock speed
FreeBSD, 2.2 GHz Athlon 84.47 46.81 9.04 185.8
Linux, 2.2 GHz Athlon 80.59 53.51 5.18 177.3
Apple, 400 MHz PPC G3 368.60 293.84 59.89 147.4
MS, P3, 863 MHz, 384 MB 885.72 155.97 180.80 764,4

In this table, the last column represents a badness factor: the highest value loses. The other three are relatively close; why is Microsoft a factor 4 to 5 slower?

The problem itself was a leftover of our old build process, and was relatively easy to fix. But what a waste of time!


Friday, 5 August 2005 Echunga
Top of page
previous day
next day
last day

After all the work I did on AUUGN, I still had a bug! Of course, it's Microsoft's fault: it seems that the ISO 9660 standard allows symbolic links on ISO images, but Microsoft doesn't understand them. And I put in one symbolic link, where it shows. Here's a proposed apology to Microsoft users in AUUG (though I still have a certain feeling “serves you right”):

This DVD went through a number of tests before mastering, but one “error” slipped through: it contains one symbolic link, and Microsoft does not understand them. It works fine on all versions of UNIX we have tried it on.

The symbolic link is:

   lr-xr-xr-x  1 root  wheel  58 Jun 25 13:08 ./Lions -> AUUGN/hackers-diary/www.lemis.com/grog/Documentation/Lions

This means that the links to the Lions Commentary will not work under Microsoft. You can still access them on the DVD as AUUGN/hackers-diary/www.lemis.com/grog/Documentation/Lions/index.html.

Why this link? This material has been on my [Greg's] web site for ever, and my diary refers to it. My "hacker's diary" made it to the disk before the commentary, so to save space I put the commentary in as a symbolic link.

The Telstra bill came today. As I had predicted, there was no mention of the refund for the Customer Service Guarantee. Called up the phone line and spent 25 minutes being transferred from Leanne in enquiries to Leanne in billing (another one) to Swat (yes, that's how she spelt it for me; I wonder if they have a list of names to give), who told me that from 21 June to 24 June was within the 2 day service guarantee. I was able to convince her that it was not, and she put me on hold; came back and said that it could sometimes take 4 to 6 weeks for the refund to come through (why?), so it might be on the next bill. The bill was issued on 2 August, more than 5 weeks after the incident; my suspicion is that they're trying to assume that the guarantee doesn't cut in until at least 24 hours after expiry. Still, I've made enough noise about this one, and I'm sure I'll get a refund of at least $12 and possibly $36; still a poor return on investment, and I suspect others who were affected will not be so lucky. I wonder if they will be compensated at all if they didn't register a fault.

Spent all day working on the build infrastructure, and finally got most of it committed. What a can of worms, and what a mess this Microsoft stuff is! Got a message from James Andrewartha suggesting a cause of the performance problem I reported yesterday:

The only windows build I've done is Mozilla, so I'm not really an expert, but possibly a reason why it's so slow is windows is not particularly fast at spawning new processes, and cygwin has to do a bunch of emulation to provide a real fork(). Since make spawns lots of short-lived child processes, this overhead quickly adds up.

That sounds reasonable, especially since the big discrepancy seems to be in system time.


Saturday, 6 August 2005 Echunga
Top of page
previous day
next day
last day

60 years since the US military destroyed Hiroshima. How the times have changed. In those days, they were the “liberators”, though even then there were doubts about the necessity. Nowadays people are more critical of the US military, with good reason.

In the morning, finally started painting a screen on the wall for the data projector. I ended up with an area of 284x160 cm, with a diagonal of 325 cm. eBay doesn't have anything that big: the biggest are 120 inch (inch? What's that? Why don't they say 300 cm?) screens, which go for $500 and up. My solution would have cost about $20. Except: in the process I discovered that the surface I was painting on hadn't been prepared properly, and much of it flaked off. One black mark for Adrian Mortimer. Now I'm left wondering what to do with the rest of the wall

Out riding after lunch. Finally we're getting into more of a routine. In the last 5 weeks we've gone riding as many times as in the previous 8 months. Darah is getting into better condition, and it was very pleasant.

After that, more attempts to install software on my SuSE machine. My first impressions may have been overly optimistic: there are some really strange things about installing software with yast2, and today I ran into a problem that I had already had with Debian: I need to install qt3 for make xconfig for the kernel install. The difference was that I managed to do so with Debian. SuSE didn't offer the qt3-devel package, and when I downloaded it from somewhere and installed it with yast2 -i, it pretended that all was well, but it still doesn't show up in the list of installed software, and make xconfig still complains about it being missing. I can find some stuff in /usr/lib/qt3, but it doesn't seem to be what the build is looking for. What a pain!


Sunday, 7 August 2005 Echunga Images for 7 August 2005
Top of page
previous day
next day
last day

Brew day today, which as usual kept me going half the day. After that, I had intended to get back to work on the SuSE box, but for some reason was tired, so didn't do much.

In the evening we ate garfish fillets. Rather than just fry them, I thought I'd marinate them and then barbecue them. I used a Chinese-style marinade with garlic, spring onions, sherry (“rice wine”) and sesame oil. They tasted terrible! Obviously not the right match for this particular fish.


Monday, 8 August 2005 Echunga
Top of page
previous day
next day
last day

Meeting day again today, and got even less done than usual: had to go and talk to my financial advisor after that, so didn't get back home until 15:30.

On the way into town, got a phone call from Telstra's complaints department. They won't pay anything for being over time, because of the floods we had at the time. What a load of nonsense! The photos I took at the time clearly show that, though it had been raining, there was no hint of flooding. Looks like another letter of complaint. Maybe we should get the TIO involved in this one.

Examining the wall on which I painted the data projector screen was depressing: just about all the paint is peeling off, so we'll have to start all over again. sigh.


Tuesday, 9 August 2005 Echunga
Top of page
previous day
next day
last day

Finally got round to doing some work on MySQL today. In the process discovered that I needed some data, so went looking for old data collections that I had maintained with Monkey 10 years ago. Spent some time trying to get it to compile with modern C++ compilers, and ultimately succeeded, though I'm left wondering whether what we have now represents progress. The following code used to be legal:

    {
    if ((validate = get (MONKEY_VALIDATE_SCREEN))           /* we have a validate function  */
        && ((*validate) (this, f) < 1) )                    /* failed screen validity check */
      return FALSE;                                         /* not valid */

get is a function returning void * (modeled on the LISP get function). As the code shows, validate is a function that takes a couple of parameters. Nowadays I have to write

    {
    if ((validate =
         (int (*) (Window *, DataField *)) get (MONKEY_VALIDATE_SCREEN)) /* we have a validate function  */
        && ((*validate) (this, f) < 1) )                    /* failed screen validity check */
      return FALSE;                                         /* not valid */

Does this really improve code quality? There's only one cast that can work, and it's implied by the type of validate.

Called up Telstra about yesterday's claim of flooding, and was told they'd call me back. They didn't.


Wednesday, 10 August 2005 Echunga
Top of page
previous day
next day
last day

More work on SQL today, and modified Monkey to export data to MySQL, which was easier than I thought. There might have been easier ways to do things, but this way I'm building up experience more quickly.


Thursday, 11 August 2005 Echunga
Top of page
previous day
next day
last day

More work on MySQL, with little to report.

In the afternoon to another ICT Council for South Australia board meeting. At least we got through it this time.


Friday, 12 August 2005 Echunga Images for 12 August 2005
Top of page
previous day
next day
last day

Now that I have the basics of MySQL sorted out, it made sense to try some of the administrative accessories, so got some information from Arjen Lentz on a the freenode #MySQL IRC channel. He suggested the use of MySQL Administrator and MySQL Query Browser . The first was easy enough: it's in the FreeBSD Ports Collection, and it installed and worked well.

Not so the Query Browser: it's not in the Ports Collection, so I decided to port it. On the face of it, that's not too complicated, but I had lots of trouble with GtkHTML, part of GNOME. Query Browser depends on it, but FreeBSD provides several packages, and it's not clear whether to install /usr/ports/www/gtkhtml and /usr/ports/www/libgtkhtml or something else.

The simple answer would be: install both, and I did. But Query Browser still couldn't find what it was looking for. Spent far too much time looking at that: the big issue is that the configure script doesn't tell you what it's looking for. Again I had to use ktrace to find what was going on:

 52433 sh       CALL  stat(0x806d53c,0xbfbfd6b0)
 52433 sh       NAMI  "/usr/X11R6/etc/libgtkhtml-3.0Conf.sh"
 52433 sh       RET   stat -1 errno 2 No such file or directory
 52433 sh       CALL  stat(0x806d53c,0xbfbfd6b0)
 52433 sh       NAMI  "/usr/local/etc/libgtkhtml-3.0Conf.sh"
 52433 sh       RET   stat -1 errno 2 No such file or directory

After that I had to find which package includes a file with a name like /usr/local/etc/libgtkhtml-3.0Conf.sh:

=== root@wantadilla (/dev/ttyp3) color="red">/usr/ports/www color="blue">636 -> grep Conf.sh *gtkhtml*/pkg-plist
gtkhtml/pkg-plist:etc/gtkhtmlConf.sh

That also gave me the port name I should be looking for (i.e. gkthtml, not the default gtkhtml-3.0). Finished that (and the day) to find the next error message:

checking for libglade-2.0
                        gthread-2.0
                        libxml-2.0 &gt;= 2.6.2
                        no
                        gtkmm-2.0...  Package no was not found in the pkg-config search path.  Perhaps you should add the directory containing `no.pc' to the PKG_CONFIG_PATH environment variable No package 'no' found

This strange format could suggest it's looking for any of these things; in fact, it's the next dependency, gtkmm. What a pain this GNOME stuff is!

In the evening, while watching TV, decided to look at some PDF documents. It turned out that teevee.lemis.com, my projector machine, doesn't have acroread, so tried to download it. That, too, failed with dependency problems. I fear that the issue of software installation and maintenance is going to make or break “Open Source” software. Maybe that's the real reason why people are migrating to Apple, despite its horrible user interface: at least software updates work.


Saturday, 13 August 2005 Echunga Images for 13 August 2005
Top of page
previous day
next day
last day

Brew day today, which kept me busy all morning. After that attended to the projector setup in the lounge room: repainted the screen, which looks worse than before—I think I should have called in a professional. Then up into the ceiling to investigate how to route cables through the ceiling, missed my footing and came down the fast way:


This should be Ceiling-1.jpeg.  Is it missing?

Image title: Ceiling 1          Dimensions:          1728 x 2304, 304 kB
Make a single page with this image Hide this image
Make this image a thumbnail Make thumbnails of all images on this page
Make this image small again Display small version of all images on this page
All images taken on Saturday, 13 August 2005, thumbnails          All images taken on Saturday, 13 August 2005, small
Diary entry for Saturday, 13 August 2005 Complete exposure details

 

Not an experience I would care to repeat. Picked myself up off the ground in not the best of conditions, and considering a prior history of severe concussions, called an ambulance, which took me off to the Royal Adelaide Hospital, where I had been treated for my last concussion in December 1998. Then I had fallen off a horse and was in a bad state, and they had treated me very well, quickly and efficiently. This time it was obvious that I wasn't so badly off, and nobody came to attend to me for 1½ hours, by which time I was feeling relatively normal, so back home. What a waste of an afternoon!


Sunday, 14 August 2005 Echunga Images for 14 August 2005
Top of page
previous day
next day
last day

Feeling somewhat sore this morning, so decided to take it easy. Somehow that didn't work, and spent a couple of hours overhauling my brewing software, without coming to a good conclusion about how to handle temperature ramps in a way that doesn't shoot you in the foot.

I have a large collection of cookery books, but few good Indian books. I've decided that photos in cookbooks are good—when they correspond to the recipe. On Monday I bought a few cookbooks in town, and today I tried a recipe for keema kofta, lamb meat balls in nut gravy, from a book entitled “Authentic Recipes from India” from Periplus publishing. The recipe looked good at first glance (or I wouldn't have chosen it), but it shows several of the worst aspects of such books:

In the end it didn't taste too bad. I've written up my own version of the Keema Kofta recipe.


Monday, 15 August 2005 Echunga
Top of page
previous day
next day
last day

As a result of my injuries from Saturday, missed the Monday meeting today. That might have been just as well for other reasons: while reading my mail, there was suddenly a loud noise, and all my machines went dead. Went round the back of the desk to find the UPS making some strange dying noises, so turned things off and investigated the damage, thinking horrible thoughts about dead disks: all my backups are cross-backed up between two of the machines (echunga.lemis.com and wantadilla.lemis.com) that had gone down.

After bypassing the UPS, echunga came up well enough modulo a bad sector on the system disk. There's no particular reason to believe that that was related to the failure. wantadilla was a different matter: dead. It didn't take long to confirm that the power supply had died. Put in a spare, which proved to have a non-functional fan, so plundered deeveear.lemis.com, the SuSE 9.3 box I'm (slowly) setting up as a DVR. After that, all went well, though this sort of situation does highlight bugs in the startup procedure.

Spent some time investigating power supplies. Clearly I need to have a couple spare. My supplier offers power supplies with suitable specs at prices ranging between $9 and $690. What to choose? Only one has a 2 year warranty; the rest are only 1 year. I'm left feeling that power supplies have been left out of the ACPI spec. There's also the (currently unanswered) question as to whether there's anything wrong with the UPS, or whether it was just dealing with an overload situation.

How I hate “Internet Banking”! Why is it that big banks have to make things so complicated? Sure, security is one thing, but navigating the screens on ANZ's Internet banking system—which doesn't even have its own URL!—is painful. And to add insult to injury all the statements are in reverse chronological order. That's all the more stupid because you can't change it, and the printed statements are the right way round. Called the bank and got told that there was nothing, but nothing, they could do about it. I suggested that telling a programmer to fix it might be a good start, but he insisted that it was impossible. Got him to forward me to his superior, who was conciliatory and understanding, but of course she couldn't do much either. At least she promised to forward it to the planners.


Tuesday, 16 August 2005 Echunga
Top of page
previous day
next day
last day

In “pick up the pieces” mode today. Did some investigation of condition of echunga's system disk, and discovered four or five areas of unreadable data. Since it's only an old 20 GB drive, decided that it was time for replacement. It seems that the “sweet spot” for drive prices is now round 200 GB, so got Yvonne to pick up one of those along with some power supplies. That left me with the usual dilemma: how to use it? The old disk wasn't full:

=== root@echunga (/dev/ttyp9) ~ color="blue">99 -> df -l
Filesystem  1048576-blocks   Used  Avail Capacity  Mounted on
/dev/ad0s1a           8921   7667    539    93%    /
/dev/ad0s1d           8569   3844   4039    49%    /home
/dev/ad1h            76285  67713   7809    90%    /dump
/dev/ad2h           150214 106037  32160    77%    /src

The other disks are used for backups (/dump) and my main source disk (/src). /dump goes through periods of overflow, so it seemed a good idea to replace it with something bigger. But putting it together with the / and /home file systems on the 200 GB drive would still be overkill, so decided to move the contents of /src to the new disk and then divide up the old /src for the three remaining file systems. It's also interesting to note that a typical size for the root file system (including /usr, and particularly /usr/local and /var) should probably now be closer to 12 GB than 8 GB.

Moving the data (with tar) wasn't as straightforward as I thought. First, it's a lot of data, and secondly (of course shortly before the end, after several hours), I found:

=== root@echunga (/dev/ttyp4) /src color="blue">24 -> tar cf - .|(cd /mnt; tar xf -)
archive_write_pax_header: 'x' header failed?!  This can't happen.

After that it just continued writing to the disk, but not storing any data. Maybe it was rewriting existing data. Spent some time considering how to handle it, and tried rsync, but rsync isn't suited to very large quantities of data, so split it up into chunks:

for i in */*/*; do echo $i; rsync -av /src/$i /mnt/`dirname $i`; done

That worked, but the problem is that it ignored too many files. Decided that, rather than attempt to fill in the gaps and to discover too late that I had not got them all, I'd start again tomorrow using GNU tar.

While waiting for all that to happen, spent some more time looking at porting MySQL Query Browser to FreeBSD. Spent far too much time looking at the dependency on gtkmm, including installing the gtkmm package. It didn't impress the MySQL Query Browser configure scripts, though, which were specifically looking for version 2.0. What a pain these dependencies are!

On the topic of pain, followed up an article in issue 15/2005 of c't discussing using apt for SuSE 9.3, mainly a reference to the gwdg site. That installed nicely and seemed to work. Then I went looking for rwhod, which is part of the base system on FreeBSD, but which didn't get installed on the SuSE box. It wasn't found. Did some investigation and discovered that it is, indeed, available on the Debian sites, but the version of apt I installed from gwdg doesn't handle .debs, only RPMs. What a pain! To add insult to injury, yast2 appeared to be hanging when I tried to access the network. Somehow I managed to get a large number of copies of the same external site information in the system, and I suspect it was trying to contact it each time. Wouldn't it be nice to know how to edit the list underneath yast2. I'm sure it's possible, and I'll find out how to do it, but it's a pain to have to.


Wednesday, 17 August 2005 Echunga Images for 17 August 2005
Top of page
previous day
next day
last day

Somehow mail kept me going for ever today, and I still didn't get my hardware upgrades sorted out. GNU tar did the trick copying my /src disk, but of course at the end of it I had the possibility that files on /src had been updated, and set to getting mklinks to list what changes it would make. That was longer than I thought: 45 MB of messages, most of them relating to broken symlinks. On the face of it, that's an error, but it's surprising how many symlinks are broken by design, and weeding out the rest is not easy. Along with the latency of large disks, that took another day.

Revisited the SuSE 9.3 software installation issues today, and as expected got at least a little further, though I'm obviously going to have to learn more about apt. It turns out that SuSE no longer supplies rwhod and friends—why?—and I had to pull down a Debian .deb package, which my installation of apt can't handle because it hasn't been told about .debs. Instead used dpkg-deb to extract it to the right place, bypassing all the Debian installation procedures. Definitely something I need to learn more about: the standard SuSE installation methods may be nice and easy to use, but they're limited in what they offer. On the other hand, had no difficulty installing MySQL Query Browser, a significant difference from the pains I've been having under FreeBSD.

In March I spent more time than I had intended in Marree while changing destroyed tyres. While we were waiting, Dad told me about Sidney Kidman, the cattle king, and shortly later I found a plaque:


This should be Marree-5.jpeg.  Is it missing?

Image title: Marree 5          Dimensions:          1728 x 2304, 448 kB
Make a single page with this image Hide this image
Make this image a thumbnail Make thumbnails of all images on this page
Make this image small again Display small version of all images on this page
All images taken on Saturday, 26 March 2005, thumbnails          All images taken on Saturday, 26 March 2005, small
Diary entry for Saturday, 26 March 2005 Complete exposure details

 

Now it seems that, like the Neanderthal Canadian we met the same evening, Kidman has come to the Adelaide Hills, this time as a horse riding track, the Kidman Trail. It's intended to go from Willunga, at the tip of Cape Jervis, to at least the Barossa, but currently it only goes as far as Echunga, where we live, and current part goes within a kilometre of our house. This evening to the Macclesfield Institute to hear a talk about the proposed next stage, which it seems will double round to the other side of our house before heading to Macclesfield. We're particularly happy because it will mean a long uninterrupted ride trail, something that's difficult to find anywhere in the world. The route is not yet cast in gravel, and that was the purpose of the meeting, to find out whether we can improve on it. Spent some time doing that. Looks like we have something specific to do on our next rides.


Thursday, 18 August 2005 Echunga Images for 18 August 2005
Top of page
previous day
next day
last day

I'm still working on upgrading my machine. Sometimes I wonder why it takes me so long when others can do it faster, but I suppose the answer is fairly clear: I want to do it right, and doing it right rather than recovering from inevitable problems does take longer. That doesn't mean I don't make mistakes, but they're getting more predictable.

Started moving my dump data to the new disk. It's refreshing to see how fast you can move bulk data nowadays, even on el-cheapo IDE disks. Here's iostat output from drive ad1 to ad3 (both slaves on different controllers), showing that I'm copying at over 30 MB/s:

      tty             ad0              ad1              ad2              ad3             cpu
 tin tout  KB/t tps  MB/s   KB/t tps  MB/s   KB/t tps  MB/s   KB/t tps  MB/s  us ni sy in id
   0 1592  9.50   4  0.04  128.00 267 33.42   0.00   0  0.00  127.59 271 33.81  18  0 32  2 48
   0 1162  0.00   0  0.00  125.80 259 31.88   0.00   0  0.00  127.56 250 31.09  15  0 39  1 46

Also still trying to install qt3 properly on deeveear. Even with the help of apt, that's not as smooth as it should be:

=== root@deeveear (/dev/pts/0) color="red">/usr/src/linux color="blue">23 -> apt-get install qt3-devel
Reading Package Lists...  Done
Building Dependency Tree...  Done
Some packages could not be installed.  This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.

Since you only requested a single operation it is extremely likely that
the package is simply not installable and a bug report against
that package should be filed.
The following information may help to resolve the situation:

The following packages have unmet dependencies:
  qt3-devel: Depends: freeglut-devel but it is not going to be installed
E: Broken packages

Installing freeglut looked pretty much the same:

=== root@deeveear (/dev/pts/0) color="red">/usr/src/linux color="blue">24 -> apt-get install freeglut-devel
Reading Package Lists...  Done
Building Dependency Tree...  Done
Some packages could not be installed.  ...

The following packages have unmet dependencies:
  freeglut-devel: Depends: freeglut (= 2.2.0-83) but 2.2.0-83.1 is to be installed
E: Broken packages

I suspect I've got my distributions mixed up. But if that's the case, it's too easy, considering that I did a complete installation and deliberately tried to stick to the rules.

Still, not everything was so difficult. I've heard from the author of kscope, which I gave some negative publicity last month, and so I owed it to him to try it out again. Installation was a breeze. Unfortunately, it didn't run. At this stage it's not clear whose fault that is, but it did run on FreeBSD. Spent some time looking at that, coming to the conclusion that I still don't like kscope, but that possibly the main reason is because it's a KDE application.


Friday, 19 August 2005 Echunga
Top of page
previous day
next day
last day

Today I finally cut over to the new echunga system disk. It was supposed to be easy, but didn't quite work out that way: I had misjumpered the disk and also used the wrong options for tar when copying the disk. Note to self: do not use the o option for copying disks. And I should really think of making an addition to chmod to understand the permissions output by ls -l.

Into a very wet Adelaide for the ADUUG lunch, spent discussing ADSL. Looks like we can expect further speed improvements in the near future.


Saturday, 20 August 2005 Echunga
Top of page
previous day
next day
last day

Quiet day today. Spent the morning bottling beer, then wrote a first cut at a filter program to set permissions and ownerships based on the format output by ls -l. Typical input is:

=== grog@wantadilla (/dev/ttypd) color="red">~/src/setperm 60 -> ls -l
total 1
-r--r--r--  1 grog  lemis    68 Aug 20 15:09 Makefile
drwxr-xr-x  2 grog  lemis   512 Aug 21 11:56 RCS
-rwxr-xr-x  1 grog  lemis  6821 Aug 20 15:22 setperm
-r--r--r--  1 grog  lemis  5267 Aug 21 11:56 setperm.c
=== grog@wantadilla (/dev/ttypd) color="red">~/src/setperm color="blue">61 -> ls -l | ./setperm 2>/dev/null
chmod 444 Makefile
chown grog:lemis Makefile
chmod 755 RCS
chown grog:lemis RCS
chmod 755 setperm
chown grog:lemis setperm
chmod 444 setperm.c
chown grog:lemis setperm.c

It's not much use like that, of course: I could pipe the stream to a shell, and it would set the ownership and permissions to exactly what they already were. But with a little massaging of the input (probably) or output stream, it could be used to reset the permissions of whole directory trees.

That's only the first cut, of course. This functionality belongs in chmod as well. I'm just wondering how best to do it. My current feeling is that it should look something like:

chmod -p -rwx-r-x-r-- foobar

Apart from that, didn't do much.


Sunday, 21 August 2005 Echunga
Top of page
previous day
next day
last day

Another quiet day. Spent some time investigating Wikipedia, with the surprising discovery that it's really quite good. I've been reading the article on Indo-European languages in Encyclopedia Britannica, and found it rather dry and uninformative going. By contrast, the information in Wikipedia more detailed and better structured. One interesting thing I found was a reference to Schleicher's Tale, a fake proto-Indo-European story. It's not found in the Encyclopedia Britannica. Of course, the version I did find was barely legible:

Click on the picture to see a medium-size version in the index

Unfortunately I can't find any way to contact the author.

That's not Wikipedia's fault, of course. I'm left wondering why it works so well. I've always had the impression that “Open Source” works well for simple things, but why should it be able to overtake the Britannica so easily?


Monday, 22 August 2005 Echunga
Top of page
previous day
next day
last day

Meeting day again today; it's amazing how a relatively short meeting can completely disrupt the day. It should have been longer today, but I had forgotten that I was supposed to be doing a tutorial on gdb, and I wasn't prepared. Next week.

After that to Dymocks to buy a book on MySQL, and on the way out saw a book on “The Languages of the World”. A bit of a toy, and I wasn't really interested in it, but browsed through it anyway and found a (legible) copy of Schleicher's Tale. Funny coincidence.

Also to Grumpy's to buy a grain mill. It was made in Italy, originally intended for kitchen use, and looks remarkably like my pasta machine. I hope it'll be more durable; Thomas says that he's been using his for years, so there's hope.

In the evening another power failure. Now that I have no UPS in place, lost just about everything. Somehow keeping computers running is still an issue.


Tuesday, 23 August 2005 Echunga
Top of page
previous day
next day
last day

Spent most of the day working on my gdb presentation next week. It's not as if I don't have preparation: it's been ten years since I started writing a book on the subject, but stopped working on it before returning to Australia. Spent most of the day bringing enough of it up to date to make it a worthwhile basis for Monday's presentation.

That in itself was a challenge. The book was originally intended for publication by O'Reilly, and at the time they used groff with the ms macros. Now I use my own evolution of the mm macros, and I've also changed the Makefile structure significantly since then. The results were more functional than pretty, and today I spent most of my time tidying things up, with surprisingly good results. At least I have some properly formatted texts.


Wednesday, 24 August 2005 Echunga
Top of page
previous day
next day
last day

More work on book production today, with good results. It's tiring, though. No wonder I left it in a mess for ten years.


Thursday, 25 August 2005 Echunga
Top of page
previous day
next day
last day

This morning Yvonne left for Broadford, of all places, to participate in an equine behaviour seminar at the AEBC, leaving me all alone in the house for the first time in quite a while: on prior occasions, Yana was there, but she's been living in Adelaide for 6 months now (and chose today to move from one flat to another).

More work on the documentation Makefile today, bringing back forcibly how the first 99% takes up 99% of the time, and the second 1% takes up the other 99% of the time. Still, it's a thing that I've been kludging around for ever, and it seems to be making progress.


Friday, 26 August 2005 Echunga Images for 26 August 2005
Top of page
previous day
next day
last day

Yvonne's away in Broadford, so I have to run the household. This morning I had to run the dishwasher. But where was the dish washer detergent? I snooped around and found this:


This should be whatsthis-1.jpeg.  Is it missing?

Image title: whatsthis 1          Dimensions:          1536 x 2048, 416 kB
Make a single page with this image Hide this image
Make this image a thumbnail Make thumbnails of all images on this page
Make this image small again Display small version of all images on this page
All images taken on Friday, 26 August 2005, thumbnails          All images taken on Friday, 26 August 2005, small
Diary entry for Friday, 26 August 2005 Complete exposure details

 

Looks like detergent, right? Yes, but there's no harm in being sure. But there's no way to be sure: neither on the front nor on the back was there anything saying what it was. Well, no, that's not quite correct: it clearly states that it is “finish PowerPoder”. But I was looking for something in English. On the back it states: “Finsh Powder [and not PowerPowder] is a unique enzyme-based formulation providing a powerful clean and sparkling shine. First time. Every time.”

Ignoring the bullshit (“First time. Every time”. Do people really believe that? Would the manufacturer guarantee that claim?), it's reasonable to assume that it's a cleaner, even if they don't say so explicitly? What kind? It's obviously dangerous, since there's a caution note written in ALL CAPS TO MAKE IT EASIER TO READ: “IRRITANT. AVOID CONTACT WITH EYES & [sic] SKIN. DO NOT SWALLOW. WASH HANDS AFTER USE”. While researching this diary entry, I also came across further proof of this problem, mentioning this particular powder. Even this article, though, doesn't seem to thing that clear labelling is important.

Clearly “open source” web sites aren't the only thing that are missing the basics. In this case, the combination of danger and lack of description make it seem particularly irresponsible to not describe what it is. What if it's some specialist powder designed only for specific types of cleaner? I called the number on the back (1 800 022 046) and spoke to Linda, who confirmed that it was, indeed, dish washer powder, apparently of a general-purpose kind (obviously; we wouldn't be using it otherwise). But she wasn't able to tell me whether they would do anything about getting it clearly marked (they're “not obligated to provide information”), and she couldn't tell me what regulatory body is responsible for clear markings. Agreed with her that I would wait a few days before filing a formal complaint.

Afterwards I took a look round the other chemicals in the kitchen. Most were clearly marked, particularly the cheaper generic ones. I suppose companies like Reckitt Benckiser, a company that doesn't appear to want to be found on the web, think that marking them in English would detract from their marketing gobbledygook. Doubtless it would; but is that a disadvantage, even for them?

More work on the debugging tutorial. My Makefile still isn't the way I want it, but it seems a lot cleaner now, and I'll have time to do it right later (isn't that what they always say?).


Saturday, 27 August 2005 Echunga
Top of page
previous day
next day
last day

Quiet day today: up early, to bed late, and didn't do much in between.

Web browsers will drive me mad! It all seems to be part of a plot to dumb down software. In the last few months I've been using firefox, without really understanding the hype. I've already stated my opinion on “tabbed browsing” and the destruction of the UNIX directory hierarchy paradigm, but lately I've been seeing more annoying things: for example, when trying to view diffs from Wikipedia, I see:

image

Well, that's what I see on tvremote, my TV remote control machine. For reasons I haven't investigated, wantadilla, my main machine, claims:

image

These are two different versions of firefox, but that shouldn't make a difference in the kind of data page.

The rest of these examples are from tvremote. Of course, firefox doesn't know anything about the PATH environment variable, and starts me off looking with a tiny window pointing into my home directory—as if the directory where I would store data would also be the directory where I keep standard programs. Even the alternative options (drop down window at the top) just go back up to the root directory and don't show any directories with executables, nor any of the directories in PATH:

image

Still, maybe it would understand if I just type in the name of the program, as shown above? After all, it has inherited PATH from its parent. But no such luck:

image

So I type in the fully qualified path name, something that became obsolete at least 30 years ago:

image

Now it finds it—and calls it emacsclient!

image

Why is there such nonsense? Even Microsoft's MS-DOS introduced a path variable as soon as it got more than one directory, over 20 years ago. Are people just trying to make things stupid? Sadly, I haven't found any browser that doesn't have this kind of nonsense.

It's interesting to note that User Friendly has also commented on the popularity of firefox.


Sunday, 28 August 2005 Echunga
Top of page
previous day
next day
last day

Had another power failure in the night, but at least I had put echunga and wantadilla back onto (separate) UPSs. Came into the office and found echunga running fine, but wantadilla was powered down. On further investigation discovered that the UPS has a total of 6 outputs, but 3 of them are “bypass”, in other words not battery backed. And of course I had to select one of them.

Apart from that, another quiet day. Spent most of it finally writing up my paper for AUUG 2005, which was none too early, since the deadline for submissions has already closed.


Monday, 29 August 2005 Echunga
Top of page
previous day
next day
last day

Back to work on document build process today. It's taking shape, and it's good to get rid of the kludges of the last 10 years, but it's also a good reminder that the first 90% of the job takes 90% of the time, and the other 10% takes the other 90% of the time (actually, it's more like 99% and 1%). The documents I had to show this evening looked worse than last week, but that should change rapidly.


Tuesday, 30 August 2005 Echunga
Top of page
previous day
next day
last day

Lots of talking on the phone today, and as a result didn't get much else done. Did manage to do some work on my documentation, but not as much as I had hoped.

The weather's been bad lately, and we had more power cuts. They didn't hit my main machines, but the switch and ADSL modem kept cycling. Took another look at the UPS that I thought had died two weeks ago, and discovered that in fact it had just blown the mains fuse. Spent some time looking for suitable fuses, but of course they were all the wrong size or rating, and ended up plundering dead equipment for something that seems to do the job. Maintaining power integrity is just too much work.


Wednesday, 31 August 2005 Echunga
Top of page
previous day
next day
last day

Woke up in the middle of the night and turned round to look at the time: there was none. The power had failed again. Cursing, out through the sound of beeping UPSs to start up the generator, after which I discovered that power had been out for 6 hours, so of course I had lost every single machine in the house. Spent some time tidying things up—there are too many loose ends round here—before going back to bed and sleeping late. The power was out for a total of 7½ hours:

Aug 30 22:55:44 wantadilla kernel: xl0: link state changed to UP
Aug 30 23:00:12 echunga kernel: nfs server wantadilla:/: not responding
Aug 30 23:00:43 echunga kernel: nfs server wantadilla:/: not responding
Aug 30 23:02:47 echunga last message repeated 4 times
Aug 30 23:06:55 echunga last message repeated 8 times
Aug 30 23:07:06 echunga ntpd[524]: sendto(203.21.37.18): No route to host
Aug 30 23:07:26 echunga kernel: nfs server wantadilla:/: not responding
Aug 30 23:07:57 echunga kernel: nfs server wantadilla:/: not responding
Aug 31 05:08:07 echunga syslogd: kernel boot file is /boot/kernel/kernel

I don't monitor power state directly, but clearly wantadilla must have died some time between 22:55:44 and 23:00:12; the power must have failed some time before that. It finally came back at about 6:45 am (not mentioned here because the system was on the generator).

Still more work on documentation Makefiles today, and finally got most of the structure working, though doubtless there will be more when I do things like review draughts and text versions. But I've broken its back.

I've had enough of firefox. I want a browser that I can use, not spend all my time trying to work out how to install which plugin. The typical situation is with Macromedia plugins. Following the link, you get:

image

(All these images are reduced; click on them to get an idea of the original size, and recall that they're displaying on a 2048x1536 display with about 140 dpi). So I installed Opera in the hope of it working better. It didn't. Even on startup, I get:

image

That's obviously an installation problem; but isn't that what the Ports Collection is there for? It doesn't seem to stop it from working, except for Java, but when I try the Shockwave plugin, I get only a little further:

image

Of course, the table of recommended “Web players” has nothing useful. It's also not clear how to set the font sizes on Opera: I set a minimum of 20 pixels in the preferences, yet the headings and the URL window are much smaller:

image

Those letters must be about 10 pixels, and on a 2048x1536 display they're legible, but barely and uncomfortably.

In the evening, restarting the TV system was interesting. It had been up for two months, during which time I had done some tweaks to the X configuration, notably specifying the screen dimensions:

screen #0:
  print screen:    no
  dimensions:    1280x736 pixels (2956x1699 millimeters)
  resolution:    11x11 dots per inch

Starting firefox gave a very similar problem to what I had had on wantadilla in the afternoon: the headings were illegible:

image

I wasn't able to fix that. My best bet is that firefox is still thinking in “points” instead of pixels. A point is roughly 1/72", so a 10 point character is about 0.139", or about 1.5 pixels. Clearly this makes no sense. Tried Opera to see if that was any different. It was:

image

Again, no way to reset the values. It looks like I'll have to stop X and not tell it the dimensions of the screen. But what a mess! If this had been done in PostScript or PDF, there would have been no problem at all.


Top of page Previous month Greg's home page Today's diary entry Next month Greg's photos Copyright information

Valid XHTML 1.0!

$Id: diary-aug2005.php,v 1.50 2017/02/04 23:13:55 grog Exp $